Description
Budgets are often simultaneously used for the conflicting purposes of planning and performance
evaluation. While economic theory suggests that firms should use separate budgets
for conflicting purposes this contrasts with existing evidence that firms rarely do so. We
address two open questions related to these observations in an experiment. Specifically,
we investigate how a planning task that is in conflict with the performance evaluation task
affects behavior in budget negotiations and their outcomes. Additionally, we analyze
whether a single budget can be effectively used for both purposes compared to two separate
budgets. We develop theory to predict that adding a planning task that is in conflict
with the superior’s performance evaluation task increases the subordinate’s cooperation
in and after the negotiation of a performance evaluation budget.
Using negotiated budgets for planning and performance
evaluation: An experimental study
q
Markus C. Arnold
a
, Robert M. Gillenkirch
b,?
a
Institute for Accounting, University of Bern, 3012 Bern, Switzerland
b
Business Department, University of Osnabrück, 49069 Osnabrück, Germany
a r t i c l e i n f o
Article history:
Available online 21 March 2015
a b s t r a c t
Budgets are often simultaneously used for the con?icting purposes of planning and perfor-
mance evaluation. While economic theory suggests that ?rms should use separate budgets
for con?icting purposes this contrasts with existing evidence that ?rms rarely do so. We
address two open questions related to these observations in an experiment. Speci?cally,
we investigate how a planning task that is in con?ict with the performance evaluation task
affects behavior in budget negotiations and their outcomes. Additionally, we analyze
whether a single budget can be effectively used for both purposes compared to two sepa-
rate budgets. We develop theory to predict that adding a planning task that is in con?ict
with the superior’s performance evaluation task increases the subordinate’s cooperation
in and after the negotiation of a performance evaluation budget. Moreover, we predict that
subordinate cooperation increases even more when the superior is restricted to use a single
budget for both purposes. Our results broadly support our hypotheses. Speci?cally, we ?nd
that when budgets are used for both planning and performance evaluation, this increases
the subordinate’s budget proposals during the negotiation and his performance after the
negotiation. These effects tend to be even larger when the superior is restricted to a single
budget rather than separate budgets for planning and performance evaluation, particularly
with respect to subordinate performance. In our experimental setting, the bene?ts of
increased subordinate cooperation even more than offset the loss in ?exibility from the
superior’s restriction to a single budget. The results of this study add to the understanding
of the interdependencies of con?icting budgeting purposes and contribute to explain why
?rms often use a single budget for multiple purposes.
Ó 2015 Elsevier Ltd. All rights reserved.
Introduction
Budgeting is one of ?rms’ most important coordination
and control mechanisms (Luft & Shields, 2003; Merchant &
Van der Stede, 2011). Usually, budgeting functions are
described as (1) operational planning and coordination,
(2) motivation and performance evaluation, and (3) goal
communication and strategy formulation (Atkinson,
Kaplan, Matsumura, & Young, 2011; Horngren, Datar, &
Rajan, 2011). Thus, budgeting functions encompass both
decision-making and control-oriented purposes of man-
agement accounting information. As these purposes oftenhttp://dx.doi.org/10.1016/j.aos.2015.02.002
0361-3682/Ó 2015 Elsevier Ltd. All rights reserved.
q
The authors greatly appreciate the helpful comments and suggestions
from Mike Shields (editor) and two anonymous reviewers, as well as from
Ramji Balakrishnan, Lynn Hannan, Tim Mitchell, Steve Salterio, Wim van
der Stede, Ivo Tafkov, seminar participants at the Universities of
Jönköping, Hannover, and Bern, and participants of the 2010 Management
Accounting Section midyear meeting, the 2009 AAA annual meeting, and
the 2009 and 2010 EAA annual meetings. We also thank Achim Hendriks
for his research assistance.
?
Corresponding author. Tel.: +49 (0) 541 969 2730.
E-mail addresses: [email protected] (M. C. Arnold),
[email protected] (R. M. Gillenkirch).
Accounting, Organizations and Society 43 (2015) 1–16
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
con?ict (Hopwood, 1972; Libby & Lindsay, 2010; Sprinkle,
2003), budget-setting can present ?rms with major
challenges.
Our study investigates how a planning task that con-
?icts with the performance evaluation task affects behav-
ior in budget negotiations and their outcomes. In our
setting, the two tasks are in con?ict because the need to
provide subordinates with suf?cient ?nancial incentives
counters the need for accurate planning and, thus, the
two budgets should be set to different levels. We further
analyze whether a single budget can be effectively used
for the two con?icting tasks as opposed to a separate bud-
get for each task because this restriction may cause more
cooperation in and after the budget negotiations.
Surprisingly little is known about the simultaneous use
of budgets for different purposes and the con?icts arising
from their interaction (Sprinkle, 2003). Prior literature sug-
gests that the budgeting purposes are not determined
independent of organizational characteristics (e.g.,
Moores & Yuen, 2001) and Shields and Shields (1998) ?nd
that uses of budget participation for decision-making and
control-oriented purposes are caused by different antece-
dents. However, most prior studies have focused on perfor-
mance evaluation in isolation and have analyzed
managers’ slack creating incentives (Brown, Evans, &
Moser, 2009; Shields & Shields, 1998; Sprinkle, 2003).
1
Only recently have studies begun to more closely analyze
interactions between budgeting functions. For example,
research suggests that ?exible budgets may be bene?cial
for planning but detrimental for performance evaluation
(Arnold & Artz, 2015; Ekholm & Wallin, 2011; Hansen &
Van der Stede, 2004). Additionally, increased budget dif?-
culty may help with performance evaluation but hurt the
communication function of budgets (Hansen & Van der
Stede, 2004). In contrast, positive externalities may arise
when a ?rm combines the resource-allocation and
performance-evaluation functions because a subordinate’s
monetary incentives to compete for resources can counter-
act his monetary incentives to create slack arising from
performance evaluation (Fisher, Maines, Peffer, & Sprinkle,
2002).
While the con?ict betweendifferent budgetary purposes
has been analyzed from an economic perspective (e.g.,
Baiman & Evans, 1983; Christensen, 1982) little is known
about their behavioral effects beyond monetary incentives.
Our study aims to contribute to the understanding of the
effects of budgets used for performance evaluation vs. per-
formance evaluation and planning. Fisher, Mitchell, Peffer,
and Webb (2014a) investigate subordinates’ budget reports
whenthe superior has a planning task and, inadditionto the
division’s net pro?t, receives payoffs from accurate fore-
casts. They ?nd that subordinates make higher cost reports
when they become aware of the superiors’ additional pay-
offs fromincreased accuracy but do not participate in them.
In a related paper, Fisher, Mitchell, Peffer, and Webb
(2014b) analyze how a superior’s reliance on cost reports
and her planning task performance depend on her
monetary incentives for planning accuracy when the sub-
ordinate is not aware of the superior’s planning task.
However, as in their research setting, the subordinate uni-
laterally determines the cost budget they do not investigate
budget negotiations. Moreover, as the superior has no
authority in determining the performance evaluation bud-
get, planning and control are unlikely to be perceived as
con?icting tasks for the superior in their setting.
In our setting, the budgeting purposes of performance
evaluation and planning are in con?ict from the superior’s
perspective because the need to provide subordinates with
suf?cient incentives to expend effort counters the need for
accurate planning. We compare a condition with only a
performance evaluation task to two conditions with an
additional planning task. In all three conditions, superior
and subordinate negotiate over the budget that is used
for performance evaluation. In the ?rst planning condition,
the superior can set a separate second budget for the plan-
ning task after the negotiation over the performance eval-
uation budget. In the second planning condition, the
superior is restricted to use a single budget for both tasks,
i.e., the planning budget automatically equals the perfor-
mance evaluation budget. As we focus on the subordinate’s
motives beyond monetary incentives, the subordinate’s
compensation function is never affected by the superior’s
planning task.
Because performance evaluation and planning are in
con?ict, budgets should be set to different levels in our set-
ting, and so a single budget cannot be used for both tasks
optimally. The superior therefore should be better off
when she can use separate budgets unless this entails sig-
ni?cant additional costs (Baiman, 1982; Barrett & Fraser,
1977; Churchill, 1984; Hopwood, 1972). This implication,
however, contrasts with empirical evidence (Merchant &
Manzoni, 1989; Umapathy, 1987) that ?rms rarely use dif-
ferent budgets for different purposes.
We develop theory to predict that with the additional
planning task, the subordinate is likely to perceive that
the superior has a more complex task and to have an
increased interpersonal accountability to the superior,
and that this makes the subordinate more cooperative in
the budget negotiation. We further predict that increased
subordinate cooperation extends to subordinate behavior
after the negotiation, leading to an increase in subordinate
performance in the presence of a planning task, particu-
larly after negotiation impasse. Finally, we also develop
theory to predict that the superior’s restriction to a single
budget can further increase the subordinate’s cooperation
during and after the negotiation.
In our experiment, participants assume the role of
either a superior or a subordinate and negotiate a budget
target for the subordinate’s performance evaluation. After
the negotiation, subordinates perform a real-effort task.
The experiment uses three settings: a baseline setting
where performance evaluation is the sole purpose of bud-
geting, a setting with separate budgets for planning and
performance evaluation, and a setting with a single budget
for both tasks. In the separate-budgets condition, planning
is operationalized by making the superior forecast the sub-
ordinate’s actual performance. In the single-budget condi-
tion, the performance evaluation budget simultaneously
1
Govindarajan (1986) measures both decision-making and control
dimensions of budgeting but collapses them into a single measure of ?rm
performance.
2 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
serves as the superior’s forecast. In both planning condi-
tions, the superior has to bear planning costs if actual per-
formance deviates from predicted performance.
Our results support our predictions. First, the superior’s
planning task increases subordinate cooperation during
the negotiation as evidenced by increased initial sub-
ordinate budget proposals. Second, after controlling for
monetary incentives, subordinate performance increases
in the presence of the planning task, particularly following
negotiation impasse. Third, when comparing the two plan-
ning conditions, the effects on subordinate behavior tend
to be even larger when the superior is restricted to a single
budget, particularly with respect to performance.
The paper makes two primary contributions to the
literature. First, to our knowledge, this is the ?rst study
focusing on the effects of the use of budgets for con?ict-
ing purposes on budget negotiations and their outcomes.
We thus respond to the call to investigate the interdepen-
dencies stemming from the simultaneous use of manage-
rial accounting information for decision-making and
control (Sprinkle, 2003). Our results provide evidence that
a superior’s additional task can affect subordinates
beyond monetary incentives in that subordinates coop-
erate more as their impact on the superior’s outcome
increases. Thus, cooperation in budget negotiations is
in?uenced not only by decreases in social distance
(Fisher, Frederickson, & Peffer, 2006) or by the informa-
tion distribution between the parties (Chalos & Haka,
1989) but also by the set of tasks and the restrictions
the superior has to deal with.
Second, our ?ndings help to reconcile the observation
that ?rms often use a single budget for multiple purposes
when economic theory predicts that they would employ
separate budgets for planning and control when these pur-
poses are in con?ict. Our results suggest that the bene?ts
from increased subordinate cooperation may compensate
for the loss in ?exibility from using a single budget. Thus,
in practice, the use of separate budgets vs. a single budget
will depend on the relative impacts of the two effects,
which implies that ?rms can pro?t by taking into account
the effects of social preferences when designing manage-
ment control systems (Church, Hannan, & Kuang, 2012;
Hannan, Rankin, & Towry, 2010).
Setting and benchmark analysis
Setting
Our setting is a modi?cation of the budget negotiation
setting of Fisher, Frederickson, and Peffer (2000). We study
superior-subordinate dyads in which the subordinate pro-
vides real effort. Prior to this effort, the parties negotiate a
budget for the subordinate’s performance evaluation, B
Eval
.
In this baseline setting, the superior has only the perfor-
mance evaluation task (‘‘performance evaluation-only con-
dition’’), and subordinate and superior receive the
following payoffs:
P
Subord
¼
F if X 6 B
Eval
F þ A Á ðX À B
Eval
Þ if X > B
Eval
ð1Þ
P
Super
¼
D Á X À F if X 6 B
Eval
D Á X À ½F þ A Á ðX À B
Eval
Þ? if X > B
Eval
ð2Þ
where F is the subordinate’s ?xed wage. X is the ?rm’s out-
put, which equals the subordinate’s actual performance. D
is the contribution margin per unit of output, and conse-
quently D Á X is the ?rm’s contribution margin that is
assigned to the superior. A is the bonus coef?cient per unit
of output for exceeding the performance evaluation bud-
get, and A Á (X À B
Eval
) is the subordinate’s bonus if X
exceeds B
Eval
. The subordinate’s performance is a function
of his ability and effort, and the superior is not informed
about her subordinate’s ability. Thus, when negotiating
the budget with the subordinate, the superior is uncertain
about the subordinate’s performance potential.
We also investigate two conditions in which the super-
ior has an additional planning task and her planning bud-
get, B
Plan
, is determined after the negotiation of B
Eval
. B
Plan
represents a forecast of actual performance, and deviations
between predicted and actual performance lead to plan-
ning costs for the superior. With the planning task, the
superior receives the following payoff:
P
super
¼
DÁ XÀF ÀCÁ jXÀB
Plan
j if X 6B
Eval
DÁ XÀ½F þAÁ ðXÀB
Eval
Þ? ÀCÁ jXÀB
Plan
j if X >B
Eval
ð3Þ
where B
Plan
is the superior’s forecast of actual performance
X in the production period and C is the coef?cient deter-
mining the superior’s planning costs when B
Plan
does not
perfectly predict actual output. B
Plan
re?ects a planned out-
put, such as the prediction of a sales volume for the instal-
lation of a production capacity, for procurement or supply-
chain coordination (Atkinson et al., 2011, chap. 10). We
study both a condition in which the superior can set a
separate second budget for the planning task after the
negotiation over the performance evaluation budget
(‘‘separate-budgets condition’’) and a condition in which
the superior must use a single budget for both tasks so that
the planning budget automatically equals the performance
evaluation budget (B
Eval
= B
Plan
: ‘‘single-budget condition’’).
Planning costs emerge from both under- and over-ful?ll-
ment of B
Plan
. If the planning budget is under-ful?lled, then
planning costs can be interpreted as, for example, costs of
unused capacity. If budgets are over-ful?lled, then extra
costs from operating at a level higher than expected arise,
such as, for example, running machines at higher speeds or
acquiring additional production capacity externally.
2
According to the superior’s payoff function (3), planning
costs are symmetric in that any deviation of actual output
X from planned output B
Plan
, whether positive or negative,
is equally costly to the superior.
3
We assume that
2
Consistent with prior literature (e.g., Balachandran, Balakrishnan, &
Sivaramakrishnan, 1997; Dunbar, 1971; Fisher, Frederickson, & Peffer,
2002), we refer to the consequences of inaccurate plans as costs because
they represent ex post inef?ciencies. Underlying is the perspective that the
superior could plan output optimally if the subordinate did not withhold or
misrepresent private information.
3
While the costs of unfavorable variances may be larger than the costs of
favorable variances (Horngren et al., 2011; Merchant & Manzoni, 1989), we
implemented symmetric planning costs in the experiment because the
symmetry in planning costs considerably facilitates the planning task for
the superior and increases experimental control.
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 3
D À A À C > 0, i.e., the superior always bene?ts from an
increase in output, even if this increase also increases the
subordinate’s bonus and the planning error. This implies
that the superior has monotonic incentives to both opti-
mally motivate the subordinate and correctly plan
performance.
The subordinate does not bear any planning costs, i.e.,
his payoff function (1) is unaffected by the planning task.
4
However, he is always informed about the superior’s plan-
ning task and planning costs. That is, at the time of the bud-
get negotiation, he is likely to be fully aware of the con?ict
between the two tasks, of his impact on the superior’s payoff
from the performance evaluation task and, in the planning
conditions, of the superior’s costs from planning errors.
This setting allows us to investigate how a superior’s
planning task affects the negotiation of a performance
evaluation budget and the outcome of the superiors’ tasks
as well as whether these effects are ampli?ed when the
superior is restricted to use a single budget for planning
and performance evaluation. Fig. 1 summarizes our
research model and relates our main variables of interest
to the whole budgeting process and its outcomes.
The upper part of Fig. 1 illustrates our hypothesis and
research question on negotiation behavior (H1 and RQ1).
We propose that the superiors’ planningtask motivates sub-
ordinates to act more cooperatively during the negotiation
and therefore increases initial budget proposals (H1 (i), left
node) and that the superiors’ restriction to a single budget
for both tasks ampli?es this effect (H1 (ii), right node).
Further, we analyze superiors’ initial counteroffers to
investigate whether and howthe planning task and the sin-
gle budget restriction in?uence a superior’s behavior in
negotiations about performance evaluationbudgets beyond
her reactions to subordinate initial proposals. As explained
in detail below, the potential direct and indirect effects of
the planning task on superiors’ initial counteroffers work
in different directions. We thus state a research question
(RQ1) to explore which of the effects dominates the super-
ior’s negotiation behavior in our experimental setting.
The lower part of Fig. 1 relates to the outcomes of the
superiors’ tasks as our second hypothesis and research
question (H2 and RQ2). Speci?cally, we propose that the
superior’s planning task makes the subordinate act more
cooperatively also after the negotiation and that this
increases performance, particularly after negotiation
impasse (H2 (i), left node). We further propose that this
effect is ampli?ed when the superior is restricted to a single
budget (H2 (ii), right node). H2 focuses on the behavioral
forces beyond monetary incentives, thereby controlling
for potentially different performance evaluation budgets.
Finally, we investigate how the single-budget restriction
in?uences the accuracy of the superior’s planning task. As
will be explained below, because the use of separate
budgets can have positive direct but also negative indirect
effects on planning accuracy we state a research question
(RQ2, right node) to investigate which forces affect plan-
ning accuracy more strongly in our experimental setting.
Benchmark analysis
Before deriving our hypotheses, we analyze the stan-
dard economic implications for the setting we have just
described. The superior’s dif?culty in setting budgets arises
from the information asymmetry about the subordinate’s
performance capability. The superior cannot expect the
subordinate to communicate his ability truthfully because
his payoff function induces slack-creating incentives
(Hansen, Otley, & Van der Stede, 2003; Murphy, 2000).
The subordinate will thus try to decrease his performance
evaluation budget and will strategically communicate dur-
ing the negotiation.
The superior’s symmetric planning costs imply that the
optimal planning budget should be set equal to the median
of the potential performance capabilities from the super-
ior’s perspective after the negotiation (Weitzman, 1976).
That means, if the superior cannot infer any additional
information about the subordinate’s capability from the
negotiation, then she will set the planning budget equal
to the median capability according to her ex ante expecta-
tions. In contrast, if, from the information obtained in the
negotiation, the superior can narrow the range of the sub-
ordinate’s potential capabilities compared to her ex ante
expectations, then she should optimally set the planning
budget to the median corresponding to this smaller range
of the potential capabilities.
In any case, however, when determining the perfor-
mance evaluation budget the superior has to take care that
the subordinate has suf?cient ?nancial incentives to
expend effort without increasing the compensation costs
(too much). If the performance evaluation budget were
equal to the median of potential capabilities (i.e., the opti-
mal planning budget), then subordinates could not reach
the performance evaluation budget in 50% of the cases.
This could be optimal only if the superior had extreme
expectations about potential capabilities.
5
Otherwise, the
expected loss in subordinate performance due to insuf?cient
?nancial incentives would be too large. Thus, the superior
should set the performance evaluation budget below the
optimal planning budget (the median capability) to increase
the likelihood of higher effort.
6
That means, in our setting,
performance evaluation and planning are in con?ict because
of the need to provide subordinates with suf?cient mone-
tary incentives to counter the need for accurate planning.
4
The subordinate does not participate in the superiors’ planning costs
for two reasons: First, this design choice increases experimental control
because it excludes any direct effect of the experimental conditions on the
subordinate’s payoff and because the pay scheme (1) is easily understand-
able. Second, while there are theoretical analyses of truth-inducing pay
schemes in which subordinates directly participate in planning accuracy
(e.g., Groves & Loeb, 1975; Weitzman, 1976), these pay schemes are rarely
found in practice (Waller & Bishop, 1990).
5
For example, if the superior expects the task to require some special
skills that half of the subordinates do not possess (i.e., their capabilities are
very low), the cost of reduced effort would be small for this group of
subordinates. In this case, the performance evaluation budget would be
optimally set close to or even above the median. Assuming that such a large
fraction of subordinates is not quali?ed for their task, though, does not
describe corporate practice.
6
This likely corresponds to actual company settings where incentive
systems usually provide suf?cient monetary incentives for the majority, not
the minority, of the employees (Baker, Jensen, & Murphy, 1988; Merchant,
2010).
4 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
Because the two budgets should be set to different
levels, the superior is better off when she can use separate
budgets unless this entails signi?cant additional costs. The
reason is that a single budget cannot solve both tasks opti-
mally at the same time (Barrett & Fraser, 1977; Hopwood,
1972; Otley, 1982). Consequently, depending on the rela-
tive importance of the two tasks and on where the single
budget is set, economic theory predicts that the restriction
to a single budget leads to increased costs from suboptimal
planning, suboptimal performance evaluation, or both.
These implications re?ect the established result that plan-
ning and performance evaluation are often con?icting bud-
geting purposes (Barrett & Fraser, 1977; Merchant &
Manzoni, 1989; Merchant & Otley, 2007) and economics
theories that predict ?rms should use separate budgets
for these different purposes (Baiman, 1982; Churchill,
1984).
Hypotheses and research questions
In contrast to the standard economic benchmark analy-
sis, we propose that behavioral factors, beyond monetary
incentives, also in?uence budget negotiations and subse-
quent performance. We suggest that with the planning
task, the subordinate is likely to perceive that the superior
has a more complex task owing to higher cognitive
demands (Bonner, 1994; Campbell, 1988) and to have an
increased interpersonal accountability to the superior
owing to his increased impact on the superior’s com-
pensation due to the potential planning costs (Kramer,
Pommerenke, & Newton, 1993). This is expected to result
in increased subordinate cooperation during and after the
negotiation. The theory we draw on to develop our
hypotheses relies on the existence of only some con?ict
between planning and performance evaluation from an
economic perspective and on the subordinate perceiving
this con?ict.
Negotiation process
Prior research suggests that a subordinate’s behavior
during and after budget negotiations is affected by fairness
concerns (Chalos & Haka, 1989; Fisher, Maines, et al.,
2002). As fairness concerns are likely to change with the
negotiation setting and with the parties’ interdependence
(De Dreu, Weingart, & Kwon, 2000; Giebels, De Dreu, &
Van de Vliert, 2000), they may be in?uenced by the super-
ior’s planning task in addition to her performance evalua-
tion task. As pointed out above, a con?icting planning
task likely increases task complexity, thereby making the
superior’s situation more dif?cult, and also increases the
subordinate’s impact on the superior’s outcome. These
effects are likely to be important in the negotiation as a
factor determining negotiation behavior is an individual’s
accountability to the counterpart (Lerner & Tetlock, 1999;
Tetlock, 1985). Accountability consists of two parts:
responsibility for another person’s outcome and an audi-
ence (a third person or the individual himself) evaluating
the individual’s decision (Lerner & Tetlock, 1999;
Schlenker, Britt, Pennington, Murphy, & Doherty, 1994).
Fig. 1. Research model and predictions.
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 5
With the planning task in addition to the performance
evaluation task during budgeting, the subordinate’s behav-
ior both during and after the budget negotiation increases
his impact on the superior’s outcome and therefore is likely
to increase his interpersonal accountability to the superior
(Kramer et al., 1993; Schlenker et al., 1994). The reason is
that planning quality and thus the superior’s planning
costs depend on her knowledge of the subordinate’s ability
and on his performance after the negotiation. If an individ-
ual’s interpersonal accountability to his or her counterpart
increases, then the individual is likely to cooperate more
and compete less (Kramer et al., 1993; Pruitt, 1981).
Therefore, the subordinate should cooperate more in the
negotiation when the superior has a planning and a perfor-
mance evaluation task.
7
Prior research on negotiations in
general (Benton, Kelley, & Liebling, 1972; Carnevale &
Pruitt, 1992) and on budget negotiations in particular
(Fisher et al., 2000; Fisher et al., 2006) suggests that less
extreme initial offers can be regarded as a signal for
cooperation. Thus, we expect the subordinates’ initial bud-
get proposals to increase when there is a planning task in
addition to the performance evaluation task.
8
Compared to the case in which the planning task can be
solved through a separate planning budget, the superior’s
restriction to a single budget for both tasks is likely to
increase the subordinate’s interpersonal accountability
even more. In this case, the superior’s ability to solve the
planning task is further restricted because the planning
budget equals the performance evaluation budget. That
means, any decrease in the performance evaluation budget
would not only increase compensation costs but also
automatically increase the superior’s planning costs
because she has no possibility to adjust the planning bud-
get. Consequently, the superior’s restriction to a single
budget for both tasks is likely to reinforce the effect on
the subordinates’ initial proposals. We predict:
H1. The subordinate’s initial budget proposal in the
negotiation (i) increases when the superior has a planning
and performance evaluation task and (ii) further increases
when the superior is restricted to the use of a single budget
for planning and performance evaluation instead of sepa-
rate budgets for each task.
In contrast to the effects of planning on subordinate
behavior, the effects on the superior are more ambiguous.
Numerous studies have illustrated the importance of
reciprocity for negotiations (e.g., Brett, Shapiro, & Lytle,
1998; Kuang & Moser, 2009; Putnam & Jones, 1982).
According to the norm of reciprocity, a party will cooperate
during a negotiation if the other party also does so (Benton
et al., 1972; Carnevale & Pruitt, 1992). Thus, if the super-
ior’s planning task makes the subordinate increase his ini-
tial bids, then the superior may reciprocate and this may
lead to lower initial superior counteroffers. Similarly, if
superiors react reciprocally and the single-budget restric-
tion further increases initial subordinate bids, then it
should further decrease initial superior counteroffers.
However, prior research also suggests that the superi-
ors’ counteroffers may be affected by anchoring in two dif-
ferent ways. First, an initial offer can serve as an anchor
that biases the counteroffer (Bazerman & Neale, 1993;
Kristensen & Gärling, 1997). This implies that higher initial
subordinate bids under planning and the additional
increase in the bids under the single-budget case may
increase superiors’ initial counteroffers. Second, anchors
for negotiation offers also may arise from other salient
information (Kahneman, 1992). With an additional plan-
ning task, a likely anchor for the superior’s initial coun-
teroffer is the optimal planning budget given the
superior’s information before the negotiation. As analyzed
above, the optimal planning budget in this case is the med-
ian of the potential performance capabilities according to
the superior’s ex ante expectations. Whether anchoring
on the ex ante optimal planning budget leads to higher
or lower counteroffers is unclear as the direction of the
effect depends on the level of initial counteroffers without
the additional planning task. Yet, in this case, superiors’
initial counteroffers should be concentrated more nar-
rowly around the expected performance capabilities than
without the planning task. As the potential behavioral
forces affecting the counteroffers point into different direc-
tions, we do not derive a directional hypothesis but use the
following research question to investigate which effect
dominates superior behavior in budget negotiations in
our experimental setting:
RQ1: How do the planning task and the restriction to
use a single budget for planning and performance eval-
uation affect a superior’s initial counteroffer in the
negotiation?
Subordinate performance
A characteristic of a budget negotiation is that subse-
quent to the negotiation, the subordinate provides non-
contractible effort. Thus, an important question is the
effect of the negotiation’s procedures and results on sub-
ordinate performance. In the following, we develop theory
to predict that the planning task leads to more subordinate
cooperation not only during, but also after the negotiation.
This increases the subordinate’s performance, particularly
after a negotiation impasse. Further, we expect the
restriction to a single budget for both tasks to reinforce
the effect.
7
Increased cooperation in the negotiation may also incorporate aspects
of strategic behavior. First, the subordinate may make higher offers because
he expects the superior to positively reciprocate his increased cooperation.
Second, the subordinate may anticipate the in?uence of the planning task
on the superior’s negotiation behavior. For example, if the subordinate
expects the planning task to reduce the superior’s concessions, he may
increase his offers to avoid negotiation impasses. However, whether
cooperation increases due to genuine or strategic concerns is largely
semantic and does not affect our predictions as long as the increase is
systematic. We provide evidence on this question in the supplemental
analysis.
8
Research on bargaining theory shows that a negotiation party can have
incentives to communicate private information when bargaining impasse is
costly to her and when she expects her counterpart to make no or few
concessions (e.g., Ausubel, Cramton, & Deneckere, 2002, chap. 50;
Rubinstein, 1985). In our setting, this might make subordinates’ offers
more informative about their capability when the superior has a planning
task. We provide evidence on this issue in the results section.
6 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
Prior research has shown that reciprocal reactions to
the superior’s negotiation behavior strongly affect sub-
ordinates’ effort choices after the negotiation (Fisher,
Frederickson, et al., 2002; Kuang & Moser, 2011). These
reciprocal reactions are more positive when superiors are
constrained as their intentions are perceived more posi-
tively in this case (Hannan, 2005). Because the presence
of a con?icting planning task is likely to make the sub-
ordinate perceive the superior’s task as more complex
and constrained, the planning task can positively affect
subordinate performance.
We propose, however, that this effect is likely to be
more pronounced after negotiation impasse than after
agreement for the following reason: If the two parties
reach an agreement, then the subordinate’s aspiration level
is likely to be ful?lled (Kuang & Moser, 2011).
Consequently, although the superior’s negotiation behav-
ior is likely perceived as more positive when a planning
task is present in addition to the performance evaluation
task, the effect of the planning task on subordinate perfor-
mance is likely to be smaller under agreement. Prior
research supports this conjecture, showing that, when
agreement is reached, the effects of all treatment condi-
tions that have been manipulated so far in similar budget
negotiation settings on subordinate performances have
been small (Arnold, 2015; Fisher et al., 2000; Fisher,
Frederickson, et al., 2002; Fisher et al., 2006).
In contrast, the effect of the planning task on perfor-
mance is likely to be larger after a budget imposition when
negotiations fail because the planning task is likely to
affect the subordinate’s attribution of the superior’s budget
imposition and therefore his interpretation of the budget
imposition. Attribution theory suggests that an important
determinant of an individual’s reaction to another person’s
actions is the attribution of that person’s behavior to exter-
nal (situational) or internal (dispositional) factors (Dobbins
& Russell, 1986; Gilbert & Malone, 1995; Ross, 1977). In the
unconstrained performance evaluation-only setting, the
subordinate is likely to attribute the superior’s budget
imposition to dispositional factors. Thus, he is likely to per-
ceive a budget imposition as unfair and to respond with
low effort (Fisher et al., 2000; Kuang & Moser, 2011). In
contrast, when the superior also has a planning task, the
subordinate cannot make unambiguous inferences about
the superior’s dispositions and may explain her behavior
by the situation rather than by disposition (Arnold, 2015;
Gilbert & Malone, 1995). That is, a budget imposition
may be more acceptable to the subordinate in a setting
with multiple budgeting tasks than in the performance
evaluation-only task, because the subordinate may attri-
bute the lack of suf?cient superior concessions during
the negotiation to the superior’s increased task complexity
and her focus on the additional planning task. Thus, the
increasing effect of the planning task on subordinate
performance is likely to be larger following a budget
imposition.
Again, the effect of the planning task on performance is
likely to be ampli?ed with a single budget for both tasks
because the superior is even more constrained and, thus,
her intentions are likely to be perceived even more posi-
tively. Speci?cally, a budget imposition may be more
justi?able with a single budget as the superior may have
no alternative to imposing a budget to keep the planning
error small. Thus, we predict the following effect of the
planning task and the restriction to a single budget on sub-
ordinate performance:
H2. Controlling for the subordinate’s monetary incentives,
(i) the positive effect of the superior’s planning task
beyond a performance evaluation task only on subordinate
performance is larger after negotiation impasse than after
agreement and (ii) the superior’s restriction to a single
budget ampli?es this effect.
Planning error
As we analyzed above, according to economic theory,
the possibility of setting separate budgets should reduce
the planning error, because the superior does not have to
trade off the con?icting tasks in a single budget.
However, from the analysis leading to H1 and H2, the
effects of the superior’s restriction to a single budget on
her planning error are unclear under both agreement and
impasse. First, after agreement, the superior cannot further
adjust the budget for the planning task in the single-
budget case. In contrast, in the separate-budgets case, the
superior can adjust her planning budget, but a reduction
in planning error may be dif?cult if the subordinate coop-
erates less during the negotiation and hides his perfor-
mance capability. Further, as predicted in H1 (ii), the
subordinates’ initial budget proposals are likely to be
higher when the superior is restricted to a single budget,
and this could lead to higher performance evaluation bud-
gets under agreement. Consequently, the performance
evaluation budget under agreement may be closer to the
subordinate’s actual performance when the superior is
restricted to a single budget.
Second, after negotiation impasse, the superior has to
trade off planning and performance evaluation in the sin-
gle-budget case and therefore may not be able to solve
the planning task as well as in the separate-budgets case.
In contrast, the arguments for H2 (ii) imply that sub-
ordinate performance is likely to be lower in the sepa-
rate-budgets case. Consequently, if the superior does not
perfectly anticipate the lower performance,
9
then the plan-
ning error after negotiation impasse may be larger with
separate budgets than with a single budget. As the potential
effects of the behavioral forces on planning accuracy point in
different directions, we state a research question to investi-
gate which of them dominates in our experimental setting:
RQ2: Will the realized planning error be different under
agreement and disagreement when the superior is
restricted to the use of a single budget than when she
can set separate budgets?
9
Prior experimental evidence shows that decision makers’ abilities to
anticipate the behavior of others in sequential bargaining games are very
limited. See, e.g., Binmore, McCarthy, Ponti, Samuelson, and Shaked (2002),
Johnson, Camerer, Sen, and Rymon (2002).
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 7
Experimental method
Participants and design
180 graduate and undergraduate students from a large
European university (90 dyads) participated in the experi-
ment—60 participants (30 dyads) in each of the three treat-
ments. Of the participants, 42% were female, and 59% were
business or economics majors. Three sessions were con-
ducted for each of the treatments. No participant partici-
pated in more than one session. Sessions lasted between
60 and 75 min. Participants received a show-up fee of €
5.00 and additional variable compensation. Average total
compensation was €14.34 and varied between €5.00 and
€23.80.
We used a nested experimental design in which we var-
ied (between subjects) the existence of a planning task—
present or absent. Nested within the planning task present,
we manipulated the restriction to use a single budget for
both purposes—present or absent. In all treatments, bud-
gets had a performance evaluation function. This nested
design resulted in three treatments: performance evalua-
tion only, separate budgets for planning and performance
evaluation (‘‘separate-budgets treatment’’), and a single
budget for planning and performance evaluation (‘‘single-
budget treatment’’). The experimental design is summar-
ized in Fig. 2.
The planning task was operationalized by having the
superior predict output, i.e., actual performance, or not.
With the planning task, the superior had to bear planning
costs if actual output deviated from predicted output, as
included in Eq. (3).
The restriction to a single budget was manipulated by
giving the superior the ability to set separate budgets for
performance evaluation and planning or not. That is, in
the single-budget treatment, the planning budget
automatically equaled the performance evaluation budget.
The subordinate’s payoff function was never affected by
any manipulation. However, as the subordinate’s ?nancial
incentives after the negotiation depended on the perfor-
mance evaluation budget, we control for these incentives
in our statistical analysis.
Procedure
The experiment, including the negotiation process, was
computerized. Instructions appeared on the computer
screens and were simultaneously read aloud.
10
The experi-
mental procedure consisted of the following steps:
(1) Upon arrival, participants were randomly assigned
to computer terminals and were separated from
each other by panels. Superior and subordinate were
paired anonymously and randomly.
(2) Then, the real effort task was explained. It consisted
of decoding two-digit numbers into letters in blocks
of two each. Participants then completed a two-min-
ute practice round.
(3) Next, all subjects completed three four-minute
training rounds. New decoding keys were dis-
tributed for each round. Subordinates were told that,
at the end of the experiment, one of the training
rounds would be randomly selected and they would
receive two points for every block correctly decoded
in this round. Subordinates were informed of the
number of correctly and incorrectly decoded blocks
after each training round. They improved their per-
formance on average by 2.18 blocks from training
rounds 1–2 and by 1.56 blocks from training rounds
2–3. Superiors were not compensated and received
no feedback in the training rounds.
(4) After the third training round, subordinates were
asked to submit a best estimate of the number of
blocks they could correctly decode in another four-
minute work round. On average, the estimate was
larger by 0.34 blocks than the maximum perfor-
mance in the training rounds.
(5) Next, participants were informed about their roles,
the setting, the payoff functions, and the budgeting
process in the experimental work round. The
parameter values were set to F = 250, A = 4, D = 10,
and C = 4.
11
They guarantee that the superior has
monotonic incentives to motivate high subordinate
performance and predict this performance accurately.
Speci?cally, for a given subordinate performance, any
unit of deviation of the planning budget decreases
the superior’s payoff by C = 4. Further, for a given
planning budget, the superior’s payoff increases mono-
tonically if subordinate performance increases. That
is, even if an increase in subordinate performance
increased both compensation and planning error,
the net effect on the superior’s payoff would still be
positive (D À A À C = 2). A and C were chosen to avoid
a parameter-speci?c focus on one of the tasks: With
A = C, an increase in compensation is as costly as an
increase in planning error. The values of A and C are
large enough to make the superior pay suf?cient
attention to both tasks. Participants were informed
that the negotiation would last at maximum three
rounds and that the superior would set the ?nal per-
formance evaluation budget if no agreement was
reached at the end of the third round. In the single-
budget treatment, participants were further told that
the planning budget would be equal to the perfor-
mance evaluation budget. In every negotiation round,
the subordinate made the initial budget proposal, and
the superior made the counteroffer, which resembles
the structure of budgeting negotiations in practice
(Anthony & Govindarajan, 2007; Umapathy, 1987).
10
The instructions used context-free language. Moreover, to avoid that in
the planning conditions, one task was framed as pro?t increasing, and the
other was framed as pro?t decreasing, we treated the effects of the two
tasks always identically. The instructions are available from the authors
upon request.
11
The minimum compensation for the subordinates was F. To avoid
having to pay the ?xed wage from their own resources, superiors were
endowed with 500 points. The remaining 250 points after the deduction of
F ensured that the superior would have suf?cient planning incentives even
in the case of low subordinate performance.
8 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
(6) Subjects then answered a quiz about the procedures
and the payoffs. The quiz included questions about
the planning budget (planning treatments only)
and the performance-evaluation budget (all treat-
ments). These questions were correctly answered
in 92% of all cases in the ?rst trial. Participants fail-
ing to give a correct answer had to read the relevant
instructions again. The experiment did not start
until all participants had answered all questions
correctly.
(7) Before the negotiations started, all participants were
informed of the results of a pretest. They were told
that 80% of the participants in the pretest had cor-
rectly decoded between 29 and 49 blocks of num-
bers in four minutes. No superior was informed
about the performance of her speci?c subordinate
in the training rounds.
(8) Next, negotiations were conducted between superi-
ors and subordinates via the computer network.
This procedure ensured anonymity and removed
any interpersonal aspects of the negotiation. At any
point in the negotiation, either party could accept
the other party’s proposal, which would end the
negotiation. If the superior and subordinate could
not reach an agreement within three rounds, then
the superior imposed the budget.
(9) After the performance evaluation budget was set,
each superior set the planning budget in the
separate-budgets treatment. Subordinates were not
informed about the planning budget so as not to
affect their performance by presenting an additional
goal. This is consistent with the informal adjustment
of performance evaluation targets for planning pur-
poses observed in practice (Berry and Otley, 1975;
Walker & Johnson, 1999). In the single-budget treat-
ment, the planning budget automatically equaled
the performance evaluation budget.
(10) Finally, subordinates completed the experimental
work round for four minutes. Subjects interacted
for only one work round to avoid ambiguities
that may arise in repeated relationships, such as
reputation building and aspects of the social
interaction. At the end of the work round, subjects
were informed of the subordinate performance
and the payoffs. They completed a post-
experimental questionnaire and received their cash
payments.
Measures
To analyze the negotiation process, we use FIRSTBID,
the subordinate’s initial budget proposal in the negotia-
tion, and FIRSTCOUNTER, the superior’s initial counterof-
fer. B
Eval
and B
Plan
represent the performance evaluation
and the planning budget, respectively. X is the sub-
ordinate’s actual performance, and
^
X is the subordinate’s
estimate of his performance capability at the end of the
training rounds. BONUS
Pot
¼ maxf0; 4 Á ð
^
X ÀB
Eval
Þg repre-
sents the potential subordinate bonus based on his esti-
mated performance capability and is used to control for
the subordinate’s monetary incentives. ERROR = |X À B
Plan
|
is used to measure the planning error. Table 1 summarizes
the measures used in the experiment.
Results
Table 2 displays descriptive statistics and shows that
the subordinates’ estimated performance capabilities
^
X
are very similar across treatments. One-way ANOVAs show
no signi?cant difference across treatments with respect to
the subordinates’ estimates and the subordinates’ maxi-
mum performances in the training round (p > .10 in both
cases). Thus, differences across treatments are not driven
by differences in performance capabilities.
As our goal is to establish a con?ict between planning
and performance evaluation, we examine the means of
B
Plan
and B
Eval
. Table 2 shows that, in the separate-budgets
treatment, mean B
Eval
is lower than mean B
Plan
and this dif-
ference is signi?cant (t-test, x ¼ 35:07 vs. 38.70, t = 3.11,
p < .01).
12
63% of the superiors adjusted the planning budget
upwards, consistent with the economic benchmark analysis.
In contrast, 27% of the superiors adjusted the planning bud-
get downwards. The majority of these cases (eight of nine)
occurred after a negotiation impasse, and the mean perfor-
mance evaluation and planning budget in these cases are
36.85 and 32.57, respectively. This result suggests that,
when setting their planning budget, some superiors allowed
for the possibility that subordinates might reduce perfor-
mance after a budget imposition. However, our ?nding that
90% of the superiors adjusted the planning budget con?rms
Performance evaluation only Planning and performance evaluation
Separate budgets Single budget
(Treatment 1) (Treatment 2) (Treatment 3)
Superior has only a
performance evaluation task
Superior has a planning and a performance evaluation task
Superior can use separate
budgets for each task
Superior is restricted to use a
single budget for both tasks
B
Eval
B
Eval
and B
Plan
B
Eval
= B
Plan
Fig. 2. Overview of the experimental setting and design.
12
We use one-tailed p-values for tests of directional predictions and two-
tailed p-values for research questions without directional prediction. The
unit of observation is the observation made for every dyad.
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 9
the success of our manipulation of con?ict between the
performance evaluation and the planning tasks.
Finally, Table 2 reveals that the mean of single budget is
close to and not signi?cantly different from the mean of
B
Eval
in the separate-budgets case (t-test, x ¼ 35:76 vs.
35.07, t = .50, p > .10) but signi?cantly lower than B
Plan
(t-test, x = 35.76 vs. 38.70, t = 2.22, p = .03). Thus, in our
experiment, performance evaluation seems to be more
important for superiors than planning.
13
Hypotheses tests
H1 predicts (i) an increase in initial budget proposals in
the planning treatments and (ii) an additional increase in
the single-budget relative to the separate-budgets treat-
ment. Consistent with these two predictions, Table 2 and
Fig. 3 show that the mean of FIRSTBID is higher in the
two planning treatments than in the performance evalua-
tion-only treatment (H1 (i)) and is higher in the single-
budget treatment than in the separate-budgets treatment
(H1 (ii)).
We conduct a nested ANOVA with FIRSTBID as the
dependent variable to test H1. The ?rst factor in the
ANOVA is whether the planning task is present or absent.
This factor tests part (i) of H1. Nested within this factor,
the second factor captures whether the superior is
restricted to a single budget. This factor tests part (ii) of
H1. As reported in Panel A of Table 3, planning signi?cantly
affects FIRSTBID (p = .03). However, even though the effect
of the single budget on FIRSTBID is in the predicted direc-
tion the nested term is not signi?cant at conventional
levels (p = .13). These results support part (i) of H1 but
not part (ii).
14
RQ1 asks how the planning task and the restriction to a
single budget affect superiors’ initial counteroffers. Table 2
and Fig. 3 show that mean FIRSTCOUNTER is very similar
across treatments and slightly above the mid-point (39)
of the pre-test span (29–49) participants were informed
about. We ran a nested ANOVA and a nested ANCOVA with
FIRSTCOUNTER as the dependent variable and the same
factors as used for testing H1 to test whether the
planning task affects FIRSTCOUNTER. In the ANOVA,
we test the overall treatment effects. In the ANCOVA,
we include FIRSTBID as a covariate to examine whether
superior counteroffers are driven by reciprocity or
anchoring on subordinates’ initial bids. If superiors react
reciprocally, then increases in FIRSTBID should decrease
FIRSTCOUNTER. In contrast, if superiors anchor
on FIRSTBID, increases in FIRSTBID should increase
FIRSTCOUNTER. Panel B of Table 3 reports the results.
They do not reveal any signi?cant effect in the ANOVA or
the ANCOVA (p > .25 in all cases). That is, we ?nd neither
a signi?cant effect of the treatment conditions nor of
FIRSTBID. This lack in signi?cant results contradicts the
view that superiors reciprocate or anchor on the sub-
ordinates’ initial bids.
Table 1
Summary of the measures used in the analysis.
Measure Description
FIRSTBID Subordinate’s initial budget
proposal
FIRSTCOUNTER Superior’s initial counteroffer
B
Eval
Performance evaluation
budget
B
Plan
Planning budget
X Subordinate performance
^
X
Subordinate’s estimate of his
performance capability
BONUS
Pot
¼ maxf0; 4 Á ð
^
X À B
Eval
Þg
Potential subordinate bonus
based on the subordinate’s
estimated performance
capability
ERROR = |X À B
Plan
| Realized planning error
Table 2
Descriptive statistics: mean and standard deviation.
Performance
evaluation only
Planning and performance
evaluation
Separate
budgets
Single
budget
(Treatment 1) (Treatment 2) (Treatment 3)
^
X
a
41.67 38.53 40.23
5.08 6.30 6.56
B
Eval
b
32.33 35.07 35.76
11.67 5.95 4.77
B
Plan
c
n.a. 38.70 35.76
n.a. 5.45 4.77
BONUS
Pot
d
43.47 20.27 23.60
41.39 19.95 25.13
FIRSTBID
e
19.93 22.20 24.53
8.79 8.15 6.28
FIRSTCOUNTER
f
40.53 40.27 41.67
17.20 6.99 8.66
X
g
30.80 28.87 37.00
19.21 19.19 14.71
ERROR
h
n.a. 14.10 12.83
n.a. 15.88 11.68
No. of obs. 30 30 30
Note: Mean and standard deviation for the performance evaluation-only
condition (Treatment 1), planning and performance evaluation with
separate budgets (Treatment 2), and planning and performance evalua-
tion with a single budget (Treatment 3).
a
^
X is the subordinate’s estimate of his performance capability.
b
B
Eval
is the budget used for performance evaluation.
c
B
Plan
is the budget used for planning. In the single-budget case, B
Plan
automatically equals B
Eval
.
d
BONUS
Pot
is the potential bonus the subordinate can earn based on his
estimated performance capability
^
X.
e
FIRSTBID is the subordinate’s initial budget proposal in the
negotiation.
f
FIRSTCOUNTER is the superior’s initial counteroffer in the negotiation.
g
X is the subordinate performance in the experimental work round.
h
ERROR = |X À B
Plan
| is the realized planning error.
13
This ?nding is consistent with the perception of performance evalua-
tion as central in practice (e.g., Otley, 1978). However, it may not be
generalizable because the optimal single budget also depends on the
parameters A and C.
14
We ?nd no evidence consistent with subordinates’ budget proposals
revealing more about their capabilities in the planning conditions. Speci?-
cally, regression analyses show that the increase in budget proposals in the
planning conditions varies neither linearly nor non-linearly (inversely
U-shaped) with subordinates’ capabilities. These results contradict that the
increased budget proposals are driven by an increased informativeness of
subordinate offers.
10 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
Finally, a test of superiors’ anchoring on the ex ante
optimal planning budget requires a measure of the concen-
tration of FIRSTCOUNTER around the mid-point of the pret-
est span. Therefore, we used the standard deviation of
FIRSTCOUNTER. Table 2 reveals that the standard deviation
is much smaller in both planning treatments than in the
performance evaluation-only treatment, and this decrease
in standard deviation of FIRSTCOUNTER in the two plan-
ning treatments relative to the performance evaluation-
only treatment is signi?cant (Bartlett’s test, v
2
> 12,
p < .01 in both cases). This result is consistent with the
notion that in our experimental setting, superiors form
an a priori expectation about subordinates’ performance
capabilities and anchor on this expectation when a plan-
ning task is present.
H2 predicts that, after controlling for monetary incen-
tives, the superior’s planning task in addition to her perfor-
mance evaluation task leads to a larger increase in
subordinate performance after negotiation impasse than
after agreement (H2 (i)). It further predicts that this effect
is ampli?ed when the superior is restricted to a single bud-
get (H2 (ii)). Panel A of Table 4 displays the descriptive
statistics for subordinates’ performances contingent on
negotiation agreement. Fig. 4 graphs the performances
relative to the estimated performance capabilities in all
experimental treatments. Table 4 and Fig. 4 show that
mean performance increases in the separate-budgets treat-
ment relative to the performance evaluation-only treat-
ment and further increases in the single-budget
treatment. As predicted, both effects are mainly driven
by the changes after budget impositions, whereas after
agreement the performances are similar across treatments.
To test H2, we run regressions with the subordinate
performance, X, as the dependent variable.
15
We use
regressions instead of ANOVAs because a Levene test indi-
cates a violation of variance homogeneity (t = 22.69,
p < .01). In the regressions, we control for heteroscedasticity
by using the Huber–White estimator. To test the interaction
predicted by part (i) of H2, we regress X on BONUS
Pot
, an
indicator variable for planning (DPlan equal to 1 in the plan-
ning treatments and 0 in the performance evaluation-only
treatment), an indicator variable for agreement (agree equal
to 1 when agreement was reached and 0 when the negotia-
tion failed), and the interaction of the two indicator vari-
ables. To test part (ii) of H2, we include only the two
planning treatments and regress X on BONUS
Pot
, Agree, an
indicator variable for the singe-budget treatment (DSingle
equal to 1 in the single-budget treatment and 0 in the sepa-
rate-budgets treatment), and the interaction of the two indi-
cator variables. In the regressions, BONUS
Pot
controls for the
effect of monetary incentives. Panel B of Table 4 reports the
results.
Consistent with H2 (i) and (ii), the interaction coef?-
cients are signi?cantly negative in both regressions. In
Model (1), the interaction coef?cient indicates that the
increase in subordinate performance (X) under planning
is smaller after agreement (b = À7.95, p = .09). This
supports H2 (i). In Model (2), the interaction coef?cient
indicates that the increase in X in the single budget-
treatment is smaller after agreement (b = À11.10, p = .04).
This result supports H2 (ii).
Moreover, the regressions allow analyzing separately
whether the changes in X are signi?cant both under
Fig. 3. Subordinates’ initial budget proposals and superiors’ initial
counteroffers. Note: The ?gure displays subordinates’ average initial
budget proposal in the negotiation (FIRSTBID) and superiors’ average
initial counteroffer (FIRSTCOUNTER) for the three experimental condi-
tions (performance evaluation-only, planning with separate budgets, and
planning with a single budget).
Table 3
Analysis of subordinates’ initial budget proposals and superiors’ initial
counteroffers.
df MS F p
Panel A: FIRSTBID
a
Planning 1 235.76 3.86 0.03
Single budget|planning 1 81.67 1.34 0.13
Error 87 61.04
Panel B: FIRSTCOUNTER
b
ANOVA
Planning 1 3.75 0.03 0.87
Single budget|planning 1 29.40 0.21 0.65
Error 87 139.90
ANCOVA
FIRSTBID 1 174.21 1.25 0.27
Planning 1 0.68
Budgets are often simultaneously used for the conflicting purposes of planning and performance
evaluation. While economic theory suggests that firms should use separate budgets
for conflicting purposes this contrasts with existing evidence that firms rarely do so. We
address two open questions related to these observations in an experiment. Specifically,
we investigate how a planning task that is in conflict with the performance evaluation task
affects behavior in budget negotiations and their outcomes. Additionally, we analyze
whether a single budget can be effectively used for both purposes compared to two separate
budgets. We develop theory to predict that adding a planning task that is in conflict
with the superior’s performance evaluation task increases the subordinate’s cooperation
in and after the negotiation of a performance evaluation budget.
Using negotiated budgets for planning and performance
evaluation: An experimental study
q
Markus C. Arnold
a
, Robert M. Gillenkirch
b,?
a
Institute for Accounting, University of Bern, 3012 Bern, Switzerland
b
Business Department, University of Osnabrück, 49069 Osnabrück, Germany
a r t i c l e i n f o
Article history:
Available online 21 March 2015
a b s t r a c t
Budgets are often simultaneously used for the con?icting purposes of planning and perfor-
mance evaluation. While economic theory suggests that ?rms should use separate budgets
for con?icting purposes this contrasts with existing evidence that ?rms rarely do so. We
address two open questions related to these observations in an experiment. Speci?cally,
we investigate how a planning task that is in con?ict with the performance evaluation task
affects behavior in budget negotiations and their outcomes. Additionally, we analyze
whether a single budget can be effectively used for both purposes compared to two sepa-
rate budgets. We develop theory to predict that adding a planning task that is in con?ict
with the superior’s performance evaluation task increases the subordinate’s cooperation
in and after the negotiation of a performance evaluation budget. Moreover, we predict that
subordinate cooperation increases even more when the superior is restricted to use a single
budget for both purposes. Our results broadly support our hypotheses. Speci?cally, we ?nd
that when budgets are used for both planning and performance evaluation, this increases
the subordinate’s budget proposals during the negotiation and his performance after the
negotiation. These effects tend to be even larger when the superior is restricted to a single
budget rather than separate budgets for planning and performance evaluation, particularly
with respect to subordinate performance. In our experimental setting, the bene?ts of
increased subordinate cooperation even more than offset the loss in ?exibility from the
superior’s restriction to a single budget. The results of this study add to the understanding
of the interdependencies of con?icting budgeting purposes and contribute to explain why
?rms often use a single budget for multiple purposes.
Ó 2015 Elsevier Ltd. All rights reserved.
Introduction
Budgeting is one of ?rms’ most important coordination
and control mechanisms (Luft & Shields, 2003; Merchant &
Van der Stede, 2011). Usually, budgeting functions are
described as (1) operational planning and coordination,
(2) motivation and performance evaluation, and (3) goal
communication and strategy formulation (Atkinson,
Kaplan, Matsumura, & Young, 2011; Horngren, Datar, &
Rajan, 2011). Thus, budgeting functions encompass both
decision-making and control-oriented purposes of man-
agement accounting information. As these purposes oftenhttp://dx.doi.org/10.1016/j.aos.2015.02.002
0361-3682/Ó 2015 Elsevier Ltd. All rights reserved.
q
The authors greatly appreciate the helpful comments and suggestions
from Mike Shields (editor) and two anonymous reviewers, as well as from
Ramji Balakrishnan, Lynn Hannan, Tim Mitchell, Steve Salterio, Wim van
der Stede, Ivo Tafkov, seminar participants at the Universities of
Jönköping, Hannover, and Bern, and participants of the 2010 Management
Accounting Section midyear meeting, the 2009 AAA annual meeting, and
the 2009 and 2010 EAA annual meetings. We also thank Achim Hendriks
for his research assistance.
?
Corresponding author. Tel.: +49 (0) 541 969 2730.
E-mail addresses: [email protected] (M. C. Arnold),
[email protected] (R. M. Gillenkirch).
Accounting, Organizations and Society 43 (2015) 1–16
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
con?ict (Hopwood, 1972; Libby & Lindsay, 2010; Sprinkle,
2003), budget-setting can present ?rms with major
challenges.
Our study investigates how a planning task that con-
?icts with the performance evaluation task affects behav-
ior in budget negotiations and their outcomes. In our
setting, the two tasks are in con?ict because the need to
provide subordinates with suf?cient ?nancial incentives
counters the need for accurate planning and, thus, the
two budgets should be set to different levels. We further
analyze whether a single budget can be effectively used
for the two con?icting tasks as opposed to a separate bud-
get for each task because this restriction may cause more
cooperation in and after the budget negotiations.
Surprisingly little is known about the simultaneous use
of budgets for different purposes and the con?icts arising
from their interaction (Sprinkle, 2003). Prior literature sug-
gests that the budgeting purposes are not determined
independent of organizational characteristics (e.g.,
Moores & Yuen, 2001) and Shields and Shields (1998) ?nd
that uses of budget participation for decision-making and
control-oriented purposes are caused by different antece-
dents. However, most prior studies have focused on perfor-
mance evaluation in isolation and have analyzed
managers’ slack creating incentives (Brown, Evans, &
Moser, 2009; Shields & Shields, 1998; Sprinkle, 2003).
1
Only recently have studies begun to more closely analyze
interactions between budgeting functions. For example,
research suggests that ?exible budgets may be bene?cial
for planning but detrimental for performance evaluation
(Arnold & Artz, 2015; Ekholm & Wallin, 2011; Hansen &
Van der Stede, 2004). Additionally, increased budget dif?-
culty may help with performance evaluation but hurt the
communication function of budgets (Hansen & Van der
Stede, 2004). In contrast, positive externalities may arise
when a ?rm combines the resource-allocation and
performance-evaluation functions because a subordinate’s
monetary incentives to compete for resources can counter-
act his monetary incentives to create slack arising from
performance evaluation (Fisher, Maines, Peffer, & Sprinkle,
2002).
While the con?ict betweendifferent budgetary purposes
has been analyzed from an economic perspective (e.g.,
Baiman & Evans, 1983; Christensen, 1982) little is known
about their behavioral effects beyond monetary incentives.
Our study aims to contribute to the understanding of the
effects of budgets used for performance evaluation vs. per-
formance evaluation and planning. Fisher, Mitchell, Peffer,
and Webb (2014a) investigate subordinates’ budget reports
whenthe superior has a planning task and, inadditionto the
division’s net pro?t, receives payoffs from accurate fore-
casts. They ?nd that subordinates make higher cost reports
when they become aware of the superiors’ additional pay-
offs fromincreased accuracy but do not participate in them.
In a related paper, Fisher, Mitchell, Peffer, and Webb
(2014b) analyze how a superior’s reliance on cost reports
and her planning task performance depend on her
monetary incentives for planning accuracy when the sub-
ordinate is not aware of the superior’s planning task.
However, as in their research setting, the subordinate uni-
laterally determines the cost budget they do not investigate
budget negotiations. Moreover, as the superior has no
authority in determining the performance evaluation bud-
get, planning and control are unlikely to be perceived as
con?icting tasks for the superior in their setting.
In our setting, the budgeting purposes of performance
evaluation and planning are in con?ict from the superior’s
perspective because the need to provide subordinates with
suf?cient incentives to expend effort counters the need for
accurate planning. We compare a condition with only a
performance evaluation task to two conditions with an
additional planning task. In all three conditions, superior
and subordinate negotiate over the budget that is used
for performance evaluation. In the ?rst planning condition,
the superior can set a separate second budget for the plan-
ning task after the negotiation over the performance eval-
uation budget. In the second planning condition, the
superior is restricted to use a single budget for both tasks,
i.e., the planning budget automatically equals the perfor-
mance evaluation budget. As we focus on the subordinate’s
motives beyond monetary incentives, the subordinate’s
compensation function is never affected by the superior’s
planning task.
Because performance evaluation and planning are in
con?ict, budgets should be set to different levels in our set-
ting, and so a single budget cannot be used for both tasks
optimally. The superior therefore should be better off
when she can use separate budgets unless this entails sig-
ni?cant additional costs (Baiman, 1982; Barrett & Fraser,
1977; Churchill, 1984; Hopwood, 1972). This implication,
however, contrasts with empirical evidence (Merchant &
Manzoni, 1989; Umapathy, 1987) that ?rms rarely use dif-
ferent budgets for different purposes.
We develop theory to predict that with the additional
planning task, the subordinate is likely to perceive that
the superior has a more complex task and to have an
increased interpersonal accountability to the superior,
and that this makes the subordinate more cooperative in
the budget negotiation. We further predict that increased
subordinate cooperation extends to subordinate behavior
after the negotiation, leading to an increase in subordinate
performance in the presence of a planning task, particu-
larly after negotiation impasse. Finally, we also develop
theory to predict that the superior’s restriction to a single
budget can further increase the subordinate’s cooperation
during and after the negotiation.
In our experiment, participants assume the role of
either a superior or a subordinate and negotiate a budget
target for the subordinate’s performance evaluation. After
the negotiation, subordinates perform a real-effort task.
The experiment uses three settings: a baseline setting
where performance evaluation is the sole purpose of bud-
geting, a setting with separate budgets for planning and
performance evaluation, and a setting with a single budget
for both tasks. In the separate-budgets condition, planning
is operationalized by making the superior forecast the sub-
ordinate’s actual performance. In the single-budget condi-
tion, the performance evaluation budget simultaneously
1
Govindarajan (1986) measures both decision-making and control
dimensions of budgeting but collapses them into a single measure of ?rm
performance.
2 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
serves as the superior’s forecast. In both planning condi-
tions, the superior has to bear planning costs if actual per-
formance deviates from predicted performance.
Our results support our predictions. First, the superior’s
planning task increases subordinate cooperation during
the negotiation as evidenced by increased initial sub-
ordinate budget proposals. Second, after controlling for
monetary incentives, subordinate performance increases
in the presence of the planning task, particularly following
negotiation impasse. Third, when comparing the two plan-
ning conditions, the effects on subordinate behavior tend
to be even larger when the superior is restricted to a single
budget, particularly with respect to performance.
The paper makes two primary contributions to the
literature. First, to our knowledge, this is the ?rst study
focusing on the effects of the use of budgets for con?ict-
ing purposes on budget negotiations and their outcomes.
We thus respond to the call to investigate the interdepen-
dencies stemming from the simultaneous use of manage-
rial accounting information for decision-making and
control (Sprinkle, 2003). Our results provide evidence that
a superior’s additional task can affect subordinates
beyond monetary incentives in that subordinates coop-
erate more as their impact on the superior’s outcome
increases. Thus, cooperation in budget negotiations is
in?uenced not only by decreases in social distance
(Fisher, Frederickson, & Peffer, 2006) or by the informa-
tion distribution between the parties (Chalos & Haka,
1989) but also by the set of tasks and the restrictions
the superior has to deal with.
Second, our ?ndings help to reconcile the observation
that ?rms often use a single budget for multiple purposes
when economic theory predicts that they would employ
separate budgets for planning and control when these pur-
poses are in con?ict. Our results suggest that the bene?ts
from increased subordinate cooperation may compensate
for the loss in ?exibility from using a single budget. Thus,
in practice, the use of separate budgets vs. a single budget
will depend on the relative impacts of the two effects,
which implies that ?rms can pro?t by taking into account
the effects of social preferences when designing manage-
ment control systems (Church, Hannan, & Kuang, 2012;
Hannan, Rankin, & Towry, 2010).
Setting and benchmark analysis
Setting
Our setting is a modi?cation of the budget negotiation
setting of Fisher, Frederickson, and Peffer (2000). We study
superior-subordinate dyads in which the subordinate pro-
vides real effort. Prior to this effort, the parties negotiate a
budget for the subordinate’s performance evaluation, B
Eval
.
In this baseline setting, the superior has only the perfor-
mance evaluation task (‘‘performance evaluation-only con-
dition’’), and subordinate and superior receive the
following payoffs:
P
Subord
¼
F if X 6 B
Eval
F þ A Á ðX À B
Eval
Þ if X > B
Eval
ð1Þ
P
Super
¼
D Á X À F if X 6 B
Eval
D Á X À ½F þ A Á ðX À B
Eval
Þ? if X > B
Eval
ð2Þ
where F is the subordinate’s ?xed wage. X is the ?rm’s out-
put, which equals the subordinate’s actual performance. D
is the contribution margin per unit of output, and conse-
quently D Á X is the ?rm’s contribution margin that is
assigned to the superior. A is the bonus coef?cient per unit
of output for exceeding the performance evaluation bud-
get, and A Á (X À B
Eval
) is the subordinate’s bonus if X
exceeds B
Eval
. The subordinate’s performance is a function
of his ability and effort, and the superior is not informed
about her subordinate’s ability. Thus, when negotiating
the budget with the subordinate, the superior is uncertain
about the subordinate’s performance potential.
We also investigate two conditions in which the super-
ior has an additional planning task and her planning bud-
get, B
Plan
, is determined after the negotiation of B
Eval
. B
Plan
represents a forecast of actual performance, and deviations
between predicted and actual performance lead to plan-
ning costs for the superior. With the planning task, the
superior receives the following payoff:
P
super
¼
DÁ XÀF ÀCÁ jXÀB
Plan
j if X 6B
Eval
DÁ XÀ½F þAÁ ðXÀB
Eval
Þ? ÀCÁ jXÀB
Plan
j if X >B
Eval
ð3Þ
where B
Plan
is the superior’s forecast of actual performance
X in the production period and C is the coef?cient deter-
mining the superior’s planning costs when B
Plan
does not
perfectly predict actual output. B
Plan
re?ects a planned out-
put, such as the prediction of a sales volume for the instal-
lation of a production capacity, for procurement or supply-
chain coordination (Atkinson et al., 2011, chap. 10). We
study both a condition in which the superior can set a
separate second budget for the planning task after the
negotiation over the performance evaluation budget
(‘‘separate-budgets condition’’) and a condition in which
the superior must use a single budget for both tasks so that
the planning budget automatically equals the performance
evaluation budget (B
Eval
= B
Plan
: ‘‘single-budget condition’’).
Planning costs emerge from both under- and over-ful?ll-
ment of B
Plan
. If the planning budget is under-ful?lled, then
planning costs can be interpreted as, for example, costs of
unused capacity. If budgets are over-ful?lled, then extra
costs from operating at a level higher than expected arise,
such as, for example, running machines at higher speeds or
acquiring additional production capacity externally.
2
According to the superior’s payoff function (3), planning
costs are symmetric in that any deviation of actual output
X from planned output B
Plan
, whether positive or negative,
is equally costly to the superior.
3
We assume that
2
Consistent with prior literature (e.g., Balachandran, Balakrishnan, &
Sivaramakrishnan, 1997; Dunbar, 1971; Fisher, Frederickson, & Peffer,
2002), we refer to the consequences of inaccurate plans as costs because
they represent ex post inef?ciencies. Underlying is the perspective that the
superior could plan output optimally if the subordinate did not withhold or
misrepresent private information.
3
While the costs of unfavorable variances may be larger than the costs of
favorable variances (Horngren et al., 2011; Merchant & Manzoni, 1989), we
implemented symmetric planning costs in the experiment because the
symmetry in planning costs considerably facilitates the planning task for
the superior and increases experimental control.
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 3
D À A À C > 0, i.e., the superior always bene?ts from an
increase in output, even if this increase also increases the
subordinate’s bonus and the planning error. This implies
that the superior has monotonic incentives to both opti-
mally motivate the subordinate and correctly plan
performance.
The subordinate does not bear any planning costs, i.e.,
his payoff function (1) is unaffected by the planning task.
4
However, he is always informed about the superior’s plan-
ning task and planning costs. That is, at the time of the bud-
get negotiation, he is likely to be fully aware of the con?ict
between the two tasks, of his impact on the superior’s payoff
from the performance evaluation task and, in the planning
conditions, of the superior’s costs from planning errors.
This setting allows us to investigate how a superior’s
planning task affects the negotiation of a performance
evaluation budget and the outcome of the superiors’ tasks
as well as whether these effects are ampli?ed when the
superior is restricted to use a single budget for planning
and performance evaluation. Fig. 1 summarizes our
research model and relates our main variables of interest
to the whole budgeting process and its outcomes.
The upper part of Fig. 1 illustrates our hypothesis and
research question on negotiation behavior (H1 and RQ1).
We propose that the superiors’ planningtask motivates sub-
ordinates to act more cooperatively during the negotiation
and therefore increases initial budget proposals (H1 (i), left
node) and that the superiors’ restriction to a single budget
for both tasks ampli?es this effect (H1 (ii), right node).
Further, we analyze superiors’ initial counteroffers to
investigate whether and howthe planning task and the sin-
gle budget restriction in?uence a superior’s behavior in
negotiations about performance evaluationbudgets beyond
her reactions to subordinate initial proposals. As explained
in detail below, the potential direct and indirect effects of
the planning task on superiors’ initial counteroffers work
in different directions. We thus state a research question
(RQ1) to explore which of the effects dominates the super-
ior’s negotiation behavior in our experimental setting.
The lower part of Fig. 1 relates to the outcomes of the
superiors’ tasks as our second hypothesis and research
question (H2 and RQ2). Speci?cally, we propose that the
superior’s planning task makes the subordinate act more
cooperatively also after the negotiation and that this
increases performance, particularly after negotiation
impasse (H2 (i), left node). We further propose that this
effect is ampli?ed when the superior is restricted to a single
budget (H2 (ii), right node). H2 focuses on the behavioral
forces beyond monetary incentives, thereby controlling
for potentially different performance evaluation budgets.
Finally, we investigate how the single-budget restriction
in?uences the accuracy of the superior’s planning task. As
will be explained below, because the use of separate
budgets can have positive direct but also negative indirect
effects on planning accuracy we state a research question
(RQ2, right node) to investigate which forces affect plan-
ning accuracy more strongly in our experimental setting.
Benchmark analysis
Before deriving our hypotheses, we analyze the stan-
dard economic implications for the setting we have just
described. The superior’s dif?culty in setting budgets arises
from the information asymmetry about the subordinate’s
performance capability. The superior cannot expect the
subordinate to communicate his ability truthfully because
his payoff function induces slack-creating incentives
(Hansen, Otley, & Van der Stede, 2003; Murphy, 2000).
The subordinate will thus try to decrease his performance
evaluation budget and will strategically communicate dur-
ing the negotiation.
The superior’s symmetric planning costs imply that the
optimal planning budget should be set equal to the median
of the potential performance capabilities from the super-
ior’s perspective after the negotiation (Weitzman, 1976).
That means, if the superior cannot infer any additional
information about the subordinate’s capability from the
negotiation, then she will set the planning budget equal
to the median capability according to her ex ante expecta-
tions. In contrast, if, from the information obtained in the
negotiation, the superior can narrow the range of the sub-
ordinate’s potential capabilities compared to her ex ante
expectations, then she should optimally set the planning
budget to the median corresponding to this smaller range
of the potential capabilities.
In any case, however, when determining the perfor-
mance evaluation budget the superior has to take care that
the subordinate has suf?cient ?nancial incentives to
expend effort without increasing the compensation costs
(too much). If the performance evaluation budget were
equal to the median of potential capabilities (i.e., the opti-
mal planning budget), then subordinates could not reach
the performance evaluation budget in 50% of the cases.
This could be optimal only if the superior had extreme
expectations about potential capabilities.
5
Otherwise, the
expected loss in subordinate performance due to insuf?cient
?nancial incentives would be too large. Thus, the superior
should set the performance evaluation budget below the
optimal planning budget (the median capability) to increase
the likelihood of higher effort.
6
That means, in our setting,
performance evaluation and planning are in con?ict because
of the need to provide subordinates with suf?cient mone-
tary incentives to counter the need for accurate planning.
4
The subordinate does not participate in the superiors’ planning costs
for two reasons: First, this design choice increases experimental control
because it excludes any direct effect of the experimental conditions on the
subordinate’s payoff and because the pay scheme (1) is easily understand-
able. Second, while there are theoretical analyses of truth-inducing pay
schemes in which subordinates directly participate in planning accuracy
(e.g., Groves & Loeb, 1975; Weitzman, 1976), these pay schemes are rarely
found in practice (Waller & Bishop, 1990).
5
For example, if the superior expects the task to require some special
skills that half of the subordinates do not possess (i.e., their capabilities are
very low), the cost of reduced effort would be small for this group of
subordinates. In this case, the performance evaluation budget would be
optimally set close to or even above the median. Assuming that such a large
fraction of subordinates is not quali?ed for their task, though, does not
describe corporate practice.
6
This likely corresponds to actual company settings where incentive
systems usually provide suf?cient monetary incentives for the majority, not
the minority, of the employees (Baker, Jensen, & Murphy, 1988; Merchant,
2010).
4 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
Because the two budgets should be set to different
levels, the superior is better off when she can use separate
budgets unless this entails signi?cant additional costs. The
reason is that a single budget cannot solve both tasks opti-
mally at the same time (Barrett & Fraser, 1977; Hopwood,
1972; Otley, 1982). Consequently, depending on the rela-
tive importance of the two tasks and on where the single
budget is set, economic theory predicts that the restriction
to a single budget leads to increased costs from suboptimal
planning, suboptimal performance evaluation, or both.
These implications re?ect the established result that plan-
ning and performance evaluation are often con?icting bud-
geting purposes (Barrett & Fraser, 1977; Merchant &
Manzoni, 1989; Merchant & Otley, 2007) and economics
theories that predict ?rms should use separate budgets
for these different purposes (Baiman, 1982; Churchill,
1984).
Hypotheses and research questions
In contrast to the standard economic benchmark analy-
sis, we propose that behavioral factors, beyond monetary
incentives, also in?uence budget negotiations and subse-
quent performance. We suggest that with the planning
task, the subordinate is likely to perceive that the superior
has a more complex task owing to higher cognitive
demands (Bonner, 1994; Campbell, 1988) and to have an
increased interpersonal accountability to the superior
owing to his increased impact on the superior’s com-
pensation due to the potential planning costs (Kramer,
Pommerenke, & Newton, 1993). This is expected to result
in increased subordinate cooperation during and after the
negotiation. The theory we draw on to develop our
hypotheses relies on the existence of only some con?ict
between planning and performance evaluation from an
economic perspective and on the subordinate perceiving
this con?ict.
Negotiation process
Prior research suggests that a subordinate’s behavior
during and after budget negotiations is affected by fairness
concerns (Chalos & Haka, 1989; Fisher, Maines, et al.,
2002). As fairness concerns are likely to change with the
negotiation setting and with the parties’ interdependence
(De Dreu, Weingart, & Kwon, 2000; Giebels, De Dreu, &
Van de Vliert, 2000), they may be in?uenced by the super-
ior’s planning task in addition to her performance evalua-
tion task. As pointed out above, a con?icting planning
task likely increases task complexity, thereby making the
superior’s situation more dif?cult, and also increases the
subordinate’s impact on the superior’s outcome. These
effects are likely to be important in the negotiation as a
factor determining negotiation behavior is an individual’s
accountability to the counterpart (Lerner & Tetlock, 1999;
Tetlock, 1985). Accountability consists of two parts:
responsibility for another person’s outcome and an audi-
ence (a third person or the individual himself) evaluating
the individual’s decision (Lerner & Tetlock, 1999;
Schlenker, Britt, Pennington, Murphy, & Doherty, 1994).
Fig. 1. Research model and predictions.
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 5
With the planning task in addition to the performance
evaluation task during budgeting, the subordinate’s behav-
ior both during and after the budget negotiation increases
his impact on the superior’s outcome and therefore is likely
to increase his interpersonal accountability to the superior
(Kramer et al., 1993; Schlenker et al., 1994). The reason is
that planning quality and thus the superior’s planning
costs depend on her knowledge of the subordinate’s ability
and on his performance after the negotiation. If an individ-
ual’s interpersonal accountability to his or her counterpart
increases, then the individual is likely to cooperate more
and compete less (Kramer et al., 1993; Pruitt, 1981).
Therefore, the subordinate should cooperate more in the
negotiation when the superior has a planning and a perfor-
mance evaluation task.
7
Prior research on negotiations in
general (Benton, Kelley, & Liebling, 1972; Carnevale &
Pruitt, 1992) and on budget negotiations in particular
(Fisher et al., 2000; Fisher et al., 2006) suggests that less
extreme initial offers can be regarded as a signal for
cooperation. Thus, we expect the subordinates’ initial bud-
get proposals to increase when there is a planning task in
addition to the performance evaluation task.
8
Compared to the case in which the planning task can be
solved through a separate planning budget, the superior’s
restriction to a single budget for both tasks is likely to
increase the subordinate’s interpersonal accountability
even more. In this case, the superior’s ability to solve the
planning task is further restricted because the planning
budget equals the performance evaluation budget. That
means, any decrease in the performance evaluation budget
would not only increase compensation costs but also
automatically increase the superior’s planning costs
because she has no possibility to adjust the planning bud-
get. Consequently, the superior’s restriction to a single
budget for both tasks is likely to reinforce the effect on
the subordinates’ initial proposals. We predict:
H1. The subordinate’s initial budget proposal in the
negotiation (i) increases when the superior has a planning
and performance evaluation task and (ii) further increases
when the superior is restricted to the use of a single budget
for planning and performance evaluation instead of sepa-
rate budgets for each task.
In contrast to the effects of planning on subordinate
behavior, the effects on the superior are more ambiguous.
Numerous studies have illustrated the importance of
reciprocity for negotiations (e.g., Brett, Shapiro, & Lytle,
1998; Kuang & Moser, 2009; Putnam & Jones, 1982).
According to the norm of reciprocity, a party will cooperate
during a negotiation if the other party also does so (Benton
et al., 1972; Carnevale & Pruitt, 1992). Thus, if the super-
ior’s planning task makes the subordinate increase his ini-
tial bids, then the superior may reciprocate and this may
lead to lower initial superior counteroffers. Similarly, if
superiors react reciprocally and the single-budget restric-
tion further increases initial subordinate bids, then it
should further decrease initial superior counteroffers.
However, prior research also suggests that the superi-
ors’ counteroffers may be affected by anchoring in two dif-
ferent ways. First, an initial offer can serve as an anchor
that biases the counteroffer (Bazerman & Neale, 1993;
Kristensen & Gärling, 1997). This implies that higher initial
subordinate bids under planning and the additional
increase in the bids under the single-budget case may
increase superiors’ initial counteroffers. Second, anchors
for negotiation offers also may arise from other salient
information (Kahneman, 1992). With an additional plan-
ning task, a likely anchor for the superior’s initial coun-
teroffer is the optimal planning budget given the
superior’s information before the negotiation. As analyzed
above, the optimal planning budget in this case is the med-
ian of the potential performance capabilities according to
the superior’s ex ante expectations. Whether anchoring
on the ex ante optimal planning budget leads to higher
or lower counteroffers is unclear as the direction of the
effect depends on the level of initial counteroffers without
the additional planning task. Yet, in this case, superiors’
initial counteroffers should be concentrated more nar-
rowly around the expected performance capabilities than
without the planning task. As the potential behavioral
forces affecting the counteroffers point into different direc-
tions, we do not derive a directional hypothesis but use the
following research question to investigate which effect
dominates superior behavior in budget negotiations in
our experimental setting:
RQ1: How do the planning task and the restriction to
use a single budget for planning and performance eval-
uation affect a superior’s initial counteroffer in the
negotiation?
Subordinate performance
A characteristic of a budget negotiation is that subse-
quent to the negotiation, the subordinate provides non-
contractible effort. Thus, an important question is the
effect of the negotiation’s procedures and results on sub-
ordinate performance. In the following, we develop theory
to predict that the planning task leads to more subordinate
cooperation not only during, but also after the negotiation.
This increases the subordinate’s performance, particularly
after a negotiation impasse. Further, we expect the
restriction to a single budget for both tasks to reinforce
the effect.
7
Increased cooperation in the negotiation may also incorporate aspects
of strategic behavior. First, the subordinate may make higher offers because
he expects the superior to positively reciprocate his increased cooperation.
Second, the subordinate may anticipate the in?uence of the planning task
on the superior’s negotiation behavior. For example, if the subordinate
expects the planning task to reduce the superior’s concessions, he may
increase his offers to avoid negotiation impasses. However, whether
cooperation increases due to genuine or strategic concerns is largely
semantic and does not affect our predictions as long as the increase is
systematic. We provide evidence on this question in the supplemental
analysis.
8
Research on bargaining theory shows that a negotiation party can have
incentives to communicate private information when bargaining impasse is
costly to her and when she expects her counterpart to make no or few
concessions (e.g., Ausubel, Cramton, & Deneckere, 2002, chap. 50;
Rubinstein, 1985). In our setting, this might make subordinates’ offers
more informative about their capability when the superior has a planning
task. We provide evidence on this issue in the results section.
6 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
Prior research has shown that reciprocal reactions to
the superior’s negotiation behavior strongly affect sub-
ordinates’ effort choices after the negotiation (Fisher,
Frederickson, et al., 2002; Kuang & Moser, 2011). These
reciprocal reactions are more positive when superiors are
constrained as their intentions are perceived more posi-
tively in this case (Hannan, 2005). Because the presence
of a con?icting planning task is likely to make the sub-
ordinate perceive the superior’s task as more complex
and constrained, the planning task can positively affect
subordinate performance.
We propose, however, that this effect is likely to be
more pronounced after negotiation impasse than after
agreement for the following reason: If the two parties
reach an agreement, then the subordinate’s aspiration level
is likely to be ful?lled (Kuang & Moser, 2011).
Consequently, although the superior’s negotiation behav-
ior is likely perceived as more positive when a planning
task is present in addition to the performance evaluation
task, the effect of the planning task on subordinate perfor-
mance is likely to be smaller under agreement. Prior
research supports this conjecture, showing that, when
agreement is reached, the effects of all treatment condi-
tions that have been manipulated so far in similar budget
negotiation settings on subordinate performances have
been small (Arnold, 2015; Fisher et al., 2000; Fisher,
Frederickson, et al., 2002; Fisher et al., 2006).
In contrast, the effect of the planning task on perfor-
mance is likely to be larger after a budget imposition when
negotiations fail because the planning task is likely to
affect the subordinate’s attribution of the superior’s budget
imposition and therefore his interpretation of the budget
imposition. Attribution theory suggests that an important
determinant of an individual’s reaction to another person’s
actions is the attribution of that person’s behavior to exter-
nal (situational) or internal (dispositional) factors (Dobbins
& Russell, 1986; Gilbert & Malone, 1995; Ross, 1977). In the
unconstrained performance evaluation-only setting, the
subordinate is likely to attribute the superior’s budget
imposition to dispositional factors. Thus, he is likely to per-
ceive a budget imposition as unfair and to respond with
low effort (Fisher et al., 2000; Kuang & Moser, 2011). In
contrast, when the superior also has a planning task, the
subordinate cannot make unambiguous inferences about
the superior’s dispositions and may explain her behavior
by the situation rather than by disposition (Arnold, 2015;
Gilbert & Malone, 1995). That is, a budget imposition
may be more acceptable to the subordinate in a setting
with multiple budgeting tasks than in the performance
evaluation-only task, because the subordinate may attri-
bute the lack of suf?cient superior concessions during
the negotiation to the superior’s increased task complexity
and her focus on the additional planning task. Thus, the
increasing effect of the planning task on subordinate
performance is likely to be larger following a budget
imposition.
Again, the effect of the planning task on performance is
likely to be ampli?ed with a single budget for both tasks
because the superior is even more constrained and, thus,
her intentions are likely to be perceived even more posi-
tively. Speci?cally, a budget imposition may be more
justi?able with a single budget as the superior may have
no alternative to imposing a budget to keep the planning
error small. Thus, we predict the following effect of the
planning task and the restriction to a single budget on sub-
ordinate performance:
H2. Controlling for the subordinate’s monetary incentives,
(i) the positive effect of the superior’s planning task
beyond a performance evaluation task only on subordinate
performance is larger after negotiation impasse than after
agreement and (ii) the superior’s restriction to a single
budget ampli?es this effect.
Planning error
As we analyzed above, according to economic theory,
the possibility of setting separate budgets should reduce
the planning error, because the superior does not have to
trade off the con?icting tasks in a single budget.
However, from the analysis leading to H1 and H2, the
effects of the superior’s restriction to a single budget on
her planning error are unclear under both agreement and
impasse. First, after agreement, the superior cannot further
adjust the budget for the planning task in the single-
budget case. In contrast, in the separate-budgets case, the
superior can adjust her planning budget, but a reduction
in planning error may be dif?cult if the subordinate coop-
erates less during the negotiation and hides his perfor-
mance capability. Further, as predicted in H1 (ii), the
subordinates’ initial budget proposals are likely to be
higher when the superior is restricted to a single budget,
and this could lead to higher performance evaluation bud-
gets under agreement. Consequently, the performance
evaluation budget under agreement may be closer to the
subordinate’s actual performance when the superior is
restricted to a single budget.
Second, after negotiation impasse, the superior has to
trade off planning and performance evaluation in the sin-
gle-budget case and therefore may not be able to solve
the planning task as well as in the separate-budgets case.
In contrast, the arguments for H2 (ii) imply that sub-
ordinate performance is likely to be lower in the sepa-
rate-budgets case. Consequently, if the superior does not
perfectly anticipate the lower performance,
9
then the plan-
ning error after negotiation impasse may be larger with
separate budgets than with a single budget. As the potential
effects of the behavioral forces on planning accuracy point in
different directions, we state a research question to investi-
gate which of them dominates in our experimental setting:
RQ2: Will the realized planning error be different under
agreement and disagreement when the superior is
restricted to the use of a single budget than when she
can set separate budgets?
9
Prior experimental evidence shows that decision makers’ abilities to
anticipate the behavior of others in sequential bargaining games are very
limited. See, e.g., Binmore, McCarthy, Ponti, Samuelson, and Shaked (2002),
Johnson, Camerer, Sen, and Rymon (2002).
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 7
Experimental method
Participants and design
180 graduate and undergraduate students from a large
European university (90 dyads) participated in the experi-
ment—60 participants (30 dyads) in each of the three treat-
ments. Of the participants, 42% were female, and 59% were
business or economics majors. Three sessions were con-
ducted for each of the treatments. No participant partici-
pated in more than one session. Sessions lasted between
60 and 75 min. Participants received a show-up fee of €
5.00 and additional variable compensation. Average total
compensation was €14.34 and varied between €5.00 and
€23.80.
We used a nested experimental design in which we var-
ied (between subjects) the existence of a planning task—
present or absent. Nested within the planning task present,
we manipulated the restriction to use a single budget for
both purposes—present or absent. In all treatments, bud-
gets had a performance evaluation function. This nested
design resulted in three treatments: performance evalua-
tion only, separate budgets for planning and performance
evaluation (‘‘separate-budgets treatment’’), and a single
budget for planning and performance evaluation (‘‘single-
budget treatment’’). The experimental design is summar-
ized in Fig. 2.
The planning task was operationalized by having the
superior predict output, i.e., actual performance, or not.
With the planning task, the superior had to bear planning
costs if actual output deviated from predicted output, as
included in Eq. (3).
The restriction to a single budget was manipulated by
giving the superior the ability to set separate budgets for
performance evaluation and planning or not. That is, in
the single-budget treatment, the planning budget
automatically equaled the performance evaluation budget.
The subordinate’s payoff function was never affected by
any manipulation. However, as the subordinate’s ?nancial
incentives after the negotiation depended on the perfor-
mance evaluation budget, we control for these incentives
in our statistical analysis.
Procedure
The experiment, including the negotiation process, was
computerized. Instructions appeared on the computer
screens and were simultaneously read aloud.
10
The experi-
mental procedure consisted of the following steps:
(1) Upon arrival, participants were randomly assigned
to computer terminals and were separated from
each other by panels. Superior and subordinate were
paired anonymously and randomly.
(2) Then, the real effort task was explained. It consisted
of decoding two-digit numbers into letters in blocks
of two each. Participants then completed a two-min-
ute practice round.
(3) Next, all subjects completed three four-minute
training rounds. New decoding keys were dis-
tributed for each round. Subordinates were told that,
at the end of the experiment, one of the training
rounds would be randomly selected and they would
receive two points for every block correctly decoded
in this round. Subordinates were informed of the
number of correctly and incorrectly decoded blocks
after each training round. They improved their per-
formance on average by 2.18 blocks from training
rounds 1–2 and by 1.56 blocks from training rounds
2–3. Superiors were not compensated and received
no feedback in the training rounds.
(4) After the third training round, subordinates were
asked to submit a best estimate of the number of
blocks they could correctly decode in another four-
minute work round. On average, the estimate was
larger by 0.34 blocks than the maximum perfor-
mance in the training rounds.
(5) Next, participants were informed about their roles,
the setting, the payoff functions, and the budgeting
process in the experimental work round. The
parameter values were set to F = 250, A = 4, D = 10,
and C = 4.
11
They guarantee that the superior has
monotonic incentives to motivate high subordinate
performance and predict this performance accurately.
Speci?cally, for a given subordinate performance, any
unit of deviation of the planning budget decreases
the superior’s payoff by C = 4. Further, for a given
planning budget, the superior’s payoff increases mono-
tonically if subordinate performance increases. That
is, even if an increase in subordinate performance
increased both compensation and planning error,
the net effect on the superior’s payoff would still be
positive (D À A À C = 2). A and C were chosen to avoid
a parameter-speci?c focus on one of the tasks: With
A = C, an increase in compensation is as costly as an
increase in planning error. The values of A and C are
large enough to make the superior pay suf?cient
attention to both tasks. Participants were informed
that the negotiation would last at maximum three
rounds and that the superior would set the ?nal per-
formance evaluation budget if no agreement was
reached at the end of the third round. In the single-
budget treatment, participants were further told that
the planning budget would be equal to the perfor-
mance evaluation budget. In every negotiation round,
the subordinate made the initial budget proposal, and
the superior made the counteroffer, which resembles
the structure of budgeting negotiations in practice
(Anthony & Govindarajan, 2007; Umapathy, 1987).
10
The instructions used context-free language. Moreover, to avoid that in
the planning conditions, one task was framed as pro?t increasing, and the
other was framed as pro?t decreasing, we treated the effects of the two
tasks always identically. The instructions are available from the authors
upon request.
11
The minimum compensation for the subordinates was F. To avoid
having to pay the ?xed wage from their own resources, superiors were
endowed with 500 points. The remaining 250 points after the deduction of
F ensured that the superior would have suf?cient planning incentives even
in the case of low subordinate performance.
8 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
(6) Subjects then answered a quiz about the procedures
and the payoffs. The quiz included questions about
the planning budget (planning treatments only)
and the performance-evaluation budget (all treat-
ments). These questions were correctly answered
in 92% of all cases in the ?rst trial. Participants fail-
ing to give a correct answer had to read the relevant
instructions again. The experiment did not start
until all participants had answered all questions
correctly.
(7) Before the negotiations started, all participants were
informed of the results of a pretest. They were told
that 80% of the participants in the pretest had cor-
rectly decoded between 29 and 49 blocks of num-
bers in four minutes. No superior was informed
about the performance of her speci?c subordinate
in the training rounds.
(8) Next, negotiations were conducted between superi-
ors and subordinates via the computer network.
This procedure ensured anonymity and removed
any interpersonal aspects of the negotiation. At any
point in the negotiation, either party could accept
the other party’s proposal, which would end the
negotiation. If the superior and subordinate could
not reach an agreement within three rounds, then
the superior imposed the budget.
(9) After the performance evaluation budget was set,
each superior set the planning budget in the
separate-budgets treatment. Subordinates were not
informed about the planning budget so as not to
affect their performance by presenting an additional
goal. This is consistent with the informal adjustment
of performance evaluation targets for planning pur-
poses observed in practice (Berry and Otley, 1975;
Walker & Johnson, 1999). In the single-budget treat-
ment, the planning budget automatically equaled
the performance evaluation budget.
(10) Finally, subordinates completed the experimental
work round for four minutes. Subjects interacted
for only one work round to avoid ambiguities
that may arise in repeated relationships, such as
reputation building and aspects of the social
interaction. At the end of the work round, subjects
were informed of the subordinate performance
and the payoffs. They completed a post-
experimental questionnaire and received their cash
payments.
Measures
To analyze the negotiation process, we use FIRSTBID,
the subordinate’s initial budget proposal in the negotia-
tion, and FIRSTCOUNTER, the superior’s initial counterof-
fer. B
Eval
and B
Plan
represent the performance evaluation
and the planning budget, respectively. X is the sub-
ordinate’s actual performance, and
^
X is the subordinate’s
estimate of his performance capability at the end of the
training rounds. BONUS
Pot
¼ maxf0; 4 Á ð
^
X ÀB
Eval
Þg repre-
sents the potential subordinate bonus based on his esti-
mated performance capability and is used to control for
the subordinate’s monetary incentives. ERROR = |X À B
Plan
|
is used to measure the planning error. Table 1 summarizes
the measures used in the experiment.
Results
Table 2 displays descriptive statistics and shows that
the subordinates’ estimated performance capabilities
^
X
are very similar across treatments. One-way ANOVAs show
no signi?cant difference across treatments with respect to
the subordinates’ estimates and the subordinates’ maxi-
mum performances in the training round (p > .10 in both
cases). Thus, differences across treatments are not driven
by differences in performance capabilities.
As our goal is to establish a con?ict between planning
and performance evaluation, we examine the means of
B
Plan
and B
Eval
. Table 2 shows that, in the separate-budgets
treatment, mean B
Eval
is lower than mean B
Plan
and this dif-
ference is signi?cant (t-test, x ¼ 35:07 vs. 38.70, t = 3.11,
p < .01).
12
63% of the superiors adjusted the planning budget
upwards, consistent with the economic benchmark analysis.
In contrast, 27% of the superiors adjusted the planning bud-
get downwards. The majority of these cases (eight of nine)
occurred after a negotiation impasse, and the mean perfor-
mance evaluation and planning budget in these cases are
36.85 and 32.57, respectively. This result suggests that,
when setting their planning budget, some superiors allowed
for the possibility that subordinates might reduce perfor-
mance after a budget imposition. However, our ?nding that
90% of the superiors adjusted the planning budget con?rms
Performance evaluation only Planning and performance evaluation
Separate budgets Single budget
(Treatment 1) (Treatment 2) (Treatment 3)
Superior has only a
performance evaluation task
Superior has a planning and a performance evaluation task
Superior can use separate
budgets for each task
Superior is restricted to use a
single budget for both tasks
B
Eval
B
Eval
and B
Plan
B
Eval
= B
Plan
Fig. 2. Overview of the experimental setting and design.
12
We use one-tailed p-values for tests of directional predictions and two-
tailed p-values for research questions without directional prediction. The
unit of observation is the observation made for every dyad.
M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16 9
the success of our manipulation of con?ict between the
performance evaluation and the planning tasks.
Finally, Table 2 reveals that the mean of single budget is
close to and not signi?cantly different from the mean of
B
Eval
in the separate-budgets case (t-test, x ¼ 35:76 vs.
35.07, t = .50, p > .10) but signi?cantly lower than B
Plan
(t-test, x = 35.76 vs. 38.70, t = 2.22, p = .03). Thus, in our
experiment, performance evaluation seems to be more
important for superiors than planning.
13
Hypotheses tests
H1 predicts (i) an increase in initial budget proposals in
the planning treatments and (ii) an additional increase in
the single-budget relative to the separate-budgets treat-
ment. Consistent with these two predictions, Table 2 and
Fig. 3 show that the mean of FIRSTBID is higher in the
two planning treatments than in the performance evalua-
tion-only treatment (H1 (i)) and is higher in the single-
budget treatment than in the separate-budgets treatment
(H1 (ii)).
We conduct a nested ANOVA with FIRSTBID as the
dependent variable to test H1. The ?rst factor in the
ANOVA is whether the planning task is present or absent.
This factor tests part (i) of H1. Nested within this factor,
the second factor captures whether the superior is
restricted to a single budget. This factor tests part (ii) of
H1. As reported in Panel A of Table 3, planning signi?cantly
affects FIRSTBID (p = .03). However, even though the effect
of the single budget on FIRSTBID is in the predicted direc-
tion the nested term is not signi?cant at conventional
levels (p = .13). These results support part (i) of H1 but
not part (ii).
14
RQ1 asks how the planning task and the restriction to a
single budget affect superiors’ initial counteroffers. Table 2
and Fig. 3 show that mean FIRSTCOUNTER is very similar
across treatments and slightly above the mid-point (39)
of the pre-test span (29–49) participants were informed
about. We ran a nested ANOVA and a nested ANCOVA with
FIRSTCOUNTER as the dependent variable and the same
factors as used for testing H1 to test whether the
planning task affects FIRSTCOUNTER. In the ANOVA,
we test the overall treatment effects. In the ANCOVA,
we include FIRSTBID as a covariate to examine whether
superior counteroffers are driven by reciprocity or
anchoring on subordinates’ initial bids. If superiors react
reciprocally, then increases in FIRSTBID should decrease
FIRSTCOUNTER. In contrast, if superiors anchor
on FIRSTBID, increases in FIRSTBID should increase
FIRSTCOUNTER. Panel B of Table 3 reports the results.
They do not reveal any signi?cant effect in the ANOVA or
the ANCOVA (p > .25 in all cases). That is, we ?nd neither
a signi?cant effect of the treatment conditions nor of
FIRSTBID. This lack in signi?cant results contradicts the
view that superiors reciprocate or anchor on the sub-
ordinates’ initial bids.
Table 1
Summary of the measures used in the analysis.
Measure Description
FIRSTBID Subordinate’s initial budget
proposal
FIRSTCOUNTER Superior’s initial counteroffer
B
Eval
Performance evaluation
budget
B
Plan
Planning budget
X Subordinate performance
^
X
Subordinate’s estimate of his
performance capability
BONUS
Pot
¼ maxf0; 4 Á ð
^
X À B
Eval
Þg
Potential subordinate bonus
based on the subordinate’s
estimated performance
capability
ERROR = |X À B
Plan
| Realized planning error
Table 2
Descriptive statistics: mean and standard deviation.
Performance
evaluation only
Planning and performance
evaluation
Separate
budgets
Single
budget
(Treatment 1) (Treatment 2) (Treatment 3)
^
X
a
41.67 38.53 40.23
5.08 6.30 6.56
B
Eval
b
32.33 35.07 35.76
11.67 5.95 4.77
B
Plan
c
n.a. 38.70 35.76
n.a. 5.45 4.77
BONUS
Pot
d
43.47 20.27 23.60
41.39 19.95 25.13
FIRSTBID
e
19.93 22.20 24.53
8.79 8.15 6.28
FIRSTCOUNTER
f
40.53 40.27 41.67
17.20 6.99 8.66
X
g
30.80 28.87 37.00
19.21 19.19 14.71
ERROR
h
n.a. 14.10 12.83
n.a. 15.88 11.68
No. of obs. 30 30 30
Note: Mean and standard deviation for the performance evaluation-only
condition (Treatment 1), planning and performance evaluation with
separate budgets (Treatment 2), and planning and performance evalua-
tion with a single budget (Treatment 3).
a
^
X is the subordinate’s estimate of his performance capability.
b
B
Eval
is the budget used for performance evaluation.
c
B
Plan
is the budget used for planning. In the single-budget case, B
Plan
automatically equals B
Eval
.
d
BONUS
Pot
is the potential bonus the subordinate can earn based on his
estimated performance capability
^
X.
e
FIRSTBID is the subordinate’s initial budget proposal in the
negotiation.
f
FIRSTCOUNTER is the superior’s initial counteroffer in the negotiation.
g
X is the subordinate performance in the experimental work round.
h
ERROR = |X À B
Plan
| is the realized planning error.
13
This ?nding is consistent with the perception of performance evalua-
tion as central in practice (e.g., Otley, 1978). However, it may not be
generalizable because the optimal single budget also depends on the
parameters A and C.
14
We ?nd no evidence consistent with subordinates’ budget proposals
revealing more about their capabilities in the planning conditions. Speci?-
cally, regression analyses show that the increase in budget proposals in the
planning conditions varies neither linearly nor non-linearly (inversely
U-shaped) with subordinates’ capabilities. These results contradict that the
increased budget proposals are driven by an increased informativeness of
subordinate offers.
10 M.C. Arnold, R.M. Gillenkirch/ Accounting, Organizations and Society 43 (2015) 1–16
Finally, a test of superiors’ anchoring on the ex ante
optimal planning budget requires a measure of the concen-
tration of FIRSTCOUNTER around the mid-point of the pret-
est span. Therefore, we used the standard deviation of
FIRSTCOUNTER. Table 2 reveals that the standard deviation
is much smaller in both planning treatments than in the
performance evaluation-only treatment, and this decrease
in standard deviation of FIRSTCOUNTER in the two plan-
ning treatments relative to the performance evaluation-
only treatment is signi?cant (Bartlett’s test, v
2
> 12,
p < .01 in both cases). This result is consistent with the
notion that in our experimental setting, superiors form
an a priori expectation about subordinates’ performance
capabilities and anchor on this expectation when a plan-
ning task is present.
H2 predicts that, after controlling for monetary incen-
tives, the superior’s planning task in addition to her perfor-
mance evaluation task leads to a larger increase in
subordinate performance after negotiation impasse than
after agreement (H2 (i)). It further predicts that this effect
is ampli?ed when the superior is restricted to a single bud-
get (H2 (ii)). Panel A of Table 4 displays the descriptive
statistics for subordinates’ performances contingent on
negotiation agreement. Fig. 4 graphs the performances
relative to the estimated performance capabilities in all
experimental treatments. Table 4 and Fig. 4 show that
mean performance increases in the separate-budgets treat-
ment relative to the performance evaluation-only treat-
ment and further increases in the single-budget
treatment. As predicted, both effects are mainly driven
by the changes after budget impositions, whereas after
agreement the performances are similar across treatments.
To test H2, we run regressions with the subordinate
performance, X, as the dependent variable.
15
We use
regressions instead of ANOVAs because a Levene test indi-
cates a violation of variance homogeneity (t = 22.69,
p < .01). In the regressions, we control for heteroscedasticity
by using the Huber–White estimator. To test the interaction
predicted by part (i) of H2, we regress X on BONUS
Pot
, an
indicator variable for planning (DPlan equal to 1 in the plan-
ning treatments and 0 in the performance evaluation-only
treatment), an indicator variable for agreement (agree equal
to 1 when agreement was reached and 0 when the negotia-
tion failed), and the interaction of the two indicator vari-
ables. To test part (ii) of H2, we include only the two
planning treatments and regress X on BONUS
Pot
, Agree, an
indicator variable for the singe-budget treatment (DSingle
equal to 1 in the single-budget treatment and 0 in the sepa-
rate-budgets treatment), and the interaction of the two indi-
cator variables. In the regressions, BONUS
Pot
controls for the
effect of monetary incentives. Panel B of Table 4 reports the
results.
Consistent with H2 (i) and (ii), the interaction coef?-
cients are signi?cantly negative in both regressions. In
Model (1), the interaction coef?cient indicates that the
increase in subordinate performance (X) under planning
is smaller after agreement (b = À7.95, p = .09). This
supports H2 (i). In Model (2), the interaction coef?cient
indicates that the increase in X in the single budget-
treatment is smaller after agreement (b = À11.10, p = .04).
This result supports H2 (ii).
Moreover, the regressions allow analyzing separately
whether the changes in X are signi?cant both under
Fig. 3. Subordinates’ initial budget proposals and superiors’ initial
counteroffers. Note: The ?gure displays subordinates’ average initial
budget proposal in the negotiation (FIRSTBID) and superiors’ average
initial counteroffer (FIRSTCOUNTER) for the three experimental condi-
tions (performance evaluation-only, planning with separate budgets, and
planning with a single budget).
Table 3
Analysis of subordinates’ initial budget proposals and superiors’ initial
counteroffers.
df MS F p
Panel A: FIRSTBID
a
Planning 1 235.76 3.86 0.03
Single budget|planning 1 81.67 1.34 0.13
Error 87 61.04
Panel B: FIRSTCOUNTER
b
ANOVA
Planning 1 3.75 0.03 0.87
Single budget|planning 1 29.40 0.21 0.65
Error 87 139.90
ANCOVA
FIRSTBID 1 174.21 1.25 0.27
Planning 1 0.68