Description
We investigate how the budget levels embedded in budget-based contracts affect individual effort and risk-taking. We
show that, from a wealth maximization perspective, a tradeoff exists between motivating effort and encouraging risktaking.
We illustrate an inverted-U relation between budget levels and effort. Budget levels and effort are positively correlated
until budgets become very difficult, at which point individuals ‘‘give up.” We illustrate an opposing, U-shaped,
relation between budget levels and risk-taking.
The e?ort and risk-taking e?ects of budget-based contracts
q
Geo?rey B. Sprinkle
a,
*
, Michael G. Williamson
b
, David R. Upton
c
a
Indiana University, Kelley School of Business, 1309 East Tenth Street, Bloomington, IN 47405, United States
b
The University of Texas at Austin, McCombs School of Business, 1 University Station B6400, Austin, TX 78712-0211, United States
c
University of North Carolina at Greensboro, Bryan School of Business and Economics, Greensboro, NC 27402-6165, United States
Abstract
We investigate how the budget levels embedded in budget-based contracts a?ect individual e?ort and risk-taking. We
show that, from a wealth maximization perspective, a tradeo? exists between motivating e?ort and encouraging risk-
taking. We illustrate an inverted-U relation between budget levels and e?ort. Budget levels and e?ort are positively cor-
related until budgets become very di?cult, at which point individuals ‘‘give up.” We illustrate an opposing, U-shaped,
relation between budget levels and risk-taking. Low budgets provide the ?exibility to take greater risks, whereas high
budgets induce individuals to ‘‘play it safe” to ensure budget attainment. Risky projects provide the greatest probability
of reaching very high (stretch) budgets. We conduct a laboratory experiment to empirically test this economic propo-
sition vis-a`-vis extant psychology research. Consistent with security-potential/aspiration theory, we ?nd that individuals
are willing to sacri?ce expected wealth to either meet the budget or increase their potential payo?s. Our results suggest
that the e?ort-risk tradeo? is mitigated at low budget levels, thereby increasing ?rm welfare, but is exacerbated at high
budget levels, reducing ?rm welfare. Collectively, our results highlight the importance of understanding how managerial
accounting practices such as budgets a?ect the various determinants of performance and not just performance per se.
Our results also help reconcile con?icting evidence regarding where budget di?culty levels should be set.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
In this paper, we investigate how the budget lev-
els embedded in budget-based contracts a?ect indi-
vidual e?ort and risk-taking. Budgeting systems
saturate all aspects of an organization’s architec-
ture, from partitioning decision rights to motivat-
ing, evaluating, and rewarding performance
(Baiman, 1990; Luft & Shields, 2003). Despite a
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.11.001
q
We thank Joel Demski, Joe Fisher, Laureen Maines, Jerry
Salamon, Mike Shields, Jim Wahlen, two anonymous review-
ers, and participants at the Indiana University Accounting
Research Workshop for their very helpful comments and
suggestions.
*
Corresponding author. Tel.: +1 812 855 3514; fax: +1 812
855 4985.
E-mail address: [email protected] (G.B. Sprinkle).
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 436–452
www.elsevier.com/locate/aos
large body of research examining the e?cacy of
budgets, particularly where budget levels should
be set to yield high performance, little evidence
exists as to how budget levels separately a?ect
the e?ort and risk-taking decisions of employees
(Ittner & Larcker, 2001).
Examining this issue is important for at least
two reasons. First, performance frequently is a
function of both e?ort and risk decisions. For
example, a manager of an investment center is
responsible for selecting projects that vary in risk
(capital budgeting decisions) and for implementing
these projects once they are selected. Second, bud-
get levels may di?erentially a?ect employees’ e?ort
and risk decisions. The budget level that motivates
high e?ort may di?er from the budget level that
encourages risk-taking.
We show that, from a wealth maximization per-
spective, a tradeo? exists between motivating e?ort
and encouraging risk-taking. We illustrate an
inverted-U relation between budget levels and
e?ort. Budget levels and e?ort are positively corre-
lated until budgets become very di?cult, at which
point e?ort drops substantially. We illustrate an
opposing, U-shaped, relation between budget lev-
els and risk-taking. Low budgets provide employ-
ees the ?exibility to allocate their e?ort to riskier
projects. High budgets induce employees to choose
lower-risk projects to ensure a high probability of
reaching the budget. However, riskier projects pro-
vide the greatest probability of reaching very high
(stretch) budgets.
We conduct a laboratory experiment to empiri-
cally test this economic proposition vis-a`-vis
extant psychology research. We ?nd that individu-
als evaluated and rewarded with a low budget sac-
ri?ced expected compensation to take on
additional risk. Consistent with security-poten-
tial/aspiration (SP/A) theory, participants work-
ing under the low budget were attracted by the
potentially high payo?s because budget attainment
was a fait accompli. This leads individuals to work
harder (beyond the point at which the expected
marginal bene?t equals the marginal cost of e?ort)
and select higher variance projects. In the context
of our setting, this result increases ?rm welfare.
In contrast, participants working under the
high budget sacri?ced expected compensation to
reduce risk. Consistent with SP/A theory, concerns
about meeting the budget increased risk aversion.
In turn, risk aversion leads individuals to exert less
e?ort (because the expected marginal bene?t
shrinks) and select ‘‘safe” projects. In the context
of our setting, both actions reduce ?rm welfare.
In essence, our experimental results suggest that
the e?ort and risk-taking tradeo? can be mitigated
at low budget levels but is exacerbated at high
budget levels.
Collectively, our results demonstrate that
increasing budget levels can have opposing e?ects
on two primary determinants of performance,
e?ort and risk.
1
The tradeo? we document under-
scores the importance of decomposing perfor-
mance into its fundamental determinants.
Managerial accounting controls, such as budgets,
that motivate organizationally desirable behavior
along one dimension of performance can simulta-
neously induce actions detrimental to the organi-
zation on another dimension of performance.
Our results also help reconcile the ?ndings of
prior studies examining the relation between bud-
get levels and performance. On the one hand,
experimental research suggests that challenging
budgets induce higher performance than easily
attainable budgets (see, e.g., Hirst & Yetton,
1999; Locke & Latham, 1990). Based primarily
on this evidence, popular management accounting
textbooks advocate evaluating employees based on
challenging budgets (see, e.g., Horngren, Datar, &
Foster, 2007, p. 183). On the other hand, archival
studies document that most organizations actually
use easily attainable budgets to evaluate and
reward employees (see, e.g., Merchant & Manzoni,
1989; Merchant & Van der Stede, 2007; Van der
Stede, 2000).
The di?erence in results may stem from the dif-
ferent research settings. The tasks used in prior
1
Scant attention has been focused on how managerial
accounting practices a?ect both e?ort and risk-taking (see,
e.g., Sprinkle, 2003). Prior research either examines how
accounting practices induce high levels of e?ort or confounds
the e?ort and risk-taking decisions. More generally, Stephen
Ross has stated, ‘‘No time has been spent on asking how
incentives a?ect the willingness of the employee to take on risk”
(Valance, 2001, p. 68).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 437
experimental studies are e?ort intensive (e.g., the
decoding of letters) and performance likely varies
little based on the risk-taking of participants. In
contrast, the archival/?eld studies were conducted
in environments where employees had more lati-
tude in project/task selection decisions. Our results
bridge these bodies of research – our ?ndings sug-
gest that high budgets may be most desirable when
employees have few decision rights and work in
non-stochastic, e?ort-intensive settings, whereas
low budgets are most desirable when both e?ort
and risk decisions are important to performance.
The remainder of this paper is organized into
four sections. The next section provides back-
ground and develops the hypothesis, and section
three explains the methods employed to test the
hypothesis. Section four presents the results, and
section ?ve provides a summary and discussion
of the results.
Background and hypothesis
Budget contract
Panel A of Fig. 1 illustrates the most common
budget-based contract in graphical form. Under
this contract, employees receive a ?xed bonus for
reaching an assigned budget and a linear piece-rate
for performance in excess of the budget (Murphy,
2001; Sprinkle & Williamson, 2004). If the
assigned budget is not achieved, then no bonus is
paid.
Because the bonus frequently is ?xed and not a
function of the budget level, employees desire a
low budget. As shown in Panel B of Fig. 1, this
allows employees to maximize their pay for any
given level of performance. For example, assume
an initial bonus of $500, a piece-rate of $5 for per-
formance above the budget target, and actual per-
formance of 80 units. If the assigned budget was 60
units, then the employee would receive a total
bonus of $600 = $500 + $5 Â (80–60 units). How-
ever, if the assigned budget was 40 units, then
the employee would receive a total bonus of
$700 = $500 + $5 Â (80 – 40).
In contrast, for any given level of perfor-
mance, employers desire a high budget because
they pay less for the same performance. This dif-
ference in preferences creates the basic tension
found under budget-based contracts. As we dis-
cuss next, it also has some interesting implica-
tions for employee e?ort and risk decisions.
2
We ?rst discuss the relation between budget lev-
els and e?ort.
Fig. 1. Budget-based contracts. Panel A – the relation between
pay and performance under the most common budget-based
contract. Panel B – the relation between pay and performance
at three budget levels.
2
While our discussion centers on budget-based contracts
where the initial bonus is not a function of the budget level, the
relations we discuss in the next two sections also hold for
budget-based contracts where the initial bonus is increasing in
the budget level.
438 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
Budget levels and e?ort
One way employees a?ect performance is
through their e?ort where, in general, ?rm returns
are increasing in e?ort. From an economic stand-
point, employees will work until the marginal cost
of their e?ort equals the marginal bene?t of the
higher pay that arises from higher e?ort. Panel A
of Fig. 2 illustrates the marginal cost and marginal
bene?t of e?ort at three budget levels.
As is typically assumed, the cost of e?ort
increases at an increasing rate. Notice that the
marginal bene?t of e?ort, pay, mirrors Panel B
of Fig. 1. For each budget level, the marginal ben-
e?t is $0 until the budget is achieved. Once
employees receive the initial bonus, the marginal
bene?t of each unit of additional performance
equals the piece-rate. At low budget levels,
employees have incentives to meet and exceed the
budget. They will work until the marginal cost of
their e?ort equals the piece-rate.
As the budget level increases, employee e?ort
also increases up to a point. This occurs because
of the initial bonus – employees will work harder
than they would under a low budget to receive
the ?xed bonus. In other words, increasing the
budget level shifts the point at which employees
receive the largest marginal bene?t, the initial
bonus. In turn, this increases the amount of e?ort
employees are willing to exert.
At some point, however, the budget becomes
too high, and the cost of reaching the budget
exceeds any monetary bene?ts accruing to the
employee. Intuitively, employees ‘‘give up” at very
high budget levels. Panel B of Fig. 2 documents
this inverted-U relation between budget and e?ort
levels. In essence, Fig. 2 helps explain the popular
prescription of ‘‘tight, but achievable” standards.
From an e?ort standpoint, it simply makes eco-
nomic sense for ?rms to set budgets as high as
employees are willing to take. In the next section,
we show that there is an opposite, or U-shaped,
relation between budget levels and risk.
Budget levels and risk
In addition to a?ecting performance via e?ort,
employees also a?ect performance via the choice
of tasks to which they allocate their e?ort (project
selection). Generally, ?rms desire their employees
to choose projects providing the highest expected
performance even though these projects are often
of higher variance (risk). However, because the
relationship between employee pay and perfor-
mance is non-linear (i.e., compensation spikes at
Fig. 2. Budget levels and e?ort. Panel A – marginal cost and
marginal bene?t of e?ort at three budget levels. Panel B – the
relationship between budget levels and e?ort. Panel A illustrates
the marginal cost and marginal bene?t of e?ort at three budget
levels. As is typically assumed, the cost of e?ort increases at an
increasing rate. For each budget level, the marginal bene?t of
e?ort is $0 until the budget is achieved. The marginal bene?t of
each unit of performance beyond the budget equals the piece-
rate. Panel B shows that as the budget level increases, e?ort also
increases up to a point – this occurs because of the initial bonus
– employees will increase their e?ort to receive this ?xed bonus.
At some point, however, the cost of reaching the budget
becomes too high and employees ‘‘give up.”
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 439
the budget level), higher risk projects with higher
expected performance do not always provide
employees the greatest expected pay. Below, we
discuss how the non-linearity in budget-based con-
tracts a?ects the relationship between budget levels
and risk-taking.
Panel A of Fig. 3 illustrates the output distribu-
tions for two projects, along with three budget lev-
els. Project 1 has a higher variance, but also a
higher mean, than project 2. Panel B of Fig. 3
shows the expected compensation for each project
at each budget level. Further, Exhibit 1 provides a
numerical example that illustrates the U-relation
between budget levels and risk-taking.
Notice that at the low budget level, employees
maximize expected compensation by choosing
project 1. Intuitively, this occurs because
employees are able to meet the budget with cer-
tainty by selecting either project. Thus, the pro-
ject with higher expected performance, project
Fig. 3. Budget levels and risk. Panel A – project outcome distributions and three budget levels. Panel B – budget levels and expected
pay for low- and high-risk projects. Panel C – the relation between budget levels and risk (project selection). Panel A illustrates the
output distributions of two projects, along with three budget levels. Project 1 has a higher variance, but also a higher mean, than
project 2. Panel B illustrates that at a low budget level, employees maximize their pay by choosing the higher risk, higher return project
(project 1). This occurs because employees are able to meet the budget with certainty by selecting either project – thus, they choose the
project with the higher expected performance. As the budget level increases, employees prefer project 2, the low risk project. This
occurs because project 2 allows them to meet the budget with certainty, whereas project 1 places their bonus in jeopardy. At very high
budget levels, the high risk project is the only way to meet the budget. Panel C documents this U-relation between budget levels and
risk (project selection).
440 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
1, provides the greatest expected compensation.
As the budget level increases, employees prefer
project 2 – this occurs because project 2 allows
them to meet the budget with certainty (and
attain the ?xed bonus) whereas project 1 places
the initial bonus in jeopardy. At very high bud-
get levels, project 1 is the only way to meet the
budget.
Panel C of Fig. 3 documents this U-Relation
between budget levels and risk (project selection).
This relation, which has not been addressed in
the budgeting literature, stands in stark contrast
to the relation between budget levels and e?ort.
It also illustrates a potential tradeo? between set-
ting budget levels to motivate high e?ort and
setting budget levels to encourage appropriate
risk-taking. This may help explain some of the
diverse results in the budgeting literature regarding
where budget levels should be set. Whether ?rms
prefer higher or lower budgets could depend on
whether e?ort or project selection is more impor-
tant to performance. Moreover, in settings where
both e?ort and project selection are equally impor-
tant to performance, a ?rm could be indi?erent
Exhibit 1.
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 441
between budget levels (after all, a U plus an
Inverted-U can average out to a straight line).
In the next section, we discuss psychology the-
ory that suggests the economic tradeo? between
e?ort and risk-taking may, in fact, be mitigated
at low budget levels but exacerbated at high bud-
get levels. In essence, organizations may get a ‘‘free
lunch” by setting low budgets.
3
Security-potential/aspiration (SP/A) theory
The prior discussion illustrates that, from an
individual wealth maximization standpoint,
increasing budget levels can have opposing e?ects
on a ?rm’s ability to motivate e?ort and encourage
risk-taking. Security-potential/aspiration (SP/A)
theory suggests that individuals will be willing to
sacri?ce expected wealth to either meet the budget
or increase their potential payo?s (i.e., the highest
possible payo?s). In contrast to conventional eco-
nomic theory that assumes risk-preferences are a
stable, innate characteristic, SP/A theory suggests
that risk-preferences would be a?ected by the bud-
get level in such a way as to promote more ?rm-
desired decisions at lower budget levels and less
desirable decisions at higher budget levels.
SP/A theory posits that individuals evaluate
alternatives in two stages when making decisions
in uncertain environments (see, e.g., Lopes, 1990;
Lopes, 1995; Lopes & Oden, 1999). In the ?rst
stage, individuals evaluate alternatives based on
the probability of meeting a minimum aspiration
level. In doing so, individuals behave in a risk-
averse manner, placing the greatest weight on out-
comes that do not meet their aspiration level. If
individuals cannot di?erentiate among alternatives
based on the ?rst criterion (e.g., multiple alterna-
tives meet the aspiration level with certainty), then,
in the second stage, individuals focus on potential
payo?s. Here, individuals behave in a risk-seeking
manner, placing the greatest weight on the highest
possible outcomes.
We are not aware of research that has examined
either the applicability or the implications of SP/A
theory in a budgeting context. However, individu-
als may adopt the budget as their aspiration level
for at least two reasons. First, individuals often
receive a substantial bonus for reaching the budget
target. Second, prior research suggests that goals
such as budgets have motivational properties inde-
pendent of their e?ects on compensation (e.g.,
Locke, Saari, Shaw, & Latham, 1981; Locke &
Latham, 1990). Moreover, employees may strive
to reach a budget simply because they have been
assigned a goal per se.
If employees do adopt budgets as their aspira-
tion level, then SP/A Theory could provide
insights into the relation between budget levels
and behavior. At low budget levels, individuals
can reach their budget with near certainty – thus,
individuals would tend to focus on the really good
payo?s. In turn, there are two ways to obtain these
really good payo?s. First, individuals can exert
more e?ort. Second, individuals can choose higher
variance projects.
4
Both activities are ‘‘risky” in
the sense that a higher variance project may not
pay o? (e.g., a low outcome could obtain). Addi-
tionally, working beyond the point at which mar-
ginal cost equals marginal bene?t can lead to a
good outcome but, by de?nition, is not a wealth
maximizing action. That said, such decisions are
preferred by the ?rm – organizations wish to elicit
as much e?ort as possible from their employees
3
We thank an anonymous Reviewer for this characterization
and help with relating security-potential/aspiration theory to
our study.
4
Projects that o?er higher levels of ?rm performance gener-
ally are of higher variance (risk). However, higher risk projects
with higher expected performance do not necessarily increase
employee remuneration. As discussed in the prior section, one
reason that these projects may not increase employee pay is the
non-linear relation between pay and performance under bud-
get-based contracts. Another reason is that higher risk projects
may have longer horizons whose returns and, in turn, e?ects on
employee pay, may not be realized for some time (Milgrom &
Roberts, 1992, Chapter 13). Moreover, in addition to alleviat-
ing agency concerns in a single-period setting, low budgets also
could alleviate intertemporal agency concerns associated with
the di?ering horizons of employees and organizations. This
argument is consistent with prior research demonstrating that
employees working under low budgets spend a greater amount
of time on projects whose performance e?ects will not be known
for over a year (Van der Stede, 2000).
442 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
and have them choose projects with the highest
expected return to the ?rm.
At higher budget levels, individuals cannot
reach their budget with certainty. Consequently,
SP/A theory suggests that individuals will place
greater weight on the worst possible outcomes
when evaluating their options (i.e., they will
become more risk-averse). Here, employees
would select projects with the highest probability
of meeting the budget – this project may or may
not be the one with the highest expected return.
Moreover, risk aversion shrinks the expected
marginal bene?t of e?ort leading, of course, to
less e?ort.
5
Both activities expose the employee
to less risk, but, naturally, are not preferred by
the ?rm.
Fig. 4 illustrates the implications of SP/A the-
ory vis-a`-vis expected wealth maximization. Panel
A illustrates that e?ort would be higher under a
low budget and lower under a high budget than
would be predicted under expected wealth maximi-
zation. Panel B shows that employees’ propensity
to select riskier projects would be higher under a
low budget and lower under a higher budget than
would be predicted under expected wealth maximi-
zation. Moreover, a ?rm would likely prefer a low
budget to a high budget in settings where both
e?ort and project selection are equally important
to performance, rather than being indi?erent as
suggested by an expected wealth maximization
perspective.
In sum, SP/A theory suggests that low budgets
may lead individuals to behave in a risk-seeking
manner, whereas high budgets may lead individu-
als to behave in a risk-averse manner. This leads to
the following hypothesis to be tested:
Hypothesis. Employees rewarded with a low
(high) budget behave in a more risk-seeking
(risk-averse) manner when making e?ort and
project selection decisions relative to employees
rewarded with a high (low) budget.
Experimental method
Participants and design
Sixty undergraduate business students volun-
teered to participate in the experiment. Partici-
5
In other words, if employees exhibit greater risk aversion
when they cannot reach their budget with certainty, then they
would discount the potential bene?ts of meeting and exceeding
the budget. In turn, this makes it more likely that the costs of
reaching the budget would exceed the uncertain bene?ts.
Fig. 4. SP/A theory versus expected wealth maximization.
Panel A – SP/A theory versus expected wealth maximization.
The relation between budget levels and e?ort. Panel B – SP/A
theory versus expected wealth maximization: The relation
between budget levels and risk. This ?gure compares security-
potential/aspiration (SP/A) theory to expected wealth maximi-
zation. Panel A compares the relation between budget levels
and e?ort, while Panel B compares the relation between budget
levels and risk decisions.
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 443
pants were randomly assigned to one of two exper-
imental conditions. We obtained these two condi-
tions by manipulating a single factor, the budget
level used to reward participants, at two levels
(low and high) between subjects. The experiment
consisted of two primary activities. In the ?rst
activity, participants worked on a time-intensive
task. In the second activity, participants chose to
play one of twelve possible lotteries that varied
in risk and expected return. This lottery choice
represented participants’ e?ort and project selec-
tion decisions. In the remainder of this section,
we describe the twelve lotteries and the experimen-
tal procedures.
The lotteries representing the e?ort and project
selection decisions
The twelve lotteries representing the e?ort and
project selection decisions for the low- and high-
budget conditions are described in Panels A and
B of Fig. 5, respectively. The twelve lotteries are
presented in a three row by four column matrix.
The lotteries are described by six pieces of infor-
mation. However, participants were only provided
with the ?rst four pieces of information. We dis-
cuss the information in each cell below.
For each lottery, the four pieces of infor-
mation provided to participants summarized
one-hundred equally likely states of nature. The
speci?c payo? function for each lottery was con-
sistent with the budget-based contract depicted
in Fig. 1. Participants received a low ?at payo?
if the state of nature from their selected lottery
was below their budget and a higher (linearly
increasing) payo? if the state of nature was
above their budget. Thus, knowing (1) the prob-
ability of meeting the budget, (2) the payo? if
the budget was not met, (3) the lowest payo?
if the budget was met, and (4) the highest payo?
if the budget was met (each amount between the
low and high payo? was equally likely) provided
the information needed to describe the 100 states
of nature in each cell.
The organization of the twelve cells in the
matrix is consistent with a multiplicative produc-
tion function – output equals participant’s e?ort
choice multiplied by a state variable drawn from
the project selected.
6
E?ort choices are represented
by rows, and project selection decisions are repre-
sented by columns. For the reasons discussed in
the prior section, we assume that increasing e?ort
(moving from row A to row B to row C) and
selecting riskier projects (by moving from left to
right across the columns) is bene?cial to the ?rm.
Below, we describe how the e?ort and project
selection decisions a?ect participants’ payo?s.
The payo?s are structured such that increasing
e?ort (moving from row A to row B to row C)
has three e?ects on the lotteries. First, the amount
received for not reaching the budget decreases
because e?ort is costly (the payment for not reach-
ing the budget represents the ?xed wage paid to
the employee less the cost of e?ort). Second, the
probability of reaching the budget either increases
or stays the same as employee e?ort increases.
Third, the variance of the payo?s increases.
7
By moving from left to right across the col-
umns, participants increase the variance of the
selected project. Choosing projects with higher
variances has the following two e?ects on the lot-
teries. First, the probability of reaching the budget
either decreases or stays the same. Second, the var-
iance of the payo?s increases.
Although we organized the cells so that the pay-
o? distributions are consistent with that of a mul-
tiplicative production function, our primary goal
in building the matrices was to create a setting
where we could observe how budget levels a?ect
the propensity of participants to trade-o? expected
compensation for risk-taking. Each matrix had
one cell (B2) that o?ered participants the highest
expected compensation ($17.25). Participants
could sacri?ce expected compensation in $0.25
increments by choosing one of the three cells with
increasingly lower variances (B1, A2, and A1).
These cells, labeled ‘‘risk-averse cells” in Fig. 5,
allowed participants to meet the budget with cer-
6
Both the tensions captured and our predictions would be
similar if we used an additive production function, where
output equals e?ort plus a state drawn from a probability
distribution, to build our lottery matrices.
7
We represent projects as uniform distributions. Naturally,
when these distributions are multiplied by larger e?ort num-
bers, the range (variance) of output and payo?s increases.
444 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
Panel A – The Effort and Project Selection Matrix for Participants inthe Low-Budget Condition
Panel B – The Effort and Project Selection Matrix for Participants in the High-Budget Condition
1 2 3 4
Probability of meeting the budget 100% 100% 65% 55%
Amount paid if the budget is not met NA NA $3.50 $3.50
Lowest amount paid if the budget is met $16.25 $16.00 $15.00 $13.00
Highest amount paid if the budget is met $16.75 $17.50 $20.00 $43.00
Cell Type Risk-Averse Risk-Averse Demo Risk-Seeking
Expected Compensation
A
$16.50 $16.75 $12.60 $16.98
Probability of meeting the budget 100% 100% 100% 60%
Amount paid if the budget is not met NA NA NA $2.50
Lowest amount paid if the budget is met $15.75 $14.75 $7.50 $6.00
Highest amount paid if the budget is met $18.25 $19.75 $26.00 $46.50
Cell Type Risk-Averse High EV Potential Risk-Seeking
Expected Compensation
B
$17.00 $17.25 $16.75 $16.75
Probability of meeting the budget 100% 100% 100% 65%
Amount paid if the budget is not met NA NA NA $0.00
Lowest amount paid if the budget is met $10.50 $10.00 $5.00 $2.00
Highest amount paid if the budget is met $18.50 $24.00 $28.00 $48.75
Cell Type Demo Potential Potential Risk-Seeking
Expected Compensation
C
$14.50 $17.00 $16.50 $16.49
1 2 3 4
Probability of meeting the budget 100% 100% 65% 55%
Amount paid if the budget is not met NA NA $3.50 $3.50
Lowest amount paid if the budget is met $16.25 $16.00 $15.00 $13.00
Highest amount paid if the budget is met $16.75 $17.50 $20.00 $43.00
Cell Type Risk-Averse Risk-Averse Demo Risk-Seeking
Expected Compensation
A
$16.50 $16.75 $12.60 $16.98
Probability of meeting the budget 100% 70% 65% 60%
Amount paid if the budget is not met NA $2.50 $2.50 $2.50
Lowest amount paid if the budget is met $15.75 $14.75 $7.50 $6.00
Highest amount paid if the budget is met $18.25 $32.40 $41.35 $46.50
Cell Type Risk-Averse High EV Potential Risk-Seeking
Expected Compensation
B
$17.00 $17.25 $16.75 $16.75
Probability of meeting the budget 100% 75% 70% 65%
Amount paid if the budget is not met NA $0.00 $0.00 $0.00
Lowest amount paid if the budget is met $10.50 $10.00 $5.00 $2.00
Highest amount paid if the budget is met $18.50 $35.35 $42.15 $48.75
Cell Type Demo Potential Potential Risk-Seeking
Expected Compensation
C
$14.50 $17.00 $16.50 $16.49
Fig. 5. E?ort and project selection matrixes for the low- and high-budget conditions. Panel A – the e?ort and project selection
matrix for participants in the low budget condition This ?gure presents the parameters used for our experiment’s low- and high-
budget conditions. Participants received the ?rst four pieces of information in each cell, which describe one-hundred states of
nature for competing lotteries. One lottery had the highest expected compensation (High EV). The three risk-averse lotteries had
lower expected values and lower variances. The potential and risk-seeking lotteries had lower expected values and higher variances.
The demo lotteries were used in the examples embedded within participants’ experimental instructions. The High Expected Value
and Potential lotteries met the budget with certainty only in the low budget condition. The payo?s are structured such that
increasing e?ort (moving from row A to row B to row C) has three e?ects on the lotteries. First, the amount received for not
reaching the budget decreases because participants pay an increasing cost for exerting higher levels of e?ort (the payment for not
reaching the budget represents the ?xed wage paid to the employee less the cost to the employee for exerting e?ort). Second, the
probability of reaching the budget either increases or stays the same as e?ort increases. Third, the variance of the payo?s increases.
Project selection choices are represented by columns. By moving from left to right across the columns, participants increase the
variance of the selected project. Choosing projects with higher variances has the following two e?ects on the payo? distributions of
the lotteries. First, the probability of reaching the budget either decreases or stays the same. Second, the variance of the payo?s
increases.
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 445
tainty. Participants also could sacri?ce expected
compensation in approximately $0.25 increments
by choosing cells with increasingly higher vari-
ances than the cell with the greatest expected com-
pensation (A4, B4, and C4). If participants chose
one of these three cells (labeled ‘‘risk-seeking
cells”), then they would meet the budget between
55% and 65% of the time. Finally, participants
could sacri?ce expected compensation in $0.25
increments by choosing one of the cells labeled
‘‘potential cells” (C2, B3, and C3). The potential
cells had higher variances than the expected value
cell but lower variances than the three risk-seeking
cells. We describe the potential cells in more detail
in the paragraphs that follow.
8
To create our low and high budget conditions,
we manipulated the probabilities of reaching the
budget in the highest expected compensation cell
and the three potential cells. Similar to the intui-
tion described in the background section, the high-
est expected compensation cell in the low budget
matrix allowed participants to reach the budget
with certainty. Further, the three potential cells
also allowed participants to reach the budget with
certainty. Thus, by choosing one of the potential
cells, participants could sacri?ce expected compen-
sation to take greater risk (increasing their poten-
tial for higher payo?s) while still reaching the
budget with certainty. Including both potential
and risk-seeking cells allows us to discriminate
between behavior consistent with SP/A theory,
where participants with low budgets choose higher
variance projects as long as they meet the budget
with certainty, and general risk-seeking behavior.
To the extent that SP/A theory is descriptive of
behavior, we would expect to see more individuals
choosing one of the potential cells and no more
individuals selecting one of the risk-seeking cells
in the low budget condition relative to the high
budget condition.
The highest expected compensation cell in the
high budget condition did not allow participants
to reach the budget with certainty. Further, partic-
ipants could not reach the budget with certainty by
taking greater risk (i.e., the potential cells did not
allow participants to reach the budget with cer-
tainty).
9
Thus, participants could only reach the
budget with certainty by sacri?cing expected com-
pensation to choose one of the cells with the lowest
variances (the risk-averse cells).
Experimental procedures
Upon arrival, we informed participants that
they would be completing several activities and
that the entire experiment would last approxi-
mately two hours. The instructions for each activ-
ity were read out loud, and all participants in each
experimental session worked at the same pace
(participants received one activity at a time).
In the ?rst activity, participants were provided
with sixty boxes each with ten rows and thirty-
one columns of random letters. Each box had a
single letter in its top-right hand corner (the search
letter). For each box, participants had to circle
every instance of the search letter and provide
the total in the top-right hand corner. Participants
learned that they would earn the right to partici-
pate in the next activity of the experiment by
working on the exercise for 45 minutes and provid-
ing acceptable totals (answers within one of the
correct answer) for at least ?fteen of the boxes.
10
At the conclusion of the 45 minute period, the
experimenter veri?ed that each participant met
the performance requirements.
8
Each matrix also has two cells (A3 and C1) that were
intentionally made unattractive to the participants. That is,
these cells were strictly dominated by other cells. We used these
two cells, labeled demo cells, for our examples in the experi-
mental instructions given to the participants.
9
To hold the expected compensation of these four cells
constant across the low and high budget conditions, we
increased the highest amount paid if the threshold is met
(increasing the slope of the linear cash payo?s above the budget
amount) in the high budget condition for these four cells.
10
Participants were informed that if they did not meet these
performance requirements then they would be paid $4 and
would be dismissed from the experiment. Further, participants
were informed that they could earn signi?cantly more money by
successfully completing this activity. All sixty participants
successfully completed the performance requirements.
446 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
The primary purpose of the ?rst activity was to
ensure participants earned the lottery choice that
they made in the second activity of the experiment.
When selecting among competing lotteries, prior
research demonstrates that participants make dif-
ferent decisions depending on whether they earn,
or are endowed with, their choice (Boylan & Sprin-
kle, 2001).
11
More generally, employees bear both
monetary (e.g., opportunity cost of foregone
wages) and non-monetary (e.g., opportunity cost
of time that could be used for leisure) costs in work
settings. Having participants actually spend time
at the experiment ensures that we have both ele-
ments in our study. In other words, our primary
interest was in examining the choices made by par-
ticipants who worked to earn their choice.
At the start of the second activity, participants
learned that they had earned the right to choose
one of twelve lotteries presented on the three row
by four column matrix described earlier and that
the outcome of the lottery would be paid to them
in cash at the end of the experiment. The instruc-
tions described each of the four pieces of informa-
tion, including the payment received for not
reaching the budget, the probability of reaching
the budget, and the variance of payments for out-
comes above the budget when moving across the
rows and down the columns.
12
Participants also
were provided with calculators and a sheet that
detailed each possible state of nature for the twelve
lotteries.
Participants studied the matrix and then
marked their cell choice on the sheet provided.
To determine the outcome of their chosen lottery,
participants drew a ball out of a bag containing
balls numbered 1 through 100. Based on this draw
and the cell chosen, participants calculated their
compensation which was veri?ed by the
experimenter.
Next, participants completed a post-experimen-
tal questionnaire. Finally, participants completed a
risk-preference instrument.
13
This instrument
required participants to make ?fteen choices
between receiving a certain amount of $2.50 and
participating in a lottery. The lotteries consisted
of a chance of winning $5 with a probability of p
and $0 with a probability 1 À p (p varied from
85% to 15% in 5% increments). Once participants
made their ?fteen choices, one of the ?fteen choices
was selected at random, the lottery was conducted,
and participants’ earnings were determined based
on their choice and the result of the lottery. Pay-
ments for activity two of the experiment and the
risk-preference instrument were totaled, and par-
ticipants were paid in cash and dismissed.
Results
Our analyses focus on the cell choices made by
participants in the second activity of our experi-
ment. Participants’ cell choices, by budget condi-
tion, are presented graphically in Fig. 6. The cells
on the graph are arranged by cell type (demo,
risk-averse, expected compensation, potential,
and risk-seeking). As expected, no participant
chose a demo cell type. Thus, we drop this cell type
from the remainder of the analyses. Further, no
participant chose the risk-averse cells A1 or A2.
That is, only the risk-averse cell with the highest
variance and expected compensation (B1) was cho-
sen by participants. All other cells, however, were
chosen by at least one participant. Because not all
cells were chosen by the participants and to
streamline the presentation of the results, we col-
lapse the cell choices of participants into their
respective cell types for our analyses. However,
our inferences remain the same when using the
disaggregated cell choices.
Table 1 presents participants’ cell choice fre-
quencies, aggregated by type, for the low- and
high-budget conditions. To test our hypothesis,
we compare the distributions of these frequencies
11
More generally, research demonstrates that individuals
make di?erent decisions depending on whether they earn, or
are endowed with a ‘‘property/decision right” (see, e.g.,
Ho?man & Spitzer, 1985; Ho?man, McCabe, Shachat, &
Smith, 1994).
12
Budgets were described as thresholds in the experimental
instructions.
13
The risk preference instrument is a modi?ed version of the
instrument used in Boylan and Sprinkle (2001).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 447
across budget conditions. In the low budget condi-
tion, we are interested in whether a disproportion-
ate number of participants chose a potential cell,
thereby sacri?cing expected compensation to
increase potential payo?s (i.e., they behaved in a
more risk-seeking manner). In the high-budget
condition, we are interested in whether a dispro-
portionate number of participants chose a risk-
0
2
4
6
8
10
12
14
Choice
F
r
e
q
u
e
n
c
y
Low 0 0 0 0 0 10 2 9 3 2 3 1
3 0 High 0 0 0 0 13 6 3 4 0 1
A3 Demo C1 Demo A1 RA A2 RA B1 RA B2 EV B3 P C2 P C3 P A4 RS B4 RS C4 RS
Fig. 6. Participants’ cell choices in the low- and high-budget conditions. This graph shows participants’ cell choices in the low-and
high-budget conditions. The cells, which are described in Fig. 5, are grouped by the following cell types: Demo (demonstration cells),
RA (risk-averse cells), EV (highest expected compensation), P (potential cells), and RS (risk-seeking cells).
Table 1
Participants’ cell choices in low- and high-budget conditions
Choice
a
Budget level
Low budget High budget
n
b
%
c
Adj. Res.
d
n
b
%
c
Adj. Res.
d
Risk-averse 0 0.0 À4.1
**
13 43.3 4.1
**
High EV 10 33.3 À1.2 6 20.0 1.2
Potential 14 46.7 À1.9
*
7 23.3 1.9
*
Risk-seeking 6 20.0 À0.7 4 13.3 0.7
v
2
(3, N = 60) = 16.73, p < 0.01
a
This column lists the possible lottery categories chosen by at least one participant during the second activity of the experiment. The
High EV choice has the greatest expected compensation. The risk-averse choices have lower expected values and variances than the
High EV choice. The potential and risk-seeking choices have lower expected values and higher variances than the High EV choice. In
the low budget condition, the High EV and Potential choices meet the budget with certainty. In the high-budget condition, these
choices do not meet the budget with certainty.
b
The number of participants that chose the lottery category.
c
The percentage of participants that chose the lottery category.
d
The observed cell frequency less the expected cell frequency assuming frequencies are identically distributed across budget level
conditions. The results of tests to determine whether the residual is signi?cantly di?erent from zero:
**
p < 0.01;
*
p = 0.06.
448 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
averse cell, thereby sacri?cing expected compensa-
tion to increase the probability of meeting the bud-
get (i.e., they behaved in a more risk-averse
manner).
The relative distributions of cell type frequen-
cies across the low- and high-budget conditions
support our hypothesis. Thirteen of the 30 partic-
ipants in the high-budget condition chose a risk-
averse cell, but no participants in the low budget
condition did so. Further, 14 of the 30 participants
in the low budget condition chose a potential cell,
while 7 of 30 chose one of these cells in the high-
budget condition. Moreover, we do not observe
large di?erences in the number of participants
choosing the highest expected compensation or
risk-seeking cells across budget conditions.
To formally test our hypothesis, we used an
independent samples Chi-square analysis to com-
pare the relative distribution of cell type frequen-
cies across the low- and high-budget conditions.
This test shows that the distributions are statisti-
cally di?erent (v
2
= 16.73, p < 0.01). To identify
the cell types driving the signi?cant result, we
examine the residual of each cell (the observed cell
frequency less the expected cell frequency given
these frequencies are identically distributed across
our budget conditions). Speci?cally, we examine
which of the cell type’s residuals adjusted for their
variances (hereafter adjusted residuals) are signi?-
cantly larger than expected.
14
The adjusted residual of À4.1 (4.1) for the risk-
averse cell type in the high (low) budget condition
is signi?cantly larger than expected (p < 0.01).
Thus, signi?cantly more (fewer) individuals chose
a risk-averse cell than expected if the cell type fre-
quencies were identically distributed across our
budget conditions. Similarly, the adjusted residual
of 1.9 (À1.9) for the potential cell type in the high
(low) budget condition is signi?cantly larger than
expected (p = 0.057). Thus, signi?cantly fewer
(more) individuals chose a potential cell than
expected if the cell type frequencies were identi-
cally distributed across budget conditions. How-
ever, the adjusted residuals for the expected
compensation and risk-seeking cell types are not
signi?cant (p = 0.23 and p = 0.48, respectively).
In essence, the results support our hypothesis.
Participants in the low budget condition sacri?ced
expected compensation to take on additional risk.
Participants in the high-budget condition, on the
other hand, sacri?ced expected compensation to
reduce risk.
15
We also examined whether participants’ dispo-
sition toward risk a?ected our results. To accom-
plish this, we partitioned our sample into
participants with high and low risk tolerances
based on their responses to the risk-preferences
instrument. As previously discussed, this instru-
ment required participants to make ?fteen choices
between receiving a certain amount of $2.50 and
participating in a lottery. Each lottery consisted
of a chance of winning $5 with a probability of p
and $0 with a probability of 1 À p. The probability
of winning $5 in the ?rst lottery was 85%. We
decreased the chance of winning by 5% in each
subsequent lottery, such that the chance of win-
ning $5 in the ?nal lottery was 15%. We recorded
the number of the choice where individuals pre-
ferred the certain amount of $2.50 to the lottery.
The lower the score on this measure, the lower
the individual’s tolerance for risk. The average
score on this measure was 6.83, and means did
14
The standardized residual is given by the following formula
(n
ij
À E
ij
)/vE
ij
where n
ij
is actual cell frequency and E
ij
is the
expected cell frequency [(n
i.
 n
.j
)/N]. We adjust the standard-
ized residual by dividing it by its variance estimated by (1 À n
i.
/
N) Â (1 À n
.j
/N) where n
i.
is the total observations in row i, n
.j
is
the total observations in column j, and N is the total
observations in the experiment. Because the variables forming
our contingency table are independent, these adjusted residuals
are approximately normally distributed with a mean of 0 and a
standard deviation of 1 (Everitt, 1977, p. 47).
15
Consistent with SP/A theory, the post-experimental ques-
tionnaire provides evidence that participants aspired to meet
their budget. Participants were asked to answer the following
statement: ‘‘Meeting the budget was important to me”. Partic-
ipants responded using a seven-point Likert scale with ‘‘1”
being ‘‘strongly disagree,” ‘‘4” being ‘‘neither agree nor
disagree,” and ‘‘7” being ‘‘strongly agree.” Participants mean
response, 5.32, was signi?cantly greater than the midpoint of
the scale (t = 6.14, p < 0.01, two-tailed). Further, mean
responses did not statistically di?er across the low- and high-
budget conditions (t = 1.17, p = 0.28, two-tailed).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 449
not statistically di?er across the low and high bud-
get conditions (t = 0.24, p = 0.81, two-tailed).
We partitioned our sample into the 27 partici-
pants with scores on the risk-preferences instru-
ment lower than the mean (the low risk-tolerance
group), and the 33 participants with scores larger
than the mean (the high risk-tolerance group). In
Table 2, we report the results of separate v
2
anal-
yses for each of these groups.
Panel A of Table 2 presents the results for the
subset of participants with low risk tolerances.
Overall, the cell choices of this subset of partici-
pants di?ered across the low and high budget con-
ditions (v
2
= 12.08, p < 0.01). Further, the
distribution of choices di?ered in a manner consis-
tent with our results on the entire sample. Relative
to participants in the high-budget condition, sig-
ni?cantly fewer participants in the low budget con-
dition chose a risk-averse cell (p < 0.01) and
signi?cantly more chose a potential cell (p < 0.05).
Panel B of Table 2 presents the results for the
subset of participants with higher risk tolerances.
The cell choices of this subset of participants mar-
ginally di?ered across the low- and high-budget
conditions (v
2
= 6.27, p = 0.10). Relative to partic-
ipants in the high budget condition, signi?cantly
fewer participants in the low budget condition
chose a risk-averse cell (p < 0.01). However, par-
ticipants choices did not di?er for potential cells
(p = 0.48).
Collectively, these results suggest that SP/A the-
ory is more descriptive of behavior in individuals
with low risk tolerances. Since individuals in work
settings are presumed to be risk-averse, this sug-
Table 2
Choices of participants in low and high budget conditions partitioned by risk tolerance
a
Choice
b
Budget level
Low budget High budget
n
c
%
d
Adj. Res.
e
n
c
%
d
Adj. Res.
e
Panel A – Analysis for the subset of participants with low risk tolerances
Risk-averse 0 0.0 À3.4
*
7 58.3 3.4
*
High EV 4 26.7 0.6 2 16.7 À0.6
Potential 8 53.3 2.0
*
2 16.7 À2.0
*
Risk-seeking 3 20.0 0.8 1 8.3 À0.8
v
2
(3, N = 27) = 12.08, p < 0.01
Panel B – Analysis for the subset of participants with high risk tolerances
Risk-averse 0 0.0 À2.5
*
6 33.3 2.5
*
High EV 6 40.0 1.1 4 22.2 À1.1
Potential 6 40.0 0.7 5 27.8 À0.7
Risk-seeking 3 20.0 0.2 3 16.7 À0.2
v
2
(3, N = 33) = 6.27, p = 0.10
a
Participants were partitioned into high and low risk tolerance groups based on their responses to a risk-preferences instrument. This
instrument required participants to make ?fteen choices between receiving a certain amount of $2.50 and a lottery. Each lottery
consisted of a chance of winning $5 with probability p and $0 with probability 1 À p. The probability of winning $5 in the ?rst lottery
was 85%. We decreased the chance of winning p in subsequent lotteries by 5%, such that the chance of winning $5 in the ?nal lottery
was 15%. We recorded the number of the choice where each individual preferred the certain amount of $2.50 to the lottery. We
partitioned participants with a score lower (higher) than the mean score into the low (high) risk tolerance group.
b
This column lists the possible lottery categories chosen by at least one participant during the second activity of the experiment. The
High EV choice has the greatest expected compensation. The risk-averse choices have lower expected values and variances than the
High EV choice. The potential and risk-seeking choices have lower expected values and higher variances than the High EV choice. In
the low budget condition, the High EV and Potential choices meet the budget with certainty. In the high-budget condition, these
choices do not meet the budget with certainty.
c
The number of participants that chose the lottery category.
d
The percentage of participants that chose the lottery category.
e
The observed cell frequency less the expected cell frequency assuming frequencies are identically distributed across budget level
conditions. The results of tests to determine whether the residual is signi?cantly di?erent from zero:
*
p < 0.05.
450 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
gests that SP/A Theory may have far-reaching
implications.
Discussion
In this paper, we examine how budget-based
contracts a?ect e?ort and risk-taking. We show
that, from a wealth maximization perspective, a
tradeo? exists between motivating e?ort and
encouraging risk-taking. We illustrate an
inverted-U relationship between budget levels
and e?ort. Budget levels and e?ort are positively
correlated until budgets become very di?cult, at
which point e?ort drops substantially. We illus-
trate an opposing, U-shaped, relationship between
budget levels and risk-taking. Low budgets pro-
vide employees the ?exibility to allocate their e?ort
to riskier projects. High budgets induce employees
to choose lower-risk projects to ensure a high
probability of reaching the budget. Riskier pro-
jects provide the greatest probability of reaching
very high (stretch) budgets.
We conduct a laboratory experiment to empiri-
cally test this economic proposition vis-a`-vis
extant psychology research. We ?nd that individu-
als evaluated and rewarded with a low budget
exhibited a greater propensity to sacri?ce expected
compensation to take on additional risk. Consis-
tent with security-potential/aspiration (SP/A) the-
ory, participants working under the low budget
were attracted by the potentially high payo?s
because budget attainment was a fait accompli.
This leads individuals to work harder (beyond
the point at which expected marginal bene?ts
equal the marginal cost of e?ort) and select higher
variance projects. In the context of our setting, this
result is desired from the ?rm’s perspective and
increases ?rm welfare.
In contrast, participants working under the
high budget exhibited a greater propensity to sac-
ri?ce expected compensation to reduce risk. Con-
sistent with SP/A theory, concerns about meeting
the budget increased risk aversion. In turn, risk
aversion leads individuals to exert less e?ort
(because the expected marginal bene?t shrinks)
and select ‘‘safe” projects. In the context of our
setting, both actions reduce ?rm welfare. In
essence, our experimental results suggest that the
e?ort and risk-taking tradeo? illustrated in our
example can be mitigated at low budget levels
but exacerbated at high budget levels.
Collectively, our results have several important
implications. First, they demonstrate that account-
ing-based performance evaluation and reward sys-
tems can have opposing e?ects on two primary
determinants of performance, e?ort and risk-tak-
ing. Control systems that motivate organization-
ally desirable behavior along one dimension of
performance can simultaneously induce actions
detrimental to the organization on another dimen-
sion of performance.
Second, these tradeo?s illustrate the importance
of decomposing performance into its fundamental
determinants. When designing control systems, for
example, organizations need to orchestrate the
assignment of decision rights with the manner in
which performance is measured and rewarded
(Brickley, Smith, & Zimmerman, 2001). Thus,
the use of high budget levels may be desirable
when employees have few decision rights and/or
work in non-stochastic settings, whereas the use
of lower budget levels may be desirable when
employees have more latitude in project/task selec-
tion decisions and/or work in environments char-
acterized by uncertainty.
Third, the speci?c trade-o? we document
between e?ort and risk-taking helps to reconcile
the mixed results of prior research. On one hand,
experimental research tends to ?nd that challeng-
ing budgets induce the highest levels of perfor-
mance (see, e.g., Locke & Latham, 1990). Based
primarily on this evidence, popular management
accounting textbooks advocate evaluating employ-
ees based on challenging budgets (Horngren et al.,
2007, p. 183). However, the tasks used in prior
experimental studies are often e?ort intensive
(e.g., the decoding of letters) and performance
likely varies little based on the risk-taking of par-
ticipants. On the other hand, prior archival studies
suggest that organizations tend to utilize lower
budget levels to evaluate employees. These studies
were conducted in environments where employees
had more latitude in project/task selection deci-
sions (Merchant & Manzoni, 1989; Simon, 1988;
Van der Stede, 2000).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 451
Certain features of our study may limit the gen-
eralizability of our ?ndings. These limitations sug-
gest a number of avenues for further inquiry. For
example, we examined a single-period setting.
Most employment relationships extend over multi-
ple periods, and it is possible that learning and
wealth e?ects would, over time, lead individuals
to maximize expected compensation. Additionally,
we examined a single-person setting and our
results may not extend to multiperson settings
where gains to trade must be split – equity and
fairness concerns may lead individuals to adopt
di?erent aspiration levels and, in turn, make di?er-
ent e?ort and risk decisions. Finally, we only
examine the e?ect of one type of incentive, bud-
get-based contracts. Future research could exam-
ine whether other incentive schemes (e.g.,
tournament or piece-rate), or combinations and
dimensions thereof, induce individuals to take
appropriate levels of risk while concurrently moti-
vating high levels of e?ort.
References
Baiman, S. (1990). Agency research in managerial accounting:
A second look. Accounting, Organizations and Society, 15,
341–371.
Boylan, S. J., & Sprinkle, G. B. (2001). Experimental evidence
on the relation between tax rates and compliance: The e?ect
of earned vs. endowed income. The Journal of the American
Taxation Association, 23, 75–90.
Brickley, J. A., Smith, C. W., Jr., & Zimmerman, J. L. (2001).
Managerial economics and organizational architecture. New
York, NY: The McGraw-Hill Companies Inc.
Everitt, B. S. (1977). The analysis of contingency tables. New
York, NY: John Wiley and Sons Inc.
Hirst, M. K., & Yetton, P. W. (1999). The e?ects of budget
goals and task interdependence on the level of and variance
in performance: A research note. Accounting, Organizations
and Society, 24, 205–216.
Ho?man, E., McCabe, K., Shachat, K., & Smith, V. (1994).
Preferences, property rights, and anonymity in bargaining
games. Games and Economic Behavior, 7, 346–380.
Ho?man, E., & Spitzer, M. L. (1985). Entitlements, rights, and
fairness: An experimental examination of subjects’ concepts
of distributive justice. Journal of Legal Studies, 14, 259–297.
Horngren, C. T., Datar, S. M., & Foster, G. (2007). Cost
accounting: a managerial emphasis. Upper Saddle River, NJ:
Prentice Hall.
Ittner, C. D., & Larcker, D. F. (2001). Assessing empirical
research in managerial accounting: A value-based manage-
ment perspective. Journal of Accounting and Economics, 32,
349–410.
Locke, E. A., & Latham, G. P. (1990). A theory of goal setting
and task performance. Englewood Cli?s, NJ: Prentice Hall.
Locke, E. A., Saari, L., Shaw, K., & Latham, G. (1981). Goal
setting and task performance: 1969–1980. Psychological
Bulletin, 1, 125–152.
Lopes, L. (1990). Re-modeling risk aversion: A comparison of
Bernoullian and rank dependent value approaches. In G. M.
Von Furstenberg (Ed.), Acting under uncertainty: Multidis-
ciplinary conceptions (pp. 267–299). Boston, MA: Kluwer
Academic Publishers.
Lopes, L. (1995). On modeling risk choice. In Contributions to
decision making I (pp. 29–50). New York, NY: Elsevier.
Lopes, L., & Oden, G. (1999). The role of aspiration level in
risky choice: A comparison of cumulative prospect theory
and SP/A theory. Journal of Mathematical Psychology, 43,
286–313.
Luft, J., & Shields, M. D. (2003). Mapping management
accounting: Graphics and guidelines for theory-consistent
empirical research. Accounting, Organizations and Society,
28, 169–249.
Merchant, K. A., & Manzoni, J. F. (1989). The achievability of
budget targets in pro?t centers: A ?eld study. The Account-
ing Review, 3, 539–558.
Merchant, K. A., & Van der Stede, W. A. (2007). Management
control systems: Performance measurement, evaluation, and
incentives. London, UK: Prentice Hall.
Milgrom, P. R., & Roberts, J. (1992). Economics, organization
& management. Upper Saddle River, NJ: Prentice Hall.
Murphy, K. (2001). Performance standards in incentive con-
tracts. Journal of Accounting and Economics, 30, 245–278.
Simon, R. (1988). Analysis of the organizational characteristics
related to tight budget goals. Contemporary Accounting
Research, 5, 267–283.
Sprinkle, G. B. (2003). Perspectives on experimental research in
managerial accounting. Accounting, Organizations and Soci-
ety, 28, 287–318.
Sprinkle, G. B., & Williamson, M. G. (2004). The Evolution
from taylorism to employee gainsharing: A case study
examining John Deere’s continuous improvement pay plan.
Issues in Accounting Education, 19, 487–503.
Valance, N. (2001). Bright minds, bigtheories. CFO(January), 68.
Van der Stede, W. A. (2000). The relationship between two
consequences of budgetary controls: Budgetary slack crea-
tion and managerial short-term orientation. Accounting,
Organizations and Society, 25, 609–622.
452 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
doc_821260073.pdf
We investigate how the budget levels embedded in budget-based contracts affect individual effort and risk-taking. We
show that, from a wealth maximization perspective, a tradeoff exists between motivating effort and encouraging risktaking.
We illustrate an inverted-U relation between budget levels and effort. Budget levels and effort are positively correlated
until budgets become very difficult, at which point individuals ‘‘give up.” We illustrate an opposing, U-shaped,
relation between budget levels and risk-taking.
The e?ort and risk-taking e?ects of budget-based contracts
q
Geo?rey B. Sprinkle
a,
*
, Michael G. Williamson
b
, David R. Upton
c
a
Indiana University, Kelley School of Business, 1309 East Tenth Street, Bloomington, IN 47405, United States
b
The University of Texas at Austin, McCombs School of Business, 1 University Station B6400, Austin, TX 78712-0211, United States
c
University of North Carolina at Greensboro, Bryan School of Business and Economics, Greensboro, NC 27402-6165, United States
Abstract
We investigate how the budget levels embedded in budget-based contracts a?ect individual e?ort and risk-taking. We
show that, from a wealth maximization perspective, a tradeo? exists between motivating e?ort and encouraging risk-
taking. We illustrate an inverted-U relation between budget levels and e?ort. Budget levels and e?ort are positively cor-
related until budgets become very di?cult, at which point individuals ‘‘give up.” We illustrate an opposing, U-shaped,
relation between budget levels and risk-taking. Low budgets provide the ?exibility to take greater risks, whereas high
budgets induce individuals to ‘‘play it safe” to ensure budget attainment. Risky projects provide the greatest probability
of reaching very high (stretch) budgets. We conduct a laboratory experiment to empirically test this economic propo-
sition vis-a`-vis extant psychology research. Consistent with security-potential/aspiration theory, we ?nd that individuals
are willing to sacri?ce expected wealth to either meet the budget or increase their potential payo?s. Our results suggest
that the e?ort-risk tradeo? is mitigated at low budget levels, thereby increasing ?rm welfare, but is exacerbated at high
budget levels, reducing ?rm welfare. Collectively, our results highlight the importance of understanding how managerial
accounting practices such as budgets a?ect the various determinants of performance and not just performance per se.
Our results also help reconcile con?icting evidence regarding where budget di?culty levels should be set.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
In this paper, we investigate how the budget lev-
els embedded in budget-based contracts a?ect indi-
vidual e?ort and risk-taking. Budgeting systems
saturate all aspects of an organization’s architec-
ture, from partitioning decision rights to motivat-
ing, evaluating, and rewarding performance
(Baiman, 1990; Luft & Shields, 2003). Despite a
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.11.001
q
We thank Joel Demski, Joe Fisher, Laureen Maines, Jerry
Salamon, Mike Shields, Jim Wahlen, two anonymous review-
ers, and participants at the Indiana University Accounting
Research Workshop for their very helpful comments and
suggestions.
*
Corresponding author. Tel.: +1 812 855 3514; fax: +1 812
855 4985.
E-mail address: [email protected] (G.B. Sprinkle).
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 436–452
www.elsevier.com/locate/aos
large body of research examining the e?cacy of
budgets, particularly where budget levels should
be set to yield high performance, little evidence
exists as to how budget levels separately a?ect
the e?ort and risk-taking decisions of employees
(Ittner & Larcker, 2001).
Examining this issue is important for at least
two reasons. First, performance frequently is a
function of both e?ort and risk decisions. For
example, a manager of an investment center is
responsible for selecting projects that vary in risk
(capital budgeting decisions) and for implementing
these projects once they are selected. Second, bud-
get levels may di?erentially a?ect employees’ e?ort
and risk decisions. The budget level that motivates
high e?ort may di?er from the budget level that
encourages risk-taking.
We show that, from a wealth maximization per-
spective, a tradeo? exists between motivating e?ort
and encouraging risk-taking. We illustrate an
inverted-U relation between budget levels and
e?ort. Budget levels and e?ort are positively corre-
lated until budgets become very di?cult, at which
point e?ort drops substantially. We illustrate an
opposing, U-shaped, relation between budget lev-
els and risk-taking. Low budgets provide employ-
ees the ?exibility to allocate their e?ort to riskier
projects. High budgets induce employees to choose
lower-risk projects to ensure a high probability of
reaching the budget. However, riskier projects pro-
vide the greatest probability of reaching very high
(stretch) budgets.
We conduct a laboratory experiment to empiri-
cally test this economic proposition vis-a`-vis
extant psychology research. We ?nd that individu-
als evaluated and rewarded with a low budget sac-
ri?ced expected compensation to take on
additional risk. Consistent with security-poten-
tial/aspiration (SP/A) theory, participants work-
ing under the low budget were attracted by the
potentially high payo?s because budget attainment
was a fait accompli. This leads individuals to work
harder (beyond the point at which the expected
marginal bene?t equals the marginal cost of e?ort)
and select higher variance projects. In the context
of our setting, this result increases ?rm welfare.
In contrast, participants working under the
high budget sacri?ced expected compensation to
reduce risk. Consistent with SP/A theory, concerns
about meeting the budget increased risk aversion.
In turn, risk aversion leads individuals to exert less
e?ort (because the expected marginal bene?t
shrinks) and select ‘‘safe” projects. In the context
of our setting, both actions reduce ?rm welfare.
In essence, our experimental results suggest that
the e?ort and risk-taking tradeo? can be mitigated
at low budget levels but is exacerbated at high
budget levels.
Collectively, our results demonstrate that
increasing budget levels can have opposing e?ects
on two primary determinants of performance,
e?ort and risk.
1
The tradeo? we document under-
scores the importance of decomposing perfor-
mance into its fundamental determinants.
Managerial accounting controls, such as budgets,
that motivate organizationally desirable behavior
along one dimension of performance can simulta-
neously induce actions detrimental to the organi-
zation on another dimension of performance.
Our results also help reconcile the ?ndings of
prior studies examining the relation between bud-
get levels and performance. On the one hand,
experimental research suggests that challenging
budgets induce higher performance than easily
attainable budgets (see, e.g., Hirst & Yetton,
1999; Locke & Latham, 1990). Based primarily
on this evidence, popular management accounting
textbooks advocate evaluating employees based on
challenging budgets (see, e.g., Horngren, Datar, &
Foster, 2007, p. 183). On the other hand, archival
studies document that most organizations actually
use easily attainable budgets to evaluate and
reward employees (see, e.g., Merchant & Manzoni,
1989; Merchant & Van der Stede, 2007; Van der
Stede, 2000).
The di?erence in results may stem from the dif-
ferent research settings. The tasks used in prior
1
Scant attention has been focused on how managerial
accounting practices a?ect both e?ort and risk-taking (see,
e.g., Sprinkle, 2003). Prior research either examines how
accounting practices induce high levels of e?ort or confounds
the e?ort and risk-taking decisions. More generally, Stephen
Ross has stated, ‘‘No time has been spent on asking how
incentives a?ect the willingness of the employee to take on risk”
(Valance, 2001, p. 68).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 437
experimental studies are e?ort intensive (e.g., the
decoding of letters) and performance likely varies
little based on the risk-taking of participants. In
contrast, the archival/?eld studies were conducted
in environments where employees had more lati-
tude in project/task selection decisions. Our results
bridge these bodies of research – our ?ndings sug-
gest that high budgets may be most desirable when
employees have few decision rights and work in
non-stochastic, e?ort-intensive settings, whereas
low budgets are most desirable when both e?ort
and risk decisions are important to performance.
The remainder of this paper is organized into
four sections. The next section provides back-
ground and develops the hypothesis, and section
three explains the methods employed to test the
hypothesis. Section four presents the results, and
section ?ve provides a summary and discussion
of the results.
Background and hypothesis
Budget contract
Panel A of Fig. 1 illustrates the most common
budget-based contract in graphical form. Under
this contract, employees receive a ?xed bonus for
reaching an assigned budget and a linear piece-rate
for performance in excess of the budget (Murphy,
2001; Sprinkle & Williamson, 2004). If the
assigned budget is not achieved, then no bonus is
paid.
Because the bonus frequently is ?xed and not a
function of the budget level, employees desire a
low budget. As shown in Panel B of Fig. 1, this
allows employees to maximize their pay for any
given level of performance. For example, assume
an initial bonus of $500, a piece-rate of $5 for per-
formance above the budget target, and actual per-
formance of 80 units. If the assigned budget was 60
units, then the employee would receive a total
bonus of $600 = $500 + $5 Â (80–60 units). How-
ever, if the assigned budget was 40 units, then
the employee would receive a total bonus of
$700 = $500 + $5 Â (80 – 40).
In contrast, for any given level of perfor-
mance, employers desire a high budget because
they pay less for the same performance. This dif-
ference in preferences creates the basic tension
found under budget-based contracts. As we dis-
cuss next, it also has some interesting implica-
tions for employee e?ort and risk decisions.
2
We ?rst discuss the relation between budget lev-
els and e?ort.
Fig. 1. Budget-based contracts. Panel A – the relation between
pay and performance under the most common budget-based
contract. Panel B – the relation between pay and performance
at three budget levels.
2
While our discussion centers on budget-based contracts
where the initial bonus is not a function of the budget level, the
relations we discuss in the next two sections also hold for
budget-based contracts where the initial bonus is increasing in
the budget level.
438 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
Budget levels and e?ort
One way employees a?ect performance is
through their e?ort where, in general, ?rm returns
are increasing in e?ort. From an economic stand-
point, employees will work until the marginal cost
of their e?ort equals the marginal bene?t of the
higher pay that arises from higher e?ort. Panel A
of Fig. 2 illustrates the marginal cost and marginal
bene?t of e?ort at three budget levels.
As is typically assumed, the cost of e?ort
increases at an increasing rate. Notice that the
marginal bene?t of e?ort, pay, mirrors Panel B
of Fig. 1. For each budget level, the marginal ben-
e?t is $0 until the budget is achieved. Once
employees receive the initial bonus, the marginal
bene?t of each unit of additional performance
equals the piece-rate. At low budget levels,
employees have incentives to meet and exceed the
budget. They will work until the marginal cost of
their e?ort equals the piece-rate.
As the budget level increases, employee e?ort
also increases up to a point. This occurs because
of the initial bonus – employees will work harder
than they would under a low budget to receive
the ?xed bonus. In other words, increasing the
budget level shifts the point at which employees
receive the largest marginal bene?t, the initial
bonus. In turn, this increases the amount of e?ort
employees are willing to exert.
At some point, however, the budget becomes
too high, and the cost of reaching the budget
exceeds any monetary bene?ts accruing to the
employee. Intuitively, employees ‘‘give up” at very
high budget levels. Panel B of Fig. 2 documents
this inverted-U relation between budget and e?ort
levels. In essence, Fig. 2 helps explain the popular
prescription of ‘‘tight, but achievable” standards.
From an e?ort standpoint, it simply makes eco-
nomic sense for ?rms to set budgets as high as
employees are willing to take. In the next section,
we show that there is an opposite, or U-shaped,
relation between budget levels and risk.
Budget levels and risk
In addition to a?ecting performance via e?ort,
employees also a?ect performance via the choice
of tasks to which they allocate their e?ort (project
selection). Generally, ?rms desire their employees
to choose projects providing the highest expected
performance even though these projects are often
of higher variance (risk). However, because the
relationship between employee pay and perfor-
mance is non-linear (i.e., compensation spikes at
Fig. 2. Budget levels and e?ort. Panel A – marginal cost and
marginal bene?t of e?ort at three budget levels. Panel B – the
relationship between budget levels and e?ort. Panel A illustrates
the marginal cost and marginal bene?t of e?ort at three budget
levels. As is typically assumed, the cost of e?ort increases at an
increasing rate. For each budget level, the marginal bene?t of
e?ort is $0 until the budget is achieved. The marginal bene?t of
each unit of performance beyond the budget equals the piece-
rate. Panel B shows that as the budget level increases, e?ort also
increases up to a point – this occurs because of the initial bonus
– employees will increase their e?ort to receive this ?xed bonus.
At some point, however, the cost of reaching the budget
becomes too high and employees ‘‘give up.”
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 439
the budget level), higher risk projects with higher
expected performance do not always provide
employees the greatest expected pay. Below, we
discuss how the non-linearity in budget-based con-
tracts a?ects the relationship between budget levels
and risk-taking.
Panel A of Fig. 3 illustrates the output distribu-
tions for two projects, along with three budget lev-
els. Project 1 has a higher variance, but also a
higher mean, than project 2. Panel B of Fig. 3
shows the expected compensation for each project
at each budget level. Further, Exhibit 1 provides a
numerical example that illustrates the U-relation
between budget levels and risk-taking.
Notice that at the low budget level, employees
maximize expected compensation by choosing
project 1. Intuitively, this occurs because
employees are able to meet the budget with cer-
tainty by selecting either project. Thus, the pro-
ject with higher expected performance, project
Fig. 3. Budget levels and risk. Panel A – project outcome distributions and three budget levels. Panel B – budget levels and expected
pay for low- and high-risk projects. Panel C – the relation between budget levels and risk (project selection). Panel A illustrates the
output distributions of two projects, along with three budget levels. Project 1 has a higher variance, but also a higher mean, than
project 2. Panel B illustrates that at a low budget level, employees maximize their pay by choosing the higher risk, higher return project
(project 1). This occurs because employees are able to meet the budget with certainty by selecting either project – thus, they choose the
project with the higher expected performance. As the budget level increases, employees prefer project 2, the low risk project. This
occurs because project 2 allows them to meet the budget with certainty, whereas project 1 places their bonus in jeopardy. At very high
budget levels, the high risk project is the only way to meet the budget. Panel C documents this U-relation between budget levels and
risk (project selection).
440 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
1, provides the greatest expected compensation.
As the budget level increases, employees prefer
project 2 – this occurs because project 2 allows
them to meet the budget with certainty (and
attain the ?xed bonus) whereas project 1 places
the initial bonus in jeopardy. At very high bud-
get levels, project 1 is the only way to meet the
budget.
Panel C of Fig. 3 documents this U-Relation
between budget levels and risk (project selection).
This relation, which has not been addressed in
the budgeting literature, stands in stark contrast
to the relation between budget levels and e?ort.
It also illustrates a potential tradeo? between set-
ting budget levels to motivate high e?ort and
setting budget levels to encourage appropriate
risk-taking. This may help explain some of the
diverse results in the budgeting literature regarding
where budget levels should be set. Whether ?rms
prefer higher or lower budgets could depend on
whether e?ort or project selection is more impor-
tant to performance. Moreover, in settings where
both e?ort and project selection are equally impor-
tant to performance, a ?rm could be indi?erent
Exhibit 1.
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 441
between budget levels (after all, a U plus an
Inverted-U can average out to a straight line).
In the next section, we discuss psychology the-
ory that suggests the economic tradeo? between
e?ort and risk-taking may, in fact, be mitigated
at low budget levels but exacerbated at high bud-
get levels. In essence, organizations may get a ‘‘free
lunch” by setting low budgets.
3
Security-potential/aspiration (SP/A) theory
The prior discussion illustrates that, from an
individual wealth maximization standpoint,
increasing budget levels can have opposing e?ects
on a ?rm’s ability to motivate e?ort and encourage
risk-taking. Security-potential/aspiration (SP/A)
theory suggests that individuals will be willing to
sacri?ce expected wealth to either meet the budget
or increase their potential payo?s (i.e., the highest
possible payo?s). In contrast to conventional eco-
nomic theory that assumes risk-preferences are a
stable, innate characteristic, SP/A theory suggests
that risk-preferences would be a?ected by the bud-
get level in such a way as to promote more ?rm-
desired decisions at lower budget levels and less
desirable decisions at higher budget levels.
SP/A theory posits that individuals evaluate
alternatives in two stages when making decisions
in uncertain environments (see, e.g., Lopes, 1990;
Lopes, 1995; Lopes & Oden, 1999). In the ?rst
stage, individuals evaluate alternatives based on
the probability of meeting a minimum aspiration
level. In doing so, individuals behave in a risk-
averse manner, placing the greatest weight on out-
comes that do not meet their aspiration level. If
individuals cannot di?erentiate among alternatives
based on the ?rst criterion (e.g., multiple alterna-
tives meet the aspiration level with certainty), then,
in the second stage, individuals focus on potential
payo?s. Here, individuals behave in a risk-seeking
manner, placing the greatest weight on the highest
possible outcomes.
We are not aware of research that has examined
either the applicability or the implications of SP/A
theory in a budgeting context. However, individu-
als may adopt the budget as their aspiration level
for at least two reasons. First, individuals often
receive a substantial bonus for reaching the budget
target. Second, prior research suggests that goals
such as budgets have motivational properties inde-
pendent of their e?ects on compensation (e.g.,
Locke, Saari, Shaw, & Latham, 1981; Locke &
Latham, 1990). Moreover, employees may strive
to reach a budget simply because they have been
assigned a goal per se.
If employees do adopt budgets as their aspira-
tion level, then SP/A Theory could provide
insights into the relation between budget levels
and behavior. At low budget levels, individuals
can reach their budget with near certainty – thus,
individuals would tend to focus on the really good
payo?s. In turn, there are two ways to obtain these
really good payo?s. First, individuals can exert
more e?ort. Second, individuals can choose higher
variance projects.
4
Both activities are ‘‘risky” in
the sense that a higher variance project may not
pay o? (e.g., a low outcome could obtain). Addi-
tionally, working beyond the point at which mar-
ginal cost equals marginal bene?t can lead to a
good outcome but, by de?nition, is not a wealth
maximizing action. That said, such decisions are
preferred by the ?rm – organizations wish to elicit
as much e?ort as possible from their employees
3
We thank an anonymous Reviewer for this characterization
and help with relating security-potential/aspiration theory to
our study.
4
Projects that o?er higher levels of ?rm performance gener-
ally are of higher variance (risk). However, higher risk projects
with higher expected performance do not necessarily increase
employee remuneration. As discussed in the prior section, one
reason that these projects may not increase employee pay is the
non-linear relation between pay and performance under bud-
get-based contracts. Another reason is that higher risk projects
may have longer horizons whose returns and, in turn, e?ects on
employee pay, may not be realized for some time (Milgrom &
Roberts, 1992, Chapter 13). Moreover, in addition to alleviat-
ing agency concerns in a single-period setting, low budgets also
could alleviate intertemporal agency concerns associated with
the di?ering horizons of employees and organizations. This
argument is consistent with prior research demonstrating that
employees working under low budgets spend a greater amount
of time on projects whose performance e?ects will not be known
for over a year (Van der Stede, 2000).
442 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
and have them choose projects with the highest
expected return to the ?rm.
At higher budget levels, individuals cannot
reach their budget with certainty. Consequently,
SP/A theory suggests that individuals will place
greater weight on the worst possible outcomes
when evaluating their options (i.e., they will
become more risk-averse). Here, employees
would select projects with the highest probability
of meeting the budget – this project may or may
not be the one with the highest expected return.
Moreover, risk aversion shrinks the expected
marginal bene?t of e?ort leading, of course, to
less e?ort.
5
Both activities expose the employee
to less risk, but, naturally, are not preferred by
the ?rm.
Fig. 4 illustrates the implications of SP/A the-
ory vis-a`-vis expected wealth maximization. Panel
A illustrates that e?ort would be higher under a
low budget and lower under a high budget than
would be predicted under expected wealth maximi-
zation. Panel B shows that employees’ propensity
to select riskier projects would be higher under a
low budget and lower under a higher budget than
would be predicted under expected wealth maximi-
zation. Moreover, a ?rm would likely prefer a low
budget to a high budget in settings where both
e?ort and project selection are equally important
to performance, rather than being indi?erent as
suggested by an expected wealth maximization
perspective.
In sum, SP/A theory suggests that low budgets
may lead individuals to behave in a risk-seeking
manner, whereas high budgets may lead individu-
als to behave in a risk-averse manner. This leads to
the following hypothesis to be tested:
Hypothesis. Employees rewarded with a low
(high) budget behave in a more risk-seeking
(risk-averse) manner when making e?ort and
project selection decisions relative to employees
rewarded with a high (low) budget.
Experimental method
Participants and design
Sixty undergraduate business students volun-
teered to participate in the experiment. Partici-
5
In other words, if employees exhibit greater risk aversion
when they cannot reach their budget with certainty, then they
would discount the potential bene?ts of meeting and exceeding
the budget. In turn, this makes it more likely that the costs of
reaching the budget would exceed the uncertain bene?ts.
Fig. 4. SP/A theory versus expected wealth maximization.
Panel A – SP/A theory versus expected wealth maximization.
The relation between budget levels and e?ort. Panel B – SP/A
theory versus expected wealth maximization: The relation
between budget levels and risk. This ?gure compares security-
potential/aspiration (SP/A) theory to expected wealth maximi-
zation. Panel A compares the relation between budget levels
and e?ort, while Panel B compares the relation between budget
levels and risk decisions.
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 443
pants were randomly assigned to one of two exper-
imental conditions. We obtained these two condi-
tions by manipulating a single factor, the budget
level used to reward participants, at two levels
(low and high) between subjects. The experiment
consisted of two primary activities. In the ?rst
activity, participants worked on a time-intensive
task. In the second activity, participants chose to
play one of twelve possible lotteries that varied
in risk and expected return. This lottery choice
represented participants’ e?ort and project selec-
tion decisions. In the remainder of this section,
we describe the twelve lotteries and the experimen-
tal procedures.
The lotteries representing the e?ort and project
selection decisions
The twelve lotteries representing the e?ort and
project selection decisions for the low- and high-
budget conditions are described in Panels A and
B of Fig. 5, respectively. The twelve lotteries are
presented in a three row by four column matrix.
The lotteries are described by six pieces of infor-
mation. However, participants were only provided
with the ?rst four pieces of information. We dis-
cuss the information in each cell below.
For each lottery, the four pieces of infor-
mation provided to participants summarized
one-hundred equally likely states of nature. The
speci?c payo? function for each lottery was con-
sistent with the budget-based contract depicted
in Fig. 1. Participants received a low ?at payo?
if the state of nature from their selected lottery
was below their budget and a higher (linearly
increasing) payo? if the state of nature was
above their budget. Thus, knowing (1) the prob-
ability of meeting the budget, (2) the payo? if
the budget was not met, (3) the lowest payo?
if the budget was met, and (4) the highest payo?
if the budget was met (each amount between the
low and high payo? was equally likely) provided
the information needed to describe the 100 states
of nature in each cell.
The organization of the twelve cells in the
matrix is consistent with a multiplicative produc-
tion function – output equals participant’s e?ort
choice multiplied by a state variable drawn from
the project selected.
6
E?ort choices are represented
by rows, and project selection decisions are repre-
sented by columns. For the reasons discussed in
the prior section, we assume that increasing e?ort
(moving from row A to row B to row C) and
selecting riskier projects (by moving from left to
right across the columns) is bene?cial to the ?rm.
Below, we describe how the e?ort and project
selection decisions a?ect participants’ payo?s.
The payo?s are structured such that increasing
e?ort (moving from row A to row B to row C)
has three e?ects on the lotteries. First, the amount
received for not reaching the budget decreases
because e?ort is costly (the payment for not reach-
ing the budget represents the ?xed wage paid to
the employee less the cost of e?ort). Second, the
probability of reaching the budget either increases
or stays the same as employee e?ort increases.
Third, the variance of the payo?s increases.
7
By moving from left to right across the col-
umns, participants increase the variance of the
selected project. Choosing projects with higher
variances has the following two e?ects on the lot-
teries. First, the probability of reaching the budget
either decreases or stays the same. Second, the var-
iance of the payo?s increases.
Although we organized the cells so that the pay-
o? distributions are consistent with that of a mul-
tiplicative production function, our primary goal
in building the matrices was to create a setting
where we could observe how budget levels a?ect
the propensity of participants to trade-o? expected
compensation for risk-taking. Each matrix had
one cell (B2) that o?ered participants the highest
expected compensation ($17.25). Participants
could sacri?ce expected compensation in $0.25
increments by choosing one of the three cells with
increasingly lower variances (B1, A2, and A1).
These cells, labeled ‘‘risk-averse cells” in Fig. 5,
allowed participants to meet the budget with cer-
6
Both the tensions captured and our predictions would be
similar if we used an additive production function, where
output equals e?ort plus a state drawn from a probability
distribution, to build our lottery matrices.
7
We represent projects as uniform distributions. Naturally,
when these distributions are multiplied by larger e?ort num-
bers, the range (variance) of output and payo?s increases.
444 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
Panel A – The Effort and Project Selection Matrix for Participants inthe Low-Budget Condition
Panel B – The Effort and Project Selection Matrix for Participants in the High-Budget Condition
1 2 3 4
Probability of meeting the budget 100% 100% 65% 55%
Amount paid if the budget is not met NA NA $3.50 $3.50
Lowest amount paid if the budget is met $16.25 $16.00 $15.00 $13.00
Highest amount paid if the budget is met $16.75 $17.50 $20.00 $43.00
Cell Type Risk-Averse Risk-Averse Demo Risk-Seeking
Expected Compensation
A
$16.50 $16.75 $12.60 $16.98
Probability of meeting the budget 100% 100% 100% 60%
Amount paid if the budget is not met NA NA NA $2.50
Lowest amount paid if the budget is met $15.75 $14.75 $7.50 $6.00
Highest amount paid if the budget is met $18.25 $19.75 $26.00 $46.50
Cell Type Risk-Averse High EV Potential Risk-Seeking
Expected Compensation
B
$17.00 $17.25 $16.75 $16.75
Probability of meeting the budget 100% 100% 100% 65%
Amount paid if the budget is not met NA NA NA $0.00
Lowest amount paid if the budget is met $10.50 $10.00 $5.00 $2.00
Highest amount paid if the budget is met $18.50 $24.00 $28.00 $48.75
Cell Type Demo Potential Potential Risk-Seeking
Expected Compensation
C
$14.50 $17.00 $16.50 $16.49
1 2 3 4
Probability of meeting the budget 100% 100% 65% 55%
Amount paid if the budget is not met NA NA $3.50 $3.50
Lowest amount paid if the budget is met $16.25 $16.00 $15.00 $13.00
Highest amount paid if the budget is met $16.75 $17.50 $20.00 $43.00
Cell Type Risk-Averse Risk-Averse Demo Risk-Seeking
Expected Compensation
A
$16.50 $16.75 $12.60 $16.98
Probability of meeting the budget 100% 70% 65% 60%
Amount paid if the budget is not met NA $2.50 $2.50 $2.50
Lowest amount paid if the budget is met $15.75 $14.75 $7.50 $6.00
Highest amount paid if the budget is met $18.25 $32.40 $41.35 $46.50
Cell Type Risk-Averse High EV Potential Risk-Seeking
Expected Compensation
B
$17.00 $17.25 $16.75 $16.75
Probability of meeting the budget 100% 75% 70% 65%
Amount paid if the budget is not met NA $0.00 $0.00 $0.00
Lowest amount paid if the budget is met $10.50 $10.00 $5.00 $2.00
Highest amount paid if the budget is met $18.50 $35.35 $42.15 $48.75
Cell Type Demo Potential Potential Risk-Seeking
Expected Compensation
C
$14.50 $17.00 $16.50 $16.49
Fig. 5. E?ort and project selection matrixes for the low- and high-budget conditions. Panel A – the e?ort and project selection
matrix for participants in the low budget condition This ?gure presents the parameters used for our experiment’s low- and high-
budget conditions. Participants received the ?rst four pieces of information in each cell, which describe one-hundred states of
nature for competing lotteries. One lottery had the highest expected compensation (High EV). The three risk-averse lotteries had
lower expected values and lower variances. The potential and risk-seeking lotteries had lower expected values and higher variances.
The demo lotteries were used in the examples embedded within participants’ experimental instructions. The High Expected Value
and Potential lotteries met the budget with certainty only in the low budget condition. The payo?s are structured such that
increasing e?ort (moving from row A to row B to row C) has three e?ects on the lotteries. First, the amount received for not
reaching the budget decreases because participants pay an increasing cost for exerting higher levels of e?ort (the payment for not
reaching the budget represents the ?xed wage paid to the employee less the cost to the employee for exerting e?ort). Second, the
probability of reaching the budget either increases or stays the same as e?ort increases. Third, the variance of the payo?s increases.
Project selection choices are represented by columns. By moving from left to right across the columns, participants increase the
variance of the selected project. Choosing projects with higher variances has the following two e?ects on the payo? distributions of
the lotteries. First, the probability of reaching the budget either decreases or stays the same. Second, the variance of the payo?s
increases.
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 445
tainty. Participants also could sacri?ce expected
compensation in approximately $0.25 increments
by choosing cells with increasingly higher vari-
ances than the cell with the greatest expected com-
pensation (A4, B4, and C4). If participants chose
one of these three cells (labeled ‘‘risk-seeking
cells”), then they would meet the budget between
55% and 65% of the time. Finally, participants
could sacri?ce expected compensation in $0.25
increments by choosing one of the cells labeled
‘‘potential cells” (C2, B3, and C3). The potential
cells had higher variances than the expected value
cell but lower variances than the three risk-seeking
cells. We describe the potential cells in more detail
in the paragraphs that follow.
8
To create our low and high budget conditions,
we manipulated the probabilities of reaching the
budget in the highest expected compensation cell
and the three potential cells. Similar to the intui-
tion described in the background section, the high-
est expected compensation cell in the low budget
matrix allowed participants to reach the budget
with certainty. Further, the three potential cells
also allowed participants to reach the budget with
certainty. Thus, by choosing one of the potential
cells, participants could sacri?ce expected compen-
sation to take greater risk (increasing their poten-
tial for higher payo?s) while still reaching the
budget with certainty. Including both potential
and risk-seeking cells allows us to discriminate
between behavior consistent with SP/A theory,
where participants with low budgets choose higher
variance projects as long as they meet the budget
with certainty, and general risk-seeking behavior.
To the extent that SP/A theory is descriptive of
behavior, we would expect to see more individuals
choosing one of the potential cells and no more
individuals selecting one of the risk-seeking cells
in the low budget condition relative to the high
budget condition.
The highest expected compensation cell in the
high budget condition did not allow participants
to reach the budget with certainty. Further, partic-
ipants could not reach the budget with certainty by
taking greater risk (i.e., the potential cells did not
allow participants to reach the budget with cer-
tainty).
9
Thus, participants could only reach the
budget with certainty by sacri?cing expected com-
pensation to choose one of the cells with the lowest
variances (the risk-averse cells).
Experimental procedures
Upon arrival, we informed participants that
they would be completing several activities and
that the entire experiment would last approxi-
mately two hours. The instructions for each activ-
ity were read out loud, and all participants in each
experimental session worked at the same pace
(participants received one activity at a time).
In the ?rst activity, participants were provided
with sixty boxes each with ten rows and thirty-
one columns of random letters. Each box had a
single letter in its top-right hand corner (the search
letter). For each box, participants had to circle
every instance of the search letter and provide
the total in the top-right hand corner. Participants
learned that they would earn the right to partici-
pate in the next activity of the experiment by
working on the exercise for 45 minutes and provid-
ing acceptable totals (answers within one of the
correct answer) for at least ?fteen of the boxes.
10
At the conclusion of the 45 minute period, the
experimenter veri?ed that each participant met
the performance requirements.
8
Each matrix also has two cells (A3 and C1) that were
intentionally made unattractive to the participants. That is,
these cells were strictly dominated by other cells. We used these
two cells, labeled demo cells, for our examples in the experi-
mental instructions given to the participants.
9
To hold the expected compensation of these four cells
constant across the low and high budget conditions, we
increased the highest amount paid if the threshold is met
(increasing the slope of the linear cash payo?s above the budget
amount) in the high budget condition for these four cells.
10
Participants were informed that if they did not meet these
performance requirements then they would be paid $4 and
would be dismissed from the experiment. Further, participants
were informed that they could earn signi?cantly more money by
successfully completing this activity. All sixty participants
successfully completed the performance requirements.
446 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
The primary purpose of the ?rst activity was to
ensure participants earned the lottery choice that
they made in the second activity of the experiment.
When selecting among competing lotteries, prior
research demonstrates that participants make dif-
ferent decisions depending on whether they earn,
or are endowed with, their choice (Boylan & Sprin-
kle, 2001).
11
More generally, employees bear both
monetary (e.g., opportunity cost of foregone
wages) and non-monetary (e.g., opportunity cost
of time that could be used for leisure) costs in work
settings. Having participants actually spend time
at the experiment ensures that we have both ele-
ments in our study. In other words, our primary
interest was in examining the choices made by par-
ticipants who worked to earn their choice.
At the start of the second activity, participants
learned that they had earned the right to choose
one of twelve lotteries presented on the three row
by four column matrix described earlier and that
the outcome of the lottery would be paid to them
in cash at the end of the experiment. The instruc-
tions described each of the four pieces of informa-
tion, including the payment received for not
reaching the budget, the probability of reaching
the budget, and the variance of payments for out-
comes above the budget when moving across the
rows and down the columns.
12
Participants also
were provided with calculators and a sheet that
detailed each possible state of nature for the twelve
lotteries.
Participants studied the matrix and then
marked their cell choice on the sheet provided.
To determine the outcome of their chosen lottery,
participants drew a ball out of a bag containing
balls numbered 1 through 100. Based on this draw
and the cell chosen, participants calculated their
compensation which was veri?ed by the
experimenter.
Next, participants completed a post-experimen-
tal questionnaire. Finally, participants completed a
risk-preference instrument.
13
This instrument
required participants to make ?fteen choices
between receiving a certain amount of $2.50 and
participating in a lottery. The lotteries consisted
of a chance of winning $5 with a probability of p
and $0 with a probability 1 À p (p varied from
85% to 15% in 5% increments). Once participants
made their ?fteen choices, one of the ?fteen choices
was selected at random, the lottery was conducted,
and participants’ earnings were determined based
on their choice and the result of the lottery. Pay-
ments for activity two of the experiment and the
risk-preference instrument were totaled, and par-
ticipants were paid in cash and dismissed.
Results
Our analyses focus on the cell choices made by
participants in the second activity of our experi-
ment. Participants’ cell choices, by budget condi-
tion, are presented graphically in Fig. 6. The cells
on the graph are arranged by cell type (demo,
risk-averse, expected compensation, potential,
and risk-seeking). As expected, no participant
chose a demo cell type. Thus, we drop this cell type
from the remainder of the analyses. Further, no
participant chose the risk-averse cells A1 or A2.
That is, only the risk-averse cell with the highest
variance and expected compensation (B1) was cho-
sen by participants. All other cells, however, were
chosen by at least one participant. Because not all
cells were chosen by the participants and to
streamline the presentation of the results, we col-
lapse the cell choices of participants into their
respective cell types for our analyses. However,
our inferences remain the same when using the
disaggregated cell choices.
Table 1 presents participants’ cell choice fre-
quencies, aggregated by type, for the low- and
high-budget conditions. To test our hypothesis,
we compare the distributions of these frequencies
11
More generally, research demonstrates that individuals
make di?erent decisions depending on whether they earn, or
are endowed with a ‘‘property/decision right” (see, e.g.,
Ho?man & Spitzer, 1985; Ho?man, McCabe, Shachat, &
Smith, 1994).
12
Budgets were described as thresholds in the experimental
instructions.
13
The risk preference instrument is a modi?ed version of the
instrument used in Boylan and Sprinkle (2001).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 447
across budget conditions. In the low budget condi-
tion, we are interested in whether a disproportion-
ate number of participants chose a potential cell,
thereby sacri?cing expected compensation to
increase potential payo?s (i.e., they behaved in a
more risk-seeking manner). In the high-budget
condition, we are interested in whether a dispro-
portionate number of participants chose a risk-
0
2
4
6
8
10
12
14
Choice
F
r
e
q
u
e
n
c
y
Low 0 0 0 0 0 10 2 9 3 2 3 1
3 0 High 0 0 0 0 13 6 3 4 0 1
A3 Demo C1 Demo A1 RA A2 RA B1 RA B2 EV B3 P C2 P C3 P A4 RS B4 RS C4 RS
Fig. 6. Participants’ cell choices in the low- and high-budget conditions. This graph shows participants’ cell choices in the low-and
high-budget conditions. The cells, which are described in Fig. 5, are grouped by the following cell types: Demo (demonstration cells),
RA (risk-averse cells), EV (highest expected compensation), P (potential cells), and RS (risk-seeking cells).
Table 1
Participants’ cell choices in low- and high-budget conditions
Choice
a
Budget level
Low budget High budget
n
b
%
c
Adj. Res.
d
n
b
%
c
Adj. Res.
d
Risk-averse 0 0.0 À4.1
**
13 43.3 4.1
**
High EV 10 33.3 À1.2 6 20.0 1.2
Potential 14 46.7 À1.9
*
7 23.3 1.9
*
Risk-seeking 6 20.0 À0.7 4 13.3 0.7
v
2
(3, N = 60) = 16.73, p < 0.01
a
This column lists the possible lottery categories chosen by at least one participant during the second activity of the experiment. The
High EV choice has the greatest expected compensation. The risk-averse choices have lower expected values and variances than the
High EV choice. The potential and risk-seeking choices have lower expected values and higher variances than the High EV choice. In
the low budget condition, the High EV and Potential choices meet the budget with certainty. In the high-budget condition, these
choices do not meet the budget with certainty.
b
The number of participants that chose the lottery category.
c
The percentage of participants that chose the lottery category.
d
The observed cell frequency less the expected cell frequency assuming frequencies are identically distributed across budget level
conditions. The results of tests to determine whether the residual is signi?cantly di?erent from zero:
**
p < 0.01;
*
p = 0.06.
448 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
averse cell, thereby sacri?cing expected compensa-
tion to increase the probability of meeting the bud-
get (i.e., they behaved in a more risk-averse
manner).
The relative distributions of cell type frequen-
cies across the low- and high-budget conditions
support our hypothesis. Thirteen of the 30 partic-
ipants in the high-budget condition chose a risk-
averse cell, but no participants in the low budget
condition did so. Further, 14 of the 30 participants
in the low budget condition chose a potential cell,
while 7 of 30 chose one of these cells in the high-
budget condition. Moreover, we do not observe
large di?erences in the number of participants
choosing the highest expected compensation or
risk-seeking cells across budget conditions.
To formally test our hypothesis, we used an
independent samples Chi-square analysis to com-
pare the relative distribution of cell type frequen-
cies across the low- and high-budget conditions.
This test shows that the distributions are statisti-
cally di?erent (v
2
= 16.73, p < 0.01). To identify
the cell types driving the signi?cant result, we
examine the residual of each cell (the observed cell
frequency less the expected cell frequency given
these frequencies are identically distributed across
our budget conditions). Speci?cally, we examine
which of the cell type’s residuals adjusted for their
variances (hereafter adjusted residuals) are signi?-
cantly larger than expected.
14
The adjusted residual of À4.1 (4.1) for the risk-
averse cell type in the high (low) budget condition
is signi?cantly larger than expected (p < 0.01).
Thus, signi?cantly more (fewer) individuals chose
a risk-averse cell than expected if the cell type fre-
quencies were identically distributed across our
budget conditions. Similarly, the adjusted residual
of 1.9 (À1.9) for the potential cell type in the high
(low) budget condition is signi?cantly larger than
expected (p = 0.057). Thus, signi?cantly fewer
(more) individuals chose a potential cell than
expected if the cell type frequencies were identi-
cally distributed across budget conditions. How-
ever, the adjusted residuals for the expected
compensation and risk-seeking cell types are not
signi?cant (p = 0.23 and p = 0.48, respectively).
In essence, the results support our hypothesis.
Participants in the low budget condition sacri?ced
expected compensation to take on additional risk.
Participants in the high-budget condition, on the
other hand, sacri?ced expected compensation to
reduce risk.
15
We also examined whether participants’ dispo-
sition toward risk a?ected our results. To accom-
plish this, we partitioned our sample into
participants with high and low risk tolerances
based on their responses to the risk-preferences
instrument. As previously discussed, this instru-
ment required participants to make ?fteen choices
between receiving a certain amount of $2.50 and
participating in a lottery. Each lottery consisted
of a chance of winning $5 with a probability of p
and $0 with a probability of 1 À p. The probability
of winning $5 in the ?rst lottery was 85%. We
decreased the chance of winning by 5% in each
subsequent lottery, such that the chance of win-
ning $5 in the ?nal lottery was 15%. We recorded
the number of the choice where individuals pre-
ferred the certain amount of $2.50 to the lottery.
The lower the score on this measure, the lower
the individual’s tolerance for risk. The average
score on this measure was 6.83, and means did
14
The standardized residual is given by the following formula
(n
ij
À E
ij
)/vE
ij
where n
ij
is actual cell frequency and E
ij
is the
expected cell frequency [(n
i.
 n
.j
)/N]. We adjust the standard-
ized residual by dividing it by its variance estimated by (1 À n
i.
/
N) Â (1 À n
.j
/N) where n
i.
is the total observations in row i, n
.j
is
the total observations in column j, and N is the total
observations in the experiment. Because the variables forming
our contingency table are independent, these adjusted residuals
are approximately normally distributed with a mean of 0 and a
standard deviation of 1 (Everitt, 1977, p. 47).
15
Consistent with SP/A theory, the post-experimental ques-
tionnaire provides evidence that participants aspired to meet
their budget. Participants were asked to answer the following
statement: ‘‘Meeting the budget was important to me”. Partic-
ipants responded using a seven-point Likert scale with ‘‘1”
being ‘‘strongly disagree,” ‘‘4” being ‘‘neither agree nor
disagree,” and ‘‘7” being ‘‘strongly agree.” Participants mean
response, 5.32, was signi?cantly greater than the midpoint of
the scale (t = 6.14, p < 0.01, two-tailed). Further, mean
responses did not statistically di?er across the low- and high-
budget conditions (t = 1.17, p = 0.28, two-tailed).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 449
not statistically di?er across the low and high bud-
get conditions (t = 0.24, p = 0.81, two-tailed).
We partitioned our sample into the 27 partici-
pants with scores on the risk-preferences instru-
ment lower than the mean (the low risk-tolerance
group), and the 33 participants with scores larger
than the mean (the high risk-tolerance group). In
Table 2, we report the results of separate v
2
anal-
yses for each of these groups.
Panel A of Table 2 presents the results for the
subset of participants with low risk tolerances.
Overall, the cell choices of this subset of partici-
pants di?ered across the low and high budget con-
ditions (v
2
= 12.08, p < 0.01). Further, the
distribution of choices di?ered in a manner consis-
tent with our results on the entire sample. Relative
to participants in the high-budget condition, sig-
ni?cantly fewer participants in the low budget con-
dition chose a risk-averse cell (p < 0.01) and
signi?cantly more chose a potential cell (p < 0.05).
Panel B of Table 2 presents the results for the
subset of participants with higher risk tolerances.
The cell choices of this subset of participants mar-
ginally di?ered across the low- and high-budget
conditions (v
2
= 6.27, p = 0.10). Relative to partic-
ipants in the high budget condition, signi?cantly
fewer participants in the low budget condition
chose a risk-averse cell (p < 0.01). However, par-
ticipants choices did not di?er for potential cells
(p = 0.48).
Collectively, these results suggest that SP/A the-
ory is more descriptive of behavior in individuals
with low risk tolerances. Since individuals in work
settings are presumed to be risk-averse, this sug-
Table 2
Choices of participants in low and high budget conditions partitioned by risk tolerance
a
Choice
b
Budget level
Low budget High budget
n
c
%
d
Adj. Res.
e
n
c
%
d
Adj. Res.
e
Panel A – Analysis for the subset of participants with low risk tolerances
Risk-averse 0 0.0 À3.4
*
7 58.3 3.4
*
High EV 4 26.7 0.6 2 16.7 À0.6
Potential 8 53.3 2.0
*
2 16.7 À2.0
*
Risk-seeking 3 20.0 0.8 1 8.3 À0.8
v
2
(3, N = 27) = 12.08, p < 0.01
Panel B – Analysis for the subset of participants with high risk tolerances
Risk-averse 0 0.0 À2.5
*
6 33.3 2.5
*
High EV 6 40.0 1.1 4 22.2 À1.1
Potential 6 40.0 0.7 5 27.8 À0.7
Risk-seeking 3 20.0 0.2 3 16.7 À0.2
v
2
(3, N = 33) = 6.27, p = 0.10
a
Participants were partitioned into high and low risk tolerance groups based on their responses to a risk-preferences instrument. This
instrument required participants to make ?fteen choices between receiving a certain amount of $2.50 and a lottery. Each lottery
consisted of a chance of winning $5 with probability p and $0 with probability 1 À p. The probability of winning $5 in the ?rst lottery
was 85%. We decreased the chance of winning p in subsequent lotteries by 5%, such that the chance of winning $5 in the ?nal lottery
was 15%. We recorded the number of the choice where each individual preferred the certain amount of $2.50 to the lottery. We
partitioned participants with a score lower (higher) than the mean score into the low (high) risk tolerance group.
b
This column lists the possible lottery categories chosen by at least one participant during the second activity of the experiment. The
High EV choice has the greatest expected compensation. The risk-averse choices have lower expected values and variances than the
High EV choice. The potential and risk-seeking choices have lower expected values and higher variances than the High EV choice. In
the low budget condition, the High EV and Potential choices meet the budget with certainty. In the high-budget condition, these
choices do not meet the budget with certainty.
c
The number of participants that chose the lottery category.
d
The percentage of participants that chose the lottery category.
e
The observed cell frequency less the expected cell frequency assuming frequencies are identically distributed across budget level
conditions. The results of tests to determine whether the residual is signi?cantly di?erent from zero:
*
p < 0.05.
450 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
gests that SP/A Theory may have far-reaching
implications.
Discussion
In this paper, we examine how budget-based
contracts a?ect e?ort and risk-taking. We show
that, from a wealth maximization perspective, a
tradeo? exists between motivating e?ort and
encouraging risk-taking. We illustrate an
inverted-U relationship between budget levels
and e?ort. Budget levels and e?ort are positively
correlated until budgets become very di?cult, at
which point e?ort drops substantially. We illus-
trate an opposing, U-shaped, relationship between
budget levels and risk-taking. Low budgets pro-
vide employees the ?exibility to allocate their e?ort
to riskier projects. High budgets induce employees
to choose lower-risk projects to ensure a high
probability of reaching the budget. Riskier pro-
jects provide the greatest probability of reaching
very high (stretch) budgets.
We conduct a laboratory experiment to empiri-
cally test this economic proposition vis-a`-vis
extant psychology research. We ?nd that individu-
als evaluated and rewarded with a low budget
exhibited a greater propensity to sacri?ce expected
compensation to take on additional risk. Consis-
tent with security-potential/aspiration (SP/A) the-
ory, participants working under the low budget
were attracted by the potentially high payo?s
because budget attainment was a fait accompli.
This leads individuals to work harder (beyond
the point at which expected marginal bene?ts
equal the marginal cost of e?ort) and select higher
variance projects. In the context of our setting, this
result is desired from the ?rm’s perspective and
increases ?rm welfare.
In contrast, participants working under the
high budget exhibited a greater propensity to sac-
ri?ce expected compensation to reduce risk. Con-
sistent with SP/A theory, concerns about meeting
the budget increased risk aversion. In turn, risk
aversion leads individuals to exert less e?ort
(because the expected marginal bene?t shrinks)
and select ‘‘safe” projects. In the context of our
setting, both actions reduce ?rm welfare. In
essence, our experimental results suggest that the
e?ort and risk-taking tradeo? illustrated in our
example can be mitigated at low budget levels
but exacerbated at high budget levels.
Collectively, our results have several important
implications. First, they demonstrate that account-
ing-based performance evaluation and reward sys-
tems can have opposing e?ects on two primary
determinants of performance, e?ort and risk-tak-
ing. Control systems that motivate organization-
ally desirable behavior along one dimension of
performance can simultaneously induce actions
detrimental to the organization on another dimen-
sion of performance.
Second, these tradeo?s illustrate the importance
of decomposing performance into its fundamental
determinants. When designing control systems, for
example, organizations need to orchestrate the
assignment of decision rights with the manner in
which performance is measured and rewarded
(Brickley, Smith, & Zimmerman, 2001). Thus,
the use of high budget levels may be desirable
when employees have few decision rights and/or
work in non-stochastic settings, whereas the use
of lower budget levels may be desirable when
employees have more latitude in project/task selec-
tion decisions and/or work in environments char-
acterized by uncertainty.
Third, the speci?c trade-o? we document
between e?ort and risk-taking helps to reconcile
the mixed results of prior research. On one hand,
experimental research tends to ?nd that challeng-
ing budgets induce the highest levels of perfor-
mance (see, e.g., Locke & Latham, 1990). Based
primarily on this evidence, popular management
accounting textbooks advocate evaluating employ-
ees based on challenging budgets (Horngren et al.,
2007, p. 183). However, the tasks used in prior
experimental studies are often e?ort intensive
(e.g., the decoding of letters) and performance
likely varies little based on the risk-taking of par-
ticipants. On the other hand, prior archival studies
suggest that organizations tend to utilize lower
budget levels to evaluate employees. These studies
were conducted in environments where employees
had more latitude in project/task selection deci-
sions (Merchant & Manzoni, 1989; Simon, 1988;
Van der Stede, 2000).
G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452 451
Certain features of our study may limit the gen-
eralizability of our ?ndings. These limitations sug-
gest a number of avenues for further inquiry. For
example, we examined a single-period setting.
Most employment relationships extend over multi-
ple periods, and it is possible that learning and
wealth e?ects would, over time, lead individuals
to maximize expected compensation. Additionally,
we examined a single-person setting and our
results may not extend to multiperson settings
where gains to trade must be split – equity and
fairness concerns may lead individuals to adopt
di?erent aspiration levels and, in turn, make di?er-
ent e?ort and risk decisions. Finally, we only
examine the e?ect of one type of incentive, bud-
get-based contracts. Future research could exam-
ine whether other incentive schemes (e.g.,
tournament or piece-rate), or combinations and
dimensions thereof, induce individuals to take
appropriate levels of risk while concurrently moti-
vating high levels of e?ort.
References
Baiman, S. (1990). Agency research in managerial accounting:
A second look. Accounting, Organizations and Society, 15,
341–371.
Boylan, S. J., & Sprinkle, G. B. (2001). Experimental evidence
on the relation between tax rates and compliance: The e?ect
of earned vs. endowed income. The Journal of the American
Taxation Association, 23, 75–90.
Brickley, J. A., Smith, C. W., Jr., & Zimmerman, J. L. (2001).
Managerial economics and organizational architecture. New
York, NY: The McGraw-Hill Companies Inc.
Everitt, B. S. (1977). The analysis of contingency tables. New
York, NY: John Wiley and Sons Inc.
Hirst, M. K., & Yetton, P. W. (1999). The e?ects of budget
goals and task interdependence on the level of and variance
in performance: A research note. Accounting, Organizations
and Society, 24, 205–216.
Ho?man, E., McCabe, K., Shachat, K., & Smith, V. (1994).
Preferences, property rights, and anonymity in bargaining
games. Games and Economic Behavior, 7, 346–380.
Ho?man, E., & Spitzer, M. L. (1985). Entitlements, rights, and
fairness: An experimental examination of subjects’ concepts
of distributive justice. Journal of Legal Studies, 14, 259–297.
Horngren, C. T., Datar, S. M., & Foster, G. (2007). Cost
accounting: a managerial emphasis. Upper Saddle River, NJ:
Prentice Hall.
Ittner, C. D., & Larcker, D. F. (2001). Assessing empirical
research in managerial accounting: A value-based manage-
ment perspective. Journal of Accounting and Economics, 32,
349–410.
Locke, E. A., & Latham, G. P. (1990). A theory of goal setting
and task performance. Englewood Cli?s, NJ: Prentice Hall.
Locke, E. A., Saari, L., Shaw, K., & Latham, G. (1981). Goal
setting and task performance: 1969–1980. Psychological
Bulletin, 1, 125–152.
Lopes, L. (1990). Re-modeling risk aversion: A comparison of
Bernoullian and rank dependent value approaches. In G. M.
Von Furstenberg (Ed.), Acting under uncertainty: Multidis-
ciplinary conceptions (pp. 267–299). Boston, MA: Kluwer
Academic Publishers.
Lopes, L. (1995). On modeling risk choice. In Contributions to
decision making I (pp. 29–50). New York, NY: Elsevier.
Lopes, L., & Oden, G. (1999). The role of aspiration level in
risky choice: A comparison of cumulative prospect theory
and SP/A theory. Journal of Mathematical Psychology, 43,
286–313.
Luft, J., & Shields, M. D. (2003). Mapping management
accounting: Graphics and guidelines for theory-consistent
empirical research. Accounting, Organizations and Society,
28, 169–249.
Merchant, K. A., & Manzoni, J. F. (1989). The achievability of
budget targets in pro?t centers: A ?eld study. The Account-
ing Review, 3, 539–558.
Merchant, K. A., & Van der Stede, W. A. (2007). Management
control systems: Performance measurement, evaluation, and
incentives. London, UK: Prentice Hall.
Milgrom, P. R., & Roberts, J. (1992). Economics, organization
& management. Upper Saddle River, NJ: Prentice Hall.
Murphy, K. (2001). Performance standards in incentive con-
tracts. Journal of Accounting and Economics, 30, 245–278.
Simon, R. (1988). Analysis of the organizational characteristics
related to tight budget goals. Contemporary Accounting
Research, 5, 267–283.
Sprinkle, G. B. (2003). Perspectives on experimental research in
managerial accounting. Accounting, Organizations and Soci-
ety, 28, 287–318.
Sprinkle, G. B., & Williamson, M. G. (2004). The Evolution
from taylorism to employee gainsharing: A case study
examining John Deere’s continuous improvement pay plan.
Issues in Accounting Education, 19, 487–503.
Valance, N. (2001). Bright minds, bigtheories. CFO(January), 68.
Van der Stede, W. A. (2000). The relationship between two
consequences of budgetary controls: Budgetary slack crea-
tion and managerial short-term orientation. Accounting,
Organizations and Society, 25, 609–622.
452 G.B. Sprinkle et al. / Accounting, Organizations and Society 33 (2008) 436–452
doc_821260073.pdf