Journal of Economic Behavior and Organization 149 (2018) 185–196

Contents lists available at ScienceDirect

Journal of Economic Behavior and Organization

journal homepage: www.elsevier.com/locate/jebo

An examination of how the effort-inducing property of

incentive compensation influences performance in
multidimensional tasks
Ling Harris a, Michael Majerczyk b,∗, Andrew H. Newman a
a
Darla Moore School of Business, University of South Carolina, Columbia SC, United States
b
J. Mack Robinson College of Business, Georgia State University, 35 Broad St. Office 517, Atlanta, GA 30309, United States

a r t i c l e i n f o a b s t r a c t

Article history: Employees often find themselves working in organizations where their jobs have multi-
Received 9 June 2017 ple dimensions. An often proposed solution to motivating employees in multidimensional
Revised 9 March 2018
environments is for organizations to use incentive compensation. Incentive compensation
Accepted 12 March 2018
can both direct employees’ attention (attention-directing property) and motivate their ef-
fort levels (effort-inducing property) but incentive compensation might not always be op-
Keywords: timal in multi-dimensional settings. Thus, it is important to better understand the two
Incentives proposed properties of incentive compensation through which incentive compensation mo-
Multidimensional tasks tivates employee performance. Conducting an experiment in a multidimensional environ-
Effort ment that contains quantity and quality task dimensions, we find that the incremental
Performance effectiveness of the effort-inducing property over the attention-directing property on per-
formance quantity and performance quality varies depending on the task dimension to-
ward which attention is directed and effort is induced. Specifically, the effort-inducing
property has a positive incremental effect over the attention-directing property on per-
formance quality when the quality dimension is incentivized; however, the effort-inducing
property does not have a positive incremental effect over the attention-directing property
on performance quantity when the quantity dimension is incentivized. This study provides
important insight regarding the properties of incentive compensation.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction

Employees often find themselves working in organizations where their jobs have multiple dimensions. Such environ-
ments may be due to the inherently multidimensional nature of a single task or due to the organizations explicitly choos-
ing to group multiple, separate tasks under a single job responsibility in order to leverage synergies across those tasks
(Lindbeck and Snower, 20 0 0). A key feature of multidimensional environments is that effort expended towards one dimen-
sion cannot be simultaneously expended towards other dimensions of the task (Hannan et al., 2013). Therefore, employees
must not only decide how much effort to expend but the amount of effort to expend along each dimension. Cognitive load
theory suggests that effort spent trying to figure out where to direct their effort leaves less effort available to actually devote
toward the task dimensions themselves (Sweller, 1988). Psychology research studying how individuals make tradeoffs finds

∗
Corresponding author.
E-mail address: mmajerczyk@gsu.edu (M. Majerczyk).

https://doi.org/10.1016/j.jebo.2018.03.010
0167-2681/© 2018 Elsevier B.V. All rights reserved.
186 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196

that individuals often take a hierarchical approach, prioritizing one task dimension over the other (Payne et al., 1993). A po-
tential issue with this approach is that an employee might prioritize a task dimension inconsistent with the dimension that
the organization wishes to prioritize. However, employers can intervene and provide cues to help employees prioritize the
desired task dimension. Accordingly, in order to best achieve the organization’s objectives, an organization not only needs to
consider how to motivate employees’ effort levels but also how to motivate employees to allocate their effort toward each
specific task dimension.
An often proposed solution for organizations to motivate employees, including in multidimensional environments, is
to use incentive compensation. Although incentive compensation might not always be optimal, it has potential usefulness
because, as Holmstrom and Milgrom (1991) identified in their multitask principal-agent model, incentive pay possesses
two properties. First, as in a single dimensional task, incentive compensation can act to invoke greater effort from the
employees (the effort-inducing property). Second, unique to a multidimensional task environment, incentive compensation
can act to direct employees on where to expend their effort (the attention-directing property). For example, if an incentive
contract weighs a certain task dimension more than another dimension employees are likely to direct their attention to
that specific task dimension. Meanwhile, the payoff parameters of the incentive contract serve to motivate an appropriate
level of effort in that desired direction. Utilizing a multidimensional environment that contains quantity and quality task
dimensions, we investigate the incremental effectiveness on performance of the effort-inducing property above assumed
benefits of the attention-directing property, including whether any incremental effectiveness of the effort-inducing property
varies depending on the task dimension toward which attention is directed and effort is induced.
Our study provides important insight regarding how to efficiently design employment contracts that better leverage the
particular incentive compensation property(ies) that is (are) most effective for the task dimension the organization wants
to emphasize. Prior research has tended to study either the overall effect of incentive compensation design on effort and
performance (see Garbers and Konradt, 2014 and Jenkins et al., 1998 for reviews of this research) or to identify settings
whereby fixed wages might be preferred to incentive compensation (e.g., Holmstrom and Milgrom, 1991). However, it is
unknown once employees’ attention has been directed to a specific task dimension, what the incremental benefit of the
effort-inducing property of incentive compensation on employee effort is.
Several theories from psychology and economics suggest that incentives will increase performance by positively moti-
vating output-based efforts. For instance, expectancy theory (see Vroom, 1964; Van Eerde and Thierry, 1996; among many
others) and agency theory (see Ross, 1973; Baiman, 1982, 1990; Eisenhardt, 1989; Holmstrom, 1989; Lambert, 2001; among
many others) suggest that incentives increase performance because incentives motivate individuals to exert effort. When
these theories are applied in a multidimensional environment, the inference is that incentive compensation has an effort-
inducing property that will increase performance. It is important, however, not to consider the effort-inducing property of
incentive compensation in isolation. Specifically, directing employees’ attention toward the quantity dimension, for instance,
should cause employees to emphasize speed of production more than quality of production and therefore motivate employ-
ees to increase their quantity output (Forster et al., 2003), compared to when the firm emphasizes the quality dimension.
Conversely, directing employees’ attention toward the quality dimension should focus employees on minimizing mistakes
and make the employees more vigilant with regard to the quality of each unit of output (Forster et al., 2003) compared to
when the firm emphasizes the quantity dimension.
As directing employees’ attention to a task dimension will increase employees’ performance on that dimension, the crit-
ical question regarding the effort-inducing property is whether or not it has an incremental effect on employees’ output
quantity or quality beyond what is driven by directing attention to that dimension. Currently, this is an open empirical
question. On the one hand, the effort-inducing property might propel employees to continue to exert effort towards that di-
mension beyond their motivation stemming from having their attention directed towards the quantity (quality) dimension.
On the other hand, simply directing attention towards quantity (quality) may be sufficient to lead employees to reach a
point where their output is maximized such that the effort-inducing property might not propel employees to increase their
output (i.e., a ceiling effect). Given the lack of a clear theoretical model capable of establishing formal predictions amenable
to testing, we view our study as providing empirical evidence that can be used to assist in the development of future theory
and formal propositions related to incentive compensation in multidimensional settings.
We investigate the incremental effectiveness on performance of the effort-inducing property above assumed benefits of
the attention-directing property by conducting an experiment whereby employees complete a multidimensional task that
requires them to allocate their effort between production quantity and production quality.1 Specifically, we examine when
incentive compensation is quantity-based whether the effort-inducing property of the incentive compensation will have a
positive effect on employees’ output quantity, whereas when incentive compensation is quality-based whether the effort-
inducing property of the incentive compensation will have a positive effect on employees’ output quality.
Using an experiment allows us the opportunity to isolate the effect of the effort-inducing property on performance
from the effect of the attention-directing property. This is because in practice it is extremely difficult to tease out how
the attention-directing and effort-inducing properties of incentive compensation separately influence performance. In our

1
In order for both the attention-directing and effort-inducing properties of incentive compensation to play a role a multidimensional setting must
be present. In a single-dimensional setting there is no need to direct attention as it is clear where employees should focus their attention. Thus, in such
settings, incentive compensation only serves an effort-inducing role. We use both a quantity and a quality dimension in our study because these dimensions
are commonly observed in practice (Kim and Gershwin, 2005).
 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196 187

experiment, we operationalize informal communication to direct participants’ attention to either the quality or quantity di-
mension of the task by using an organization’s mission statement and utilize this informal organization communication as
our experimental proxy for the attention-directing property of incentive compensation. Prior research suggests that while
incentive contracts can serve an attention-directing role, organizations also often have non-incentive compensation related
mechanisms in place that can direct employees’ attention in a particular direction (Umpathy, 1987; Newman, 2014). Infor-
mation communication can serve as an experimental proxy because fundamentally it should convey the same information
as incentives do in terms of where the organization wants its employees to direct their attention; however, the information
is not related to their compensation.
We conduct a 2 × 2 between-participants experiment where employees allocate their effort between production quantity
and production quality. We manipulate the informal communication by varying whether the organization’s mission state-
ment focuses on quantity or quality-based production. We choose to have the informal communication always be present
because, given that organizations usually provide some level of guidance to their employees, we are more interested in
how the focus of the communication influences performance than how its presence influences performance. We manipulate
the presence of incentive compensation by varying participants’ compensation scheme at two levels (i.e., flat-wage com-
pensation or incentive-based compensation). Incentive compensation is always based on the emphasized task dimension
(i.e., when quantity (quality) is emphasized via the mission statement, then the incentive compensation is based on higher
quantity (quality)).
The results of our experiment show that the effectiveness of the effort-inducing property varies depending on the task
dimension. Specifically, offering incentive compensation for quantity in addition to emphasizing quantity has no incremental
positive effect on quantity performance compared to only emphasizing quantity; in contrast, we find that offering incentive
compensation for quality in addition to emphasizing quality does have an incremental positive effect on quality perfor-
mance compared to only emphasizing quality. Taken together, our results suggest that in terms of incentive compensation
influencing employee performance, the effort-inducing property has a positive incremental effect on performance relative to
the attention-directing property on the quality dimension but not on the quantity dimension. Supplemental analysis regard-
ing the effort-inducing property finds that incentivizing quantity has no significant spillover effect on quality performance
whereas incentivizing quality increases quantity performance.
The rest of the paper is structured as follows. In Section 2, we present in detail our experimental design and procedures.
In Section 3 we provide results with Section 4 concluding.

2. Experimental design and procedures

2.1. Task description

The experiment was conducted in a computer lab at a large state university in the United States. We recruited partici-
pants from business courses at this university and seventy-three business students took part in our study.2 The mean age
of our participants is 23.1 years, and 39.7 percent are male. There are no significant differences in age or gender across
conditions (all p’s > 0.483 ). Therefore, we do not include age or gender in our analyses.
We adapt a contextual, computerized production task used by Farrell et al. (2008, 2012). Specifically, participants assume
the role of a sandwich-maker working for a virtual sandwich shop and complete customers’ sandwich orders using a com-
puter. Each order is selected from a pool of pre-programmed sandwiches, with each sandwich containing from six to eight
ingredients. The sequence of sandwich orders in each round is identical for all participants.
Participants completed the task for two production rounds (after a practice round).4 Each production round lasts six
minutes and a clock appears on the screen to help participants keep track of their time. Fig. 1 shows a sample screen of
the sandwich-making task. After a customer’s sandwich order appears in the “Order” window on the top of the computer
screen, participants can access the sandwich’s ingredient list using a drop-down “Menu” box. The ingredient list disappears
from the screen when participants move the computer mouse away from the “Menu” box. In order to access the ingredient
list again participants must re-select the sandwich from the drop-down “Menu.” Participants use the drop-down menus
for the five different types of ingredients (breads, meats, cheeses, vegetables, and condiments) to assemble a sandwich
and participants can add ingredients in any order they choose. Once an ingredient is added to the sandwich an image of
that ingredient appears in the production space in the center of the screen. Participants can remove ingredients from the
sandwich by using the “Remove Ingredients” button located at the top of the screen. Participants click on the “Finished”
button to complete a sandwich order.

2
Seventy-seven students participated in the experiment. Computer failure rendered the data from four participants (across all conditions) incomplete.
Thus, we report results related to the usable data from seventy-three participants.
3
All reported p-Values in the text are two-tailed.
4
In total, participants completed four production rounds in the lab. We designed the first two production rounds to test our research questions outlined
in this paper. We designed the final two production rounds to test separate research questions for another study that focuses on how individuals perform
when they receive additional information beyond that provided in the first two rounds. In order to do so we subjected participants to an additional
manipulation after the second production round. We would not expect this additional manipulation to have any effect on performance in the first two
rounds as participants were unaware of this manipulation at the time they completed the first two rounds. Because their behavior and performance on the
final two rounds is beyond the scope of this study, we focus our discussion only on performance for the first two rounds going forward.
188 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196

Fig. 1. Sample experimental screen.

Notes: The above screen shot is a sample screen of the sandwich making task. A customer’s sandwich order appears in the “Order” window on the top
of the computer screen. Participants can access the sandwich’s ingredient list using a drop-down “Menu” box. The ingredient list disappears from the
screen when participants move the computer mouse away from the “Menu” box. In order to access the ingredient list again participants must reselect the
sandwich from the drop-down “Menu.” Participants use the drop-down menus for the five different types of ingredients (breads, meats, cheese, vegetables,
and condiments) to assemble a sandwich and participants can add ingredients in any order they choose. Once an ingredient is added to the sandwich
an image of that ingredient appears in the production space in the center of the screen. Participants can remove ingredients from a sandwich by using
the “Remove Ingredients” button located at the top of the screen. Participants click on the “Finished” button to complete a sandwich order. At this point,
the computer program determines a sandwich’s acceptability by checking the assembled sandwich against its ingredient list to determine the number of
mistakes. A sandwich is considered acceptable and added to a participant’s production tally as long as it has less than four mistakes. For each acceptable
sandwich, an “Order Complete” message appears on the screen, and the next sandwich order appears automatically in the “Order” window. If the sandwich
has four or more mistakes, the participant receives a message stating that there are too many mistakes and that the participant must redo the order. Each
production round lasts six minutes and the time remaining in the current production round is displayed on the screen at all times.

At this point, the computer program determines a sandwich’s acceptability by checking the assembled sandwich against
its ingredient list to determine the number of mistakes. A sandwich is considered acceptable and is added to a participant’s
production tally as long as it has less than four mistakes.5 For each acceptable sandwich, an “Order Complete” message
appears on the screen, and the next sandwich order appears automatically in the “Order” window. If the sandwich has four
or more mistakes, the participant receives a message stating that there are too many mistakes and that the participant must
redo the order. This prevents participants from trying to increase their production tally by skipping the sandwiches with
more ingredients.
All participants automatically receive individual performance feedback at the end of each round. Participants learn the
total number of acceptable sandwiches they completed as well as how many sandwiches they made with zero, one, two,
three, and four or more mistakes, respectively. Since sandwiches with four or more mistakes have to be remade, these
sandwiches are not included in computing the total number of acceptable sandwiches completed. The performance feedback
remains on the screen for thirty seconds.
The sandwich-making production task is a multidimensional task that includes both quantity and quality dimensions. We
use two primary dependent variables. Quantity performance measures the quantity dimension of the task by counting the
number of acceptable sandwiches a participant completed during a round (i.e., the number of sandwiches with less than
four mistakes). Quality performance measures the quality dimension of the task by calculating a quality ratio that divides the

5
It is explicitly stated in the instructions that an acceptable sandwich has less than four mistakes; therefore, participants are fully aware what constitutes
an acceptable sandwich before beginning the task.
 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196 189

number of total mistakes a participant made in a round by the total number of acceptable sandwiches the same participant
completed in that round.

2.2. Task dimension emphasis manipulation

As discussed previously, in our experiment, we operationalize an organization’s mission statement to direct participants’
attention to either the quality or quantity dimension of the task and utilize this informal organization communication as
our experimental proxy for the attention-directing property of incentive compensation. In the Quantity Emphasis condition,
we present participants with a mission statement that states, “(W)hile we at HMN Sandwich Shop maintain high quality, we
seek to serve as many customers as possible and be the quantity-leader in the industry.” In the Quality Emphasis condition,
we present participants with a mission statement that states, “(W)hile we at HMN Sandwich Shop want to serve as many
customers as possible, we seek to maintain the highest quality in our sandwiches and to be the quality-leader in the indus-
try.” Thus, the organization always emphasizes one dimension of the task but we vary between-subjects which dimension
it emphasizes.

2.3. Compensation manipulation

We manipulate compensation at two levels. In the Flat-Wage condition, which serves as our baseline, each participant
receives $2.50 per production round in addition to a $5 show-up fee. In the Incentive Compensation condition, participants
are told that each of them may earn additional pay based on their performance in addition to a $5 show-up fee. Specifi-
cally, in the Incentive Compensation condition, when quantity is emphasized participants are told that the more acceptable
sandwiches they produce the more money they will earn, whereas when quality is emphasized participants are told that the
fewer mistakes they make when producing sandwiches the more money they will earn. Participants received their compen-
sation at the end of the experiment according to their total compensation for all rounds. Note that the variable incentive
compensation criteria are always aligned with the emphasis conveyed in the organization’s mission statement. We did this
because verbally emphasizing one dimension of a task while financially incentivizing another dimension of a task sends
conflicting messages, which we believe is less likely to occur in the real world. Also, even if they do occur in practice, given
our focus on investigating the performance effects of the motivating properties of incentive compensation examining how
participants resolve conflicting messages is beyond the scope of our study.
Having documented a performance level associated with the mission statement proxy in the presence of a flat wage for
the attention-directing property, we can then use our experimental design to provide evidence on any incremental perfor-
mance effect of the effort-inducing property by examining whether incentive compensation on either task dimension affects
employee performance relative to the mission statement.
Participants have no uncertainty regarding which task dimension performance does (does not) lead to higher compen-
sation. We also inform participants of their expected payout range ($9–$21, including $5 for showing up). However, we
purposely do not inform these participants of the specific formulaic basis by which increased performance translates into ad-
ditional compensation.6 We made this design decision to keep any specific incentive-related parameters from inadvertently
affecting participant behavior on our variables of interest. Although this design approach creates some noise for participants,
the level of noise is constant across the two incentive compensation conditions.7 Such an incentive contract design not only
mirrors many settings in practice whereby noise is present in employees’ compensation structures (Gibbs et al., 2003; It-
tner et al., 2003; Rajan and Reichelstein, 2006), but also has been widely used by prior research (e.g., Bloomfield et al.,
20 09a, 20 09b) when researchers want to avoid specific incentive compensation parameters from having unintended effect
on individual behavior. Finally, for any non-incentivized dimension of the production task, we employ an internal control in
each round.8 The purpose of the control is to keep participants thinking about all dimensions of the task. This is consistent
with organizations taking measures to keep their employees from ignoring certain task dimensions altogether when these
dimensions are not the primary focus of the organization.

6
The specific payout for participants in the incentive conditions is determined as follows. In each production round, participants were ranked into
quartiles for the number of acceptable sandwiches (quantity compensation condition) or for the total number of mistakes (quality compensation condition).
Participants in the top 25 percentile were awarded $4 in a round. Participants in the next quartiles in the same round were then awarded $3, $2, and $1,
respectively.
7
We discuss in the Discussion and Conclusion section the opportunity for future research to investigate whether varying the incentive compensation
parameters (e.g., vague versus specific) influences the effect of the effort-inducing property of incentive compensation on employee performance.
8
Specifically, at 90-s intervals within each 360-s (6-min) round a pop-up box with performance feedback appears on the screen. The pop-up box
remains on the screen for 10 s during which time participants cannot actively work on assembling sandwiches. The round clock is paused during this
10 s time to ensure participants have the entire 360 s in each round to perform the production task. When incentive compensation is based on quality
(quantity) performance, we provide the participant with periodic in-round feedback on the production quantity (quality) by indicating the number of
acceptable sandwiches completed (mistakes made) by this participant thus far in the current round. When neither of the task dimensions is incentivized
and participants receive a flat-wage for the round, the periodic feedback informs participants both of their number of acceptable sandwiches and their
total number of mistakes thus far in the current round.
190 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196

Table 1
Descriptive statistics

Panel A: descriptive statistics of quantity performance

Flat-wage Incentive-compensation
(1) (2)

Quantity emphasis 9.06 9.75

(3.38) (3.45)
Quality emphasis 6.80 8.32
(2.18) (1.49)
Subjects 36 37

Panel B: descriptive statistics of quality performance

Flat-wage Incentive-compensation
(1) (2)

Quantity emphasis 1.12 1.27

(1.03) (1.21)
Quality emphasis 0.55 0.11
(0.96) (0.15)
Subjects 36 37

Notes: See notes to Fig. 2 for the descriptions of each variable.

2.4. Session timeline

We conducted a total of eight experimental sessions and participants were assigned to an experimental condition based
on which session they attended. At the start of each session, we provided participants with written instructions detailing
the experiment. Following a review of the instructions by one of the experimenters, participants completed a practice round
in order to familiarize themselves with the sandwich-making task. Next, participants took a short quiz to ensure that they
understood the experimental task and their compensation structure, as well as could recall the task dimension emphasized
via the mission statement. After all participants in the session successfully completed the quiz, they began the first pro-
duction round. Upon completion of the production task, participants answered a series of post-experimental questions that
allowed us to gather demographic and process-related information. Following this, participants received payment. Partici-
pants received an average of $15, including a show-up fee of $5.

3. Results

3.1. Descriptive statistics

Recall that we are investigating the incremental effectiveness on performance of the effort-inducing property above as-
sumed benefits of the attention-directing property and whether any incremental effectiveness of the effort-inducing prop-
erty varies depending on the task dimension toward which attention is directed and effort is induced. Table 1 and Fig. 2
report descriptive statistics for the dependent variables used to test these research questions. Panels A and Panel B provide
descriptive statistics for participants’ quantity performance and quality performance, respectively. We average together per-
formance across the two production rounds when calculating each dependent measure.9 Performance on a dimension (e.g.,
quality dimension) appears to be higher when that dimension was emphasized by the mission statement (e.g., quality di-
mension) compared to when the other dimension was emphasized (e.g., quantity dimension). Furthermore, performance ap-
pears to improve on each dimension when incentive compensation related to the emphasis is also present. The combination
of these two descriptive results suggests that performance is potentially benefiting from both properties (attention-directing
and effort-inducing) of incentive compensation. We provide the box and whisker plots in Fig. 3 for a visualization of our
dependent measures by condition averaged across rounds.

3.2. Test of proxy for attention-directing property

We begin our analysis by providing support for our attention-directing proxy of informal communication before moving
to our central research question of the incremental effectiveness of the effort-inducing property on each specific dimension.
Accordingly, we first focus our tests within the Flat-Wage condition because this controls for any effort-inducing effect on
performance from incentive compensation. Specifically, we test (a) whether communicating a preference for the quantity
dimension rather than the quality dimension leads to comparably higher output quantity, and (b) whether communicating a
preference for the quality dimension rather than the quantity dimension leads to comparably higher output quality.

9
All findings reported in this section are inferentially unchanged by conducting individual analysis on each of the two rounds.
 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196 191

10

9.5

Acceptable Sandwiches Made

9

8.5

8 Quality Emphasis
7.5 Quantity Emphasis

6.5

6
Flat-wage Incentive
Compensation Compensation

1.4

1.2
Acceptable Sandwich
Average Mistakes per

0.8
Quality Emphasis
0.6
Quantity Emphasis
0.4

0.2

0
Flat-wage Incentive
Compensation Compensation

Fig. 2. Performance by compensation condition Panel A: quantity performance. Panel B: quality performance.
Notes to Panel A: Quantity Performance is the average number of acceptable sandwiches participants made in the two production rounds. A sandwich is
considered to be acceptable when it has fewer than four mistakes. In the Quality Emphasis condition, participants are presented with a mission statement
that reads, “(w)hile we at HMN Sandwich Shop want to serve as many customers as possible, we seek to maintain the highest quality in our sandwiches
and to be the quality-leader in the industry.” In the Quantity Emphasis condition, participants are presented with a mission statement that reads, “(w)hile
we at HMN Sandwich Shop maintain high quality, we seek to serve as many customers as possible and be the quantity-leader in the industry.” In the Flat-
Wage condition participants receive $2.50 in each production round regardless of their performance. In the Incentive Compensation condition participants are
paid based on their performance. Specifically, in the Quantity-Emphasis condition participants are told that the more acceptable sandwiches they produce
the more money they will earn. In the Quality-Emphasis condition, participants are told that the fewer mistakes they make when producing sandwiches
the more money they will earn. Notes to Panel B: Quality Performance is the average number of total mistakes in the two production rounds divided by
the average number of acceptable sandwiches produced in the two production rounds. In the Quality Emphasis condition, participants are presented with a
mission statement that reads, “(w)hile we at HMN Sandwich Shop want to serve as many customers as possible, we seek to maintain the highest quality
in our sandwiches and to be the quality-leader in the industry.” In the Quantity Emphasis condition, participants are presented with a mission statement
that reads, “(w)hile we at HMN Sandwich Shop maintain high quality, we seek to serve as many customers as possible and be the quantity-leader in the
industry.” In the Flat-Wage condition participants receive $2.50 in each production round regardless of their performance. In the Incentive Compensation
condition participants are paid based on their performance. Specifically, in the Quantity-Emphasis condition participants are told that the more acceptable
sandwiches they produce the more money they will earn. In the Quality-Emphasis condition, participants are told that the fewer mistakes they make when
producing sandwiches the more money they will earn.

As reported in Panel A of Table 1, we find that when the organization communicated a quantity emphasis under a flat-
wage contract participants completed an average of 9.06 acceptable sandwiches compared to 6.80 when the organization
communicated a quality emphasis under a flat-wage contract. We use a two-side Mann–Whitney U test to conduct a pair-
wise comparison.10 Results of this test yield a p-value equal to 0.057 (not tabulated). As reported in Panel B of Table 1, we
find that when the organization communicated a quality emphasis under a flat-wage contract, participants made an aver-
age of 0.55 mistakes per acceptable sandwich compared to 1.12 mistakes per acceptable sandwich when the organization

10
We use a nonparametric test for the analyses because the Shapiro test indicates that our data may not follow a normal distribution.
192 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196

Fig. 3. Boxplots of dependent measures Panel A: quantity performance by condition. Panel B: quality performance by condition.
Notes: See notes to Fig. 2 for the descriptions of each variable.

communicated a quantity emphasis under a flat-wage contract. We also use a two-side Mann–Whitney U test to conduct
this pairwise comparison. Results of this test yield a p-value equal to 0.018 (not tabulated).11 The combination of these

11
As stated previously, we measure quality as a ratio of the number of a participant’s total mistakes in a round divided by the total number of acceptable
sandwiches that participant produced in that round. Our results are inferentially identical if we use other ways to measure “quality” such as the total
mistakes in a round (regardless of the number of acceptable sandwiches made) or by calculating a ratio that excludes the mistakes made for unacceptable
sandwiches.
 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196 193

two significant pairwise comparison tests provides support that our proxy was effective in directing attention towards the
emphasized dimension.

3.3. Test of effect of effort-inducing property

Having used a proxy for the attention-directing property of incentive compensation via the use of informal communica-
tion of the organization’s mission statement, we examine whether there is an incremental benefit from the effort inducing
property of incentive compensation. To test our research question, we hold organization emphasis constant and examine the
effects of the compensation structure (i.e., flat-wage versus incentive compensation) on quantity and quality performance.
Doing so allows us to examine the incremental performance effect of the effort-inducing property of incentive compensation
while controlling for any performance effect associated with the attention-directing property.
When the mission statement emphasized quantity, participants completed an average of 9.06 acceptable sandwiches in
the Flat-Wage condition compared to 9.75 sandwiches in the Incentive Compensation condition (see Panel A of Table 1).
We conduct two different analyses to test for differences in quantity performance between these two conditions. First, we
perform a two-sided Mann–Whitney U test, which yields a p-value equal to 0.589 (not tabulated). Second, we use OLS re-
gression and Panel A of Table 2 tabulates the results. Given the multiperiod nature of the design, we run independent OLS
regressions for each production round individually (columns 1 and 2) and we also run an OLS regression with clustered
standard errors for the data as a whole (column 3). In the third column, standard errors are clustered at the individual
level using robust clustering to control for the within-subject observations. The dependent variable is the number of ac-
ceptable sandwiches. As shown in Panel A of Table 2, there is no observable effect of incentive compensation on quantity
performance when quantity is emphasized via a mission statement. Taken together, the Mann–Whitney analysis and the
regression analysis suggest that when the organization emphasizes and compensates quantity there is no incremental ben-
efit of the effort-inducing property on participants’ quantity performance over and above the positive performance effect
associated with the attention-directing ability of informal communication.
When the mission statement emphasized quality, participants made an average of 0.55 mistakes per acceptable sand-
wich in the Flat-Wage condition compared to 0.11 mistakes in the Incentive Compensation condition (see Panel B of Table 1).
Again, we conduct two different analyses to test for differences in quality performance between these two conditions. First,
we perform a two-sided Mann–Whitney U test, which yields a p value equal to 0.137 (not tabulated). Second, we use a
TOBIT regression model.12 Panel B of Table 2 tabulates the results of the TOBIT regressions. The dependent variable is the
number of mistakes per acceptable sandwich. Similar to the OLS regressions with quantity performance, we use robust clus-
tering of the standard error when testing across periods and run individual regressions for each round separately (columns
1 and 2) in addition to the test of the overall data (column 3). As shown in Panel B of Table 2, there is also an observ-
able effect of incentive compensation on quality performance when quality is emphasized via a mission statement. Taken
together, the Mann–Whitney analysis and the regression analysis provide evidence that when the organization emphasizes
and compensates quality, the effort-inducing property of incentive compensation has a positive performance effect on qual-
ity performance over and above the positive performance effect associated with the attention-directing ability of informal
communication
Thus, we find evidence that the incremental effect of the effort-inducing property of incentive compensation is condi-
tional. While there was no observable incremental effect of the effort-inducing property on quantity performance when
quantity was emphasized in the mission statement, we observe evidence of an incremental benefit on quality performance
when quality is emphasized. These results suggest directing employees’ attention to the quantity dimension may be suffi-
cient to maximize quantity output (i.e., a ceiling effect) whereas there is room for both directing employees’ attention to the
quality dimension and utilizing incentive compensation to have a positive influence on quality output. In the next section
we examine spillover effects on the non-emphasized dimension.

3.4. Supplemental analysis – performance spillover

A performance spillover occurs in a multidimensional task when incentive compensation not only influences employee
performance on the primary task dimension toward which the compensation is focused, but also influences employees’
performance on other dimensions of the task. While Holmstrom and Milgrom (1991) highlight the potential for such in-
complete contracts to harm performance on the non-compensated task dimensions, empirical evidence tends to document
that such incomplete contracts either have no performance effect on the non-incentivized task dimensions (see Bonner and
Sprinkle, 2002 for a review) or could actually have a positive performance effect on those dimensions due to a spillover
effect (Hecht et al., 2012). Given the mixed evidence in prior research, we examine whether there are any spillover effects
of the effort-inducing property of incentive compensation. Similar to the test of our research questions, we hold organiza-
tion emphasis constant and examine the effects of the compensation structure (flat-wage versus incentive compensation)
on quantity and quality performance so we can examine the incremental performance effect of the effort-inducing property

12
We use a TOBIT model because the data for participants’ quality performance is censored to the left based on the high frequency of zero mistakes by
participants and to the right based on the cap on mistakes per acceptable sandwiches (4 or more mistakes result in an unaccepted sandwich).
194 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196

Table 2
Statistical tests

Panel A: effect of effort-inducing property on quantity performancea

Quantity emphasis

Round 1 Round 2 Total

Incentive compensation 0.61 0.78 0.69

(0.61) (0.59) (0.61)
Intercept 8.50∗ ∗ ∗ 9.61∗ ∗ ∗ 8.46∗ ∗ ∗
(11.98) (10.38) (11.47)
Round 1.19∗ ∗ ∗
(4.68)
Obs 36 36 72
b
Panel B: effect of effort-inducing property on quality performance

Quality emphasis

Round 1 Round 2 Total

∗
Incentive compensation −0.88 −0.62 −0.78∗
(1.49) (1.73) (1.71)
Intercept 0.07 0.05 0.17
(0.17) (0.21) (0.52)
Round −0.30
(1.01)
Obs 37 37 74

Panel C: spillover effect of effort-inducing property on quantity performancea

Quality emphasis

Round 1 Round 2 Total

Incentive compensation 1.44∗ ∗ 1.58∗ ∗ 1.51∗ ∗

(2.28) (2.52) (2.45)
Intercept 6.61∗ ∗ ∗ 7.00∗ ∗ ∗ 6.57∗ ∗ ∗
(14.60) (15.61) (12.69)
Round 0.46∗ ∗ ∗
(3.06)
Obs 37 37 74

Panel D: spillover effect of effort-inducing property on quality performanceb

Quantity emphasis

Round 1 Round 2 Total

Incentive compensation 0.09 0.03 0.06

(0.21) (0.16) (0.12)
Intercept 0.83∗ ∗ ∗ 1.20∗ ∗ ∗ 0.83∗ ∗ ∗
(2.78) (3.43) (2.91)
Round 0.37∗ ∗ ∗
(3.07)
Obs 36 36 72

See notes to Fig. 2 for the descriptions of each variable.

∗
Indicates p-value < 0.10.
∗∗
Indicates p-value < 0.05.
∗∗∗
Indicates p-value < 0.01.
a
Tests reflect coefficients (and t-statistics) from OLS with clustered standard errors. All values are
two-tailed.
b
Tests reflect coefficients (and t-statistics) from TOBIT regression with clustered standard errors. All
values are two-tailed.

of incentive compensation while controlling for any performance effect associated with the attention-directing property of
our attention-directing proxy.
When the mission statement emphasized quality, we compare quantity performance when participants are paid a flat-
wage to when participants receive incentive compensation based on their quality performance. As reported in Panel A of
Table 1, the average number of acceptable sandwiches made is 6.80 in the Flat-Wage condition and 8.32 in the Incentive
Compensation condition. Similar to our main analysis, we utilize two statistical tests to examine whether there are spillover
effects. First, we perform a two-sided Mann–Whitney U test, which yields a p-value equal to 0.003 (not tabulated). Second,
we perform OLS regression analysis. As reported in Panel C of Table 2, the regression result indicates that there is a statisti-
cally significant difference in quantity performance in these two conditions overall and separately in each round. Specifically,
we find that performance on the non-emphasized quantity dimension improves when incentive compensation is provided
on the emphasized quality dimension. Taken together, this analysis is consistent with there being a positive spillover effect
 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196 195

of the effort-inducing property of incentive compensation on quantity performance when quality is incentivized. In other
words, incentivizing quality leads to higher quantity performance as well.
A closer examination of this spillover analysis reveals that the 8.32 sandwiches made when participants receive incen-
tive compensation based on their quality performance does not significantly differ from the 9.06 sandwiches participants
make when quantity performance is emphasized and participants receive a flat-wage (Mann–Whitney test; p = 0.679, not
tabulated). This implies that the positive performance spillover effect on quantity performance associated with the effort-
inducing property of quality-based incentive compensation (i.e., the increase in quantity from 6.80 to 8.32) serves to neu-
tralize the negative performance spillover effect on quantity performance associated with the directing of attention towards
quality via the mission statement (i.e., the decrease in quantity from 9.06 to 6.80).
When the mission statement emphasized quantity, we compare quality performance when participants are paid a flat-
wage to when participants receive incentive compensation based on their production quantity. As reported in Panel B of
Table 1, the average number of mistakes per acceptable sandwich is 1.12 in the Flat-Wage condition and 1.27 in the In-
centive Compensation condition. Similar to our main analysis, we utilize two statistical tests to examine whether there are
spillover effects. First, we perform a two-sided Mann–Whitney U test, which yields a p-value equal to 0.857 (not tabu-
lated). Second, we perform regression analysis. Specifically, when the mission statement emphasized quantity, we compare
quality performance when participants are paid a flat-wage to when participants receive incentive compensation based on
their production quantity. Again, because of the censored nature of the data related to quality performance we use a TOBIT
model. As reported in Panel D of Table 2, we find that there is no statistically significant difference in quality performance in
these two conditions. Taken together, the analysis indicates that no performance spillover is present, which combined with
the analysis in the previous section, implies that when the mission statement emphasized quantity that the effort-inducing
property of the incentive compensation has no incremental effect on either quantity or quality performance.

4. Discussion and conclusion

Despite the abundant research on incentive compensation in economics and other disciplines, the literature related to
the properties of incentive compensation in multidimensional tasks laid out in Holmstrom and Milgrom’s (1991) multitask
principal-agent model has received limited attention in the experimental literature. Specifically, although incentive compen-
sation has both attention-directing and effort-inducing properties in multidimensional tasks, research tends to focus on the
overall effects of incentive compensation and has rarely considered the incremental effectiveness of these two properties or
what may influence their relative effectiveness. This paper provides empirical data on the effort-inducing property’s incre-
mental effectiveness, controlling for the attention-directing property of incentive compensation. Our results show that the
effectiveness of the effort-inducing property is conditional on the task dimension. Specifically, the effort-inducing property
has a positive incremental effect on performance relative to the attention-directing property on the quality dimension but
not on the quantity dimension.
Our study helps organizations consider how to best motivate employees’ effort in a multidimensional task with quan-
tity and quality task dimensions. To our knowledge we are the first study to provide empirical evidence in an attempt to
disentangle the effectiveness of the attention-getting and effort-inducing properties of incentive compensation on employee
performance. In doing so, we are able to offer important insight regarding the role the properties play in making incen-
tive compensation an effective effort-motivating mechanism in multidimensional environments. This insight allows us to
contribute to contract design research focused on better understanding the value of incentive compensation in multidimen-
sional tasks (e.g., Holmstrom and Milgrom, 1991; Feltham and Xie, 1994; Datar et al., 2001; Farrell et al., 2008; Kachelmeier
et al., 2008; Hecht et al., 2012) as well as to the broader stream of research focused on understanding the effect of financial
incentives on performance in general (e.g., see reviews by Bonner and Sprinkle, 2002, and Garbers and Konradt, 2014).
Our study also contributes to incentive compensation research on spillover effects (e.g., Holmstrom and Milgrom, 1991;
Bonner and Sprinkle, 2002; Hecht et al., 2012). Prior studies have attempted to balance incentives and foster spillover ef-
fects through interventions such as contingent audits (Sinclair-Desgangé, 1999) and internal competitive financial indicators
(Gautier and Wauthy, 2007). Participants in our study always work under an incomplete contract in that incentives are either
not offered at all (emphasis-only condition) or offered but not on all dimensions of the task, with no additional contractual
mechanisms. While prior analytical research (Holmstrom and Milgrom, 1991) highlights the potential for such incomplete
contracts to harm performance on the non-compensated task dimensions, empirical evidence tends to document that such
incomplete contracts have no performance effect on the non-incentivized task dimensions (e.g., see reviews by Bonner and
Sprinkle, 2002 and Jenkins et al., 1998) or recent research even finds they could actually have a positive performance ef-
fect on those dimensions (Hecht et al., 2012). Regarding the mixed findings of prior empirical research, our analysis of the
spillover effects in our setting provides insight that suggests that the underlying task dimension being incentivized can
influence whether no spillover effect or a positive spillover effect occurs.
We conclude by acknowledging limitations of our study and identifying potential future research opportunities. First,
we use the informal communication of an organization’s preferences in the form of a mission statement to capture the
attention-getting property of incentive compensation because it allows us to design an experiment that separately mea-
sures the effects on performance of the attention-getting and effort-inducing properties of incentive compensation. While
we believe that it is a reasonable and novel way to do so, we acknowledge that informal communication may not be a
perfect proxy. Future research can test the robustness of this proxy as well as identify alternative proxies. Second, we focus
196 L. Harris et al. / Journal of Economic Behavior and Organization 149 (2018) 185–196

on a two-dimensional task that has quality and quantity dimensions. We focus on these two dimensions because they are
common to many tasks and are often studied. However, there can be other dimensions to tasks (e.g., creativity) and tasks
can also have more than two dimensions. Future research could examine whether our results are influenced by the specific
types of task dimensions present or the number of task dimensions. Third, we did not inform participants in the incentive
conditions of the exact relation between quality/quantity and their additional compensation. We did so because we felt it
was important to first examine a more straightforward setting before considering the potential effects of added complexities
of providing incentives on multiple task dimensions or the potential effects associated with the specific level of piece-rate
compensation. While prior research has examined how the form, structure, and level of incentive compensation influences
performance in general, future research could examine whether (and how) different forms, structures, or levels of incentive
compensation influences the effect that the underlying attention-getting and effort-inducing properties of incentive com-
pensation have on performance. Finally, just as research has identified person, task, environmental, and incentive scheme
variables that influence incentive compensation’s general effectiveness (see review by Bonner and Sprinkle, 2002), future
research could do so with regard to the effectiveness of incentive compensations’ underlying motivational properties in a
multidimensional environment.

Acknowledgments

We are grateful to Daniel Houser (Editor), and two anonymous reviewers, as well as, Jeremy Douthit, Daniel Jones, Joel
Owens, Bryan Stikeleather, Ivo Tafkov, Nate Waddoups, Crystal Zhan, Flora Zhou, workshop participants at Virginia Tech, PhD
seminar participants at Georgia State University, and participants at the Accounting, Behavior, and Organizations Research
Conference for their helpful comments. We gratefully acknowledge the Moore School of Business at the University of South
Carolina for funding this research.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jebo.2018.03.010.
References

Baiman, S., 1982. Agency research in managerial accounting: a survey. J. Account. Lit. 1, 154–213.
Baiman, S., 1990. Agency research in managerial accounting. Account. Organ. Soc. 15 (4), 341–371.
Bloomfield, R.J., O’Hara, M., Saar, G., 2009a. How noise trading affects markets: an experimental analysis. Rev. Financ. Stud. 22, 2275–2302.
Bloomfield, R.J., O’Hara, M., Saar, G., Tayler, W.B., Zhou, F., 2009b. Momentum, reversal, and uninformed traders in laboratory markets. J. Financ. 64,
2535–2558.
Bonner, S.E., Sprinkle, G.B., 2002. The effects of monetary incentives on effort and task performance: theories, evidence, and a framework for research.
Account. Organ. Soc. 27, 303–345.
Datar, S., Kulp, S.C., Lambert, R.A., 2001. Balancing performance measures. J. Account. Res. 39, 75–92.
Eisenhardt, K.M., 1989. Agency theory: an assessment and review. Acad. Manag. Rev. 14 (1), 57–74.
Farrell, A.M., Kadous, K., Towry, K.L., 2008. Contracting on contemporaneous versus forward-looking measures: an experimental investigation. Contemp.
Account. Res. 25 (3), 773–802.
Farrell, AM., Kadous, K., Towry, K.L., 2012. Does the communication of causal linkages improve employee effort allocations and firm performance? An
experimental investigation. J. Manag. Account. Res. 24, 77–102.
Feltham, G.A., Xie, J., 1994. Performance measure congruity and diversity in multi-task principal/agent relations. Account. Rev. 69, 429–453.
Forster, J., Higgins, E.T., Bianco, A.T., 2003. Speed/accuracy decisions in task performance: built-in trade-off or separate strategic concerns. Organ. Behav.
Hum. Decis. Process. 90, 148–164.
Garbers, Y., Konradt, U., 2014. The effect of financial incentives on performance: a quantitative review of individual and team-based financial incentives. J.
Occup. Organ. Psychol. 87, 102–137.
Gautier, A., Wauthy, X., 2007. Teaching versus research: a multi-tasking approach to multi-department universities. Eur. Econ. Rev. 51, 273–295.
Gibbs, M., Merchant, K.A., Van der Stede, W.A., Vargus, M.E., 2003. Determinants and effects of subjectivity in incentives. Account. Rev. 79 (2), 409–436.
Hannan, R.L., McPhee, G., Newman, A.H., Tafkov, I., 2013. The effect of relative performance information on performance and effort allocation in a multi-task
environment. Account. Rev. 88 (2), 553–575.
Hecht, G., Tafkov, I., Towry, K.L., 2012. Performance spillover in a multi-task environment. Contemp. Account. Res. 29 (2), 563–589.
Holmstrom, B., 1989. Agency costs and innovation. J. Econ. Behav. Organ. 12 (3), 305–327.
Holmstrom, B., Milgrom, P., 1991. Multitask principal–agent analyses: incentive contracts, asset ownership, and job design. J. Law Econ. Organ. 7, 24–52.
Ittner, C.D., Larcker, D.F., Meyer, M.W., 2003. Subjectivity and the weighting of performance measures: evidence from a balanced scorecard. Account. Rev.
78, 725–758.
Jenkins Jr, G.D., Mitra, A., Gupta, N., Shaw, J.D., 1998. Are financial incentives related to performance? A meta-analytic review of empirical research. J. Appl.
Psychol. 83, 777–787.
Kachelmeier, S.J., Reichert, B.E., Williamson, M.G., 2008. Measuring and motivating quantity, creativity, or both. J. Account. Res. 46, 341–373.
Kim, J., Gershwin, S.B., 2005. Integrated quality and quantity modeling of a production line. OR Spectr. 27, 287–314.
Lambert, R.A., 2001. Contracting theory and accounting. J. Account. Econ. 32, 3–87.
Lindbeck, A., Snower, D.J., 20 0 0. Multitask learning and the reorganization of work: from Tayloristic to holistic organization. J. Labor Econ. 18 (3), 353–376.
Newman, A.H., 2014. An investigation of how the informal communication of firm preferences influences managerial honesty. Account. Organ. Soc. 39 (3),
195–207.
Payne, J.W., Bettman, J.R., Johnson, E.J., 1993. The Adaptive Decision Maker. Cambridge University Press, Cambridge, UK.
Rajan, M.V., Reichelstein, S., 2006. Subjective performance indicators and discretionary bonus pools. J. Account. Res. 44 (3), 585–618.
Ross, S.A., 1973. The economic theory of agency: the principal’s problem. Am. Econ. Rev. 63 (2), 134–139.
Sinclair-Desgangé, B., 1999. How to restore higher-powered incentives in multitask agencies. J. Law Econ. Organ. 15 (2), 418–433.
Sweller, J., 1988. Cognitive load during problem solving: effects on learning. Cognit. Sci. 12 (2), 257–285.
Umpathy, S., 1987. Current Budget Practices in U.S. Industry. Quorum Books, New York, NY.
Van Eerde, W., Thierry, H., 1996. Vroom’s expectancy models and work-related criteria: a meta-analysis. J. Appl. Psychol. 81 (5), 575–586.
Vroom, V., 1964. Work and Motivation. Wiley, New York.