Description
Using data from three production units of a large manufacturing plant that employs production teams in its assembly
operations, this paper examines how changes made to an existing team-based incentive plan affects labor productivity,
product quality, and worker absenteeism. The firm switched from a piece-rates contract and an attendance bonus and instituted
a two tier incentive plan comprising two different but complementary performance-based bonus schemes: one tier
based on individual team performance and the other tier on plant-wide performance. The incentives were introduced concurrently
with management control initiatives intended to facilitate cooperation and monitoring among the teams. I find
significant productivity gains and improvements in quality and absenteeism associated with the new incentive plan. These
findings underscore two important points that have not been emphasized in existing empirical studies of incentive pay:
incentive contracts for teams generate superior performance using combination of incentives, and the need to introduce
organizational changes to facilitate cooperation and peer monitoring in tandem with incentive pay to capture greater
incentive effects in team production.
An analysis of changes to a team-based incentive plan and
its e?ects on productivity, product quality, and absenteeism
Francisco J. Roma´n
*
Rawls College of Business, Texas Tech University, Department of Accounting, P.O. Box 42101, Lubbock, TX 79409, United States
Abstract
Using data from three production units of a large manufacturing plant that employs production teams in its assembly
operations, this paper examines how changes made to an existing team-based incentive plan a?ects labor productivity,
product quality, and worker absenteeism. The ?rm switched from a piece-rates contract and an attendance bonus and insti-
tuted a two tier incentive plan comprising two di?erent but complementary performance-based bonus schemes: one tier
based on individual team performance and the other tier on plant-wide performance. The incentives were introduced con-
currently with management control initiatives intended to facilitate cooperation and monitoring among the teams. I ?nd
signi?cant productivity gains and improvements in quality and absenteeism associated with the new incentive plan. These
?ndings underscore two important points that have not been emphasized in existing empirical studies of incentive pay:
incentive contracts for teams generate superior performance using combination of incentives, and the need to introduce
organizational changes to facilitate cooperation and peer monitoring in tandem with incentive pay to capture greater
incentive e?ects in team production.
Ó 2008 Elsevier Ltd. All rights reserved.
Introduction
In recent years, ?rms have increasingly turned to
group and team-based incentive compensation to
reward worker productivity (Blinder, 1999; Brown
& Armstrong, 1999; DeMatteo, Eby, & Sundstrom,
1998; Milgrom & Roberts, 1992; Weitzman &
Kruse, 1999).
1
Recent surveys underscore the
importance of team-incentive pay and incentive
plans for teams among ?rms. For example, Brown
and Armstrong (1999) assert that over 50% of major
US and European corporations employ some form
of group- and team-based bonus scheme in their
compensation plans.
Despite its importance, team pay has not received
the attention it should from accounting researchers
and their e?ects on performance are poorly under-
stood (Indjejikian, 1999; Sprinkle, 2003; Young,
1999).
2
Though related literature from other ?elds
0361-3682/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2008.08.004
*
Tel.: +1 806 742 3138; fax: +1 806 742 3182.
E-mail address: [email protected]
1
For further discussion on the role of team-based incentive
plans and group compensation, refer to Blinder (1999).
2
Although various experimental studies in accounting had
examined various issues concerning teams (see, for example,
Rankin, 2004; Young, Fisher, & Lindquist, 1993), and most
recently, other papers have addressed the implications of incen-
tive compensation on group performance (see, Fisher, Pe?er, &
Sprinkle, 2003), little empirical evidence exists documenting the
performance implications of incentive plans for teams in ?eld-
settings.
Available online at www.sciencedirect.com
Accounting, Organizations and Society 34 (2009) 589–618
www.elsevier.com/locate/aos
document improvements in productivity with the
use of ?rm-wide incentives, such as pro?t-sharing
and gain-sharing plans (see Hansen, 1997; Knez &
Simester, 2001; Weitzman & Kruse, 1999), there is
very little research that explicitly addresses the per-
formance e?ects of incentive plans speci?cally
designed for production teams using ?eld data.
Accordingly, the focus of this research study is to
shed new light about the e?ects of incentive plans
for teams in ?eld-settings. Drawing on data from
three production units of a large manufacturing ?rm
(a lock manufacturer) that employs production
teams in its assembly operations, the study investi-
gates how workers react to changes made to an exit-
ing incentive plan and whether the underlying
changes led to improvements in labor productivity
and product quality, and in reducing absenteeism.
More speci?cally, I compare changes in perfor-
mance in these outcomes after this ?rm switched
from an incentive plan that relied on piece-rates to
a two tier incentive plan that rewarded performance
at the team-level and at the plant-level. Under the
new bonus structure, workers were rewarded at
the team-level with a bonus when team output and
quality meets or exceeds a pre-speci?ed target, and
at the plant-level with a gain-sharing incentive plan
that pays quarterly bonuses when production teams
achieve productivity and quality targets for the
entire factory. Additionally, a central feature of
the new incentive plan is that it was implemented
in conjunction with a new management control sys-
tem, which comprised several initiatives intended to
facilitate cooperation, reinforce monitoring, and
encourage peer pressure between workers. These
include the distribution of performance reports to
teams, mechanisms to facilitate the exchange of
information to improve performance, and rules to
punish shirking and uncooperativeness.
This is a rich research setting to explore the
e?ects of team-based incentive plans on perfor-
mance. Of particular interest is the fact that this ?rm
not only made substantive changes to the bonus
structure to its existing compensation plan, but as
mentioned above, it adopted several initiatives in
tandem with the incentives seeking better results.
Given the inherent di?culties underlying team
pay, e.g., the free-rider problem, possible collusion
arrangements, investigating whether such manage-
ment control initiatives when o?ered jointly with
incentive pay are more e?ective in motivating work-
ers to raise performance is particularly important.
Also, this issue warrants further investigation given
the fact that changes in compensation systems sel-
dom occur without other major transformations
within the ?rm. For example, Ichniowski, Shaw,
and Prennushi (1997) and Ichniowski and Shaw
(2003) note that incentive compensation is usually
accompanied with human resources practices to eli-
cit optimal performance from the workforce. None-
theless, the empirical evidence as to how incentive
pay and organizational practices in?uence perfor-
mance is limited. Bonner and Sprinkle (2002) elabo-
rate further on this point when they argue that
research concerning the e?ects of monetary incen-
tives should provide more insight into the mecha-
nisms by which incentives in?uence performance.
Lastly, many ?rms nowadays introduce various
bonus schemes in combination aiming for better
results, but as pointed out by Sprinkle (2003) little
is known as to how combination of incentives a?ects
worker productivity.
It must be noted that my aim is not to test the
optimality of the incentive plan, or make a direct test
of theoretical predictions concerning team-incen-
tives of agency models (Alchian & Demsetz, 1972;
Holmstrom, 1982), or behavioral theories Deutsch,
1949, 1973). Instead, my goals is to broaden our
understanding of howteam-based incentive compen-
sations works in practice, and more importantly,
documents how ?rms can design more e?ective
incentive plans for teams. In this regard, the article
resembles a case study similar in scope to other stud-
ies which had examined the performance e?ects of
incentive pay in ?eld-settings (Banker, Lee, & Potter,
1996b; Hansen, 1997; Knez & Simester, 2001).
After controlling for numerous factors that are
known to in?uence productivity and product qual-
ity in a multivariate regression model, the results
show that by moving from piece-rates to the new
incentives and by adopting the organizational
changes, the factory had a 31% rise in productivity
and a 95% drop in product defects. Moreover, the
factory experienced a substantial decrease in worker
absenteeism. Although it is di?cult to causally link
these improvements speci?cally to higher e?ort,
more cooperation, improved monitoring, and peer
pressure (all outcomes are unobservable latent
constructs), qualitative evidence gathered from
management and workers attributes these improve-
ments to positive changes in worker behavior asso-
ciated with the incentive plan.
This research contributes to the literature in sev-
eral important ways. First, I provide exploratory
evidence on the interplay between incentive pay
590 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
and management control initiatives within ?rms.
These results emphasize a need to jointly consider
combination of incentive pay schemes along with
imposition of control mechanisms to facilitate com-
munication and monitoring between workers as we
design compensation systems for teams. These ?nd-
ings are in line with recent observations in the con-
tracting literature. For example, Bonner and
Sprinkle (2002) and Ichniowski and Shaw (2003)
calls for greater integration between incentive pay
and elements in job design when designing compen-
sation systems to enhance the incentive e?ects of
monetary incentives. Likewise, these ?ndings pro-
vide further evidence about the importance of creat-
ing complementarities between incentive pay and
organizational design as a viable strategy to enhance
organizational performance, coinciding with the
?ndings of theoretical studies (Baker, Jensen, &
Murphy, 1988; Holmstrom & Milgrom, 1994; Mil-
grom & Roberts, 1995). The study may also help
to guide ?rms in the design and implementation of
better incentive plans for teams.
The remainder of the paper is organized as fol-
lows: the next section discusses the research setting
and compares the original and new incentive plan.
Hypotheses development reviews the literature and
develops hypotheses. Data and methodology
describes the sample and the empirical methodol-
ogy. Results are discussed in Empirical results.
The paper concludes with a discussion of the main
?ndings and o?er suggestions for future research.
Research setting
The research site is a large manufacturing facility
operated by a division of a Fortune 500 US Com-
pany. The factory assembles a large assortment of
security metal locks and it is dominant world pro-
ducer of security locks.
3
The site consists of three
production units to serve distinct markets. The
retail unit assembles locks for retail customers, the
institutional unit combination locks for schools,
the military, and sports facilities, and the commer-
cial unit customizes locks according to detailed
customer speci?cations. The plant operates in a
non-unionized environment and employs approxi-
mately 800 production workers.
Before presenting the structure of the original and
new incentive plan, it is useful to describe the nature
of the production process to better understand the
motives of using team-based compensation.
Process technology and production teams
Relevant to our discussion is the fact that the
plant employs production teams and/or production
cells to perform all assembly operations. The vari-
ous assembly tasks required in the fabrication of
locks are performed in their entirety by these teams,
which range in size from 8 to 15 members.
4
All three
units share more or less the same production tech-
nology. The assembly process is primarily labor-
intensive and each member of a team is responsible
for completing a speci?c task. In synthesis, these in-
clude the perforation, attachment, and insertion of
various metal and rubber components inside the
body of a metal padlock, assisted by small machin-
ery and specialized tooling. One important aspect of
the manufacturing process is that these all assembly
tasks are complementary and sequentially interde-
pendent. In this process, each team is responsible
for monitoring quality and in executing machinery
setups between production runs. Teams works un-
der the direction of a production supervisor who
is responsible for monitoring the team’s output
and quality. Hence, given the nature of the assembly
process in which all assembly steps are highly com-
plementary, the ?rm has always used some form of
incentive bonus linked to the team’s performance in
its compensation plan.
Original incentive plan
The original compensation plan consisted of
three elements: base pay, an individual attendance
bonus, and a piece-rate bonus linked to the team’s
output.
5
Base pay consisted of a daily salary, which
varied according to seniority and skill-level. To
reduce absenteeism, which is problematic in the
maquiladora industry, workers received a monthly
attendance bonus. This portion of pay was mainly
3
The facility is located in Northern Mexico and it is part of the
maquiladora industry. Under this type of outsourcing production
arrangement, ?rms import duty-free materials and components
from the US into Mexico for use in the assembly of its products.
After assembly, the ?nished products are re-exported to the US
or to the plant’s parent country of origin.
4
At the time of the study, there were approximately 50
production teams in operation.
5
For ease in exposition, the main elements of the previous and
new incentive plan are summarized here. Further details of the
speci?cs under each plan are provided in the appendix.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 591
?xed and accounted for approximately 70% of a
worker’s total earnings. The remaining portion
was contingent upon each team’s performance.
Each team was rewarded on the basis of piece-rates
and worker pay was proportional to the number of
locks produced daily by each team; resulting cash
bonuses were calculated on a weekly basis and
shared equally among members of a team.
6
A follow up question is why did this ?rm decide to
make changes to the compensation plan? Manage-
ment pointed out several limitations with the existing
plan. First, the consensus was that piece-rates were
only partially e?ective in motivating teams to raise
productivity consistently across all teams in the fac-
tory. Therefore, productivity for the plant was less
than optimal. Although some teams were highly pro-
ductive, in particular teams consisting of more senior
workers were more cohesive and motivated to work
harder to earn higher bonuses, the same level of pro-
ductivity was di?cult to achieve consistently for
many teams. To illustrate, I refer the reader to
Fig. 1, which graphs the factory’s monthly productiv-
ity. Clearly, it can be seen that productivity in the
months preceding the changes to the incentive plan
was low; it averaged around 60%.
The problem with low productivity was further
eroded as a result of high turnover and absenteeism
at the plant. E?ort levels were generally sub-optimal
for teams experiencing high turnover. The frequent
arrival of new workers tended to retard production,
resulting in lower output levels and smaller bonuses
for these teams. As a result, workers were not moti-
vated to work as hard to raise output. Additionally,
it was evident hat the attendance bonus was not suc-
ceeding in reducing absenteeism. Besides harming
productivity, absenteeism also made it di?cult for
teams to adhere to a group norm to monitor and
enforce good behavior among themselves (reduce
shirking, enhance cooperation, etc.). Hence, it was
imperative to ?nd an alternative solution to reduc-
ing chronic levels of absenteeism and turnover.
Another problem experienced with piece-rates is
that it contributed to quality problems. Because
the piece-rate bonus was tied to output, teams
mainly focused at raising output at the expense of
quality and were not as motivated to inspect, pre-
vent, or reduce product defects. Although, quality
was closely monitored and management refused to
pay a bonus for defective units, the number of prod-
uct defects was exceptionally high. To illustrate,
right before the ?rm adopted the new incentive plan,
product defects, measured as parts per million
(PPMs), averaged around 12,000 defects per month.
Lastly, the existing compensation plan appeared
to provide few incentives for teams to cooperate
with one another and share information to improve
the plant’s performance. Several modes of coopera-
tion were seen as critical. First, it was important to
motivate teams to communicate with one another so
they could exchange information regarding process
improvements. For instance, seemingly trivial inno-
vations such opening combination locks faster to
test their functionality or inserting components in
a di?erent way can have a discernible impact on
labor productivity. Management wanted teams to
share this knowledge with their peers; however,
piece-rates provided no incentive for teams to do
so.
7
It was also imperative to motivate teams across
Study Period
Team-level incentive Plant-level incentive
Original Incentive Plan New Incentive Plan
Piece-rates Team output-target scheme Team incentive, plant incentive, mgmt. control initiatives
January 1999
April 2000 July 2000
September 2003
Fig. 1. Implementation timeline of incentive plan.
6
For purposes of the bonus calculation, the piece-rate factor
was tied to the total standard hours of output produced during
the day. Standard hours of output equal the number of locks
produced in a given day times the standard labor time based on-
time and motion calculations required to assemble a given lock
model.
7
While the motives as to why teams did not share information
with one another could vary, a valid explanation is that
oftentimes workers are reluctant to do so for concerns that this
could lead to a rise in performance standards. For instance, a rise
in ?rm-wide productivity could potentially lead to the ratcheting
performance standards in such a way that it lowers their bonuses.
Further discussion of this argument can be found in Bandiera,
Barankay, and Rasul (2005) and Brown and Phillips (1986).
592 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
the three production units to share valuable knowl-
edge to improve the plant’s performance in general.
Teams at each production unit had developed
unique core competencies. For instance, production
teams in the commercial unit were highly skilled in
executing rapid set ups between production runs.
New incentive plan
To address all of the above concerns, the ?rm
restructured the existing compensation plan. Under
the new plan the ?rm wanted to motivate workers to
improve productivity and quality not only through
greater e?ort levels but also through heightened
cooperation at all levels: within each team, between
teams, and across production units. Restructuring
compensation around a single incentive scheme,
such as the existing piece-rate contract, presented
limitations since it neglected inter-team coopera-
tion. Hence, in changing the structure of the com-
pensation package the ?rm decided to adopt a two
tier bonus plan to reward individual team perfor-
mance as well as the combined performance of all
teams.
Under the new pay structure, each worker main-
tained the same level of base pay, but the attendance
bonus and piece-rates were phased out. The piece-
rate contract was replaced by a team-level bonus
and the attendance bonus by a plant-level bonus.
Before describing in detail each bonus scheme, it is
important to highlight several important details
under the new incentive plan. First, the new plan
contain larger variation in incentive pay. Workers
could earn more in total compensation, but because
a larger portion of their earnings now dependent on
the combined performance of all the teams, the level
of risk increased as well, and consequently, workers
could earn less pay. Second, several organizational
control initiatives were instituted to reinforce the
plan. Last, the implementation dates of the team
and plant bonus varied. The team bonus was intro-
duced in April 2000 and the plant bonus in July
2000.
8
Fig. 1 provides a chronology on these events.
I now turn to discuss in explicit detail the bonus
schemes and the organizational control initiatives.
Team-incentive bonus (output-target pay scheme)
In sharp contrast to piece-rates, which rewarded
teams for any positive level of output, the new
team-incentive bonus (hereafter referred as an out-
put-target pay scheme) provides a daily ?xed cash
bonus to each worker only when the respective team
reaches an output quota (a ?xed quantity of locks)
and a quality quota (a minimum number of product
defects). To illustrate, at the start of each workday,
a production supervisor informs each team the
number of locks that must be assembled and a min-
imum threshold of acceptable defects based on the
mix of locks in each production batch, as well as
the supervisor’s expectation of the team’s productiv-
ity. Failure to meet both targets automatically dis-
quali?es the team from earning the bonus. Instead
each worker receives his/her daily wage. Moreover,
after reaching the day’s output quota, teams can
earn an additional bonus for each lock produced
in excess of the set target. The bonus for this extra
output is paid using the same piece-rate factor
employed under the old plan. The ?xed and variable
portions of pay make the bonus scheme a combina-
tion of a budget-based and linear (piece-rate) con-
tract.
9
This sort of bonus scheme has been
referred to as a budget-linear scheme in the compen-
sation literature in accounting (see Fisher et al.,
2003).
Plant-incentive bonus (gain-sharing pay scheme)
The plant-incentive bonus rewards all teams with
a quarterly cash bonus for meeting plant-wide quar-
terly performance targets on productivity and qual-
ity. It is structured in such way that all teams must
coordinate e?orts to achieve the underlying perfor-
mance targets. The amount of bonus is levered so
that the percentage of bonus earned increases as e?-
ciency levels improve. To illustrate, at the com-
mencement of each quarter, workers are presented
with three performance targets with varying degree
of di?culty for each of the two criteria. The amount
of bonus depends on attaining any one of three set
targets for the quarter. For example, if performance
targets for productivity are set at 70%, 80% and 90%
(similar targets are set for minimum number of
product defects), and the lowest target is reached,
8
Management opted to implement the changes to the incentive
plan gradually so its workforce could be more receptive and
productivity would not su?er as much by giving them more time
to adapt to the new system. However, workers were informed
several months in advanced about the changes that were going to
occur to the compensation plan.
9
Management pointed out that the intention of maintaining a
portion of the piece-rates contract under this budget-scheme was
to stimulate teams (especially the most dynamic) to raise output
during periods of high demand without ratcheting output
standards which can potentially demoralize and discourage
workers.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 593
then each worker receives a cash bonus of 6% of the
accumulated quarterly earnings. By reaching the
middle and upper targets, workers are entitled to
an 8% and 10% bonus, respectively. No bonus is
paid when the plant fails to meet the lowest set tar-
gets. As noted previously, the plant bonus replaced
an individual attendance bonus; therefore, an addi-
tional provision under the plant bonus is that work-
ers must maintain near perfect attendance during
the quarter to be entitled to the bonus.
Organizational control initiatives to reinforce the new
incentive plan
Several organizational control initiatives were
instituted as part of the new incentive plan to facil-
itate the exchange of information, cooperation
between teams, and monitoring within and between
teams. First, to strength the team bonus, each team
now is provided with feedback of their performance.
Speci?cally, teams are given a report that depicts the
actual number of locks and the number of product
defects accumulated throughout the day. The
intended purpose is twofold. Teams can assess
how well they are doing relative to the standards.
If the team trails the output and quality quotas,
workers can either intensify e?ort levels, or help
one another to meet their quotas. So, the intended
purpose is to stimulate e?ort and intra-team
cooperation.
Moreover, this mechanism could reduce the
extent of free-riding by way of mutual monitoring.
To illustrate consider this example. Suppose that a
particular production team reaches its output and
quality targets on a regular basis, and neither target
is ratcheted-up in subsequent periods. If on a given
day, this team signi?cantly trails the output-target,
absent external or extraordinary circumstances in
the assembly process (e.g., material shortages, bot-
tlenecks, etc.), it would indicate that the team’s
e?ort levels are not as intense as in previous periods,
signaling shirking. Hence, reporting each team’s
performance on a timely basis induces workers to
monitor each other’s e?orts more rigorously, allow-
ing them in this way to minimize shirking.
The second organizational initiative consisted of
providing feedback on performance to the entire
factory. A weekly report is generated showing the
performance of each team on several key metrics:
productivity, product defects, material waste, the
number of production orders completed on-time,
absenteeism, and turnover. This information is dis-
seminated publicly to all teams. The intended pur-
pose is to help promote cooperation and
information sharing across the three units. For
example, low-performing teams could identify bet-
ter performers and seek help so their performance
can improve. Similarly, high performers can identify
low performers and volunteer to help them improve.
In addition, the distribution of performance reports
could facilitate monitoring among teams as a way to
control the free-rider problem by way of peer pres-
sure as discuss in more detail in the hypotheses
section.
A third and ?nal mechanism was instituted to
punish de?cient worker behavior, such as shirking
or uncooperativeness. Based on a team’s suggestion,
workers performing sub-optimally with the rest of
the team can be dismissed after repeated warnings
and eventually be laid-o?. Previously, shirking was
not disciplined. A potential e?ect of this mechanism
is that it could also help to instigate peer pressure
within each team.
Hypotheses development
This section develops hypotheses concerning the
sensitivity of worker productivity and product qual-
ity to the new incentive plan and describes how the
various elements on the plan could have led to
improvements in productivity and product quality,
as well as a reduction in worker absenteeism. An
important observation is that given the fact that
the new incentive plan comprised a combination
of incentive schemes which were introduced at dif-
ferent dates, plus the plan was implemented in con-
junction with the other initiatives of organizational
control, it makes it nearly impossible to make test-
able predictions on the e?ects on productivity and
quality attributed speci?cally to each element in
the incentive plan. Therefore, my predictions are
based on how the various elements in the plan could
have impacted productivity and quality as a whole
rather than individually. My predictions are guided
by studies in the compensation literature in account-
ing and labor economics, as well as in organiza-
tional behavior.
A starting point in our analysis is the switch from
piece-rates to the output-target scheme. Building on
prior research ?ndings, one might expect that work-
ers would be motivated by an underlying piece-rates
contract and work hard to earn large bonuses. For
example, numerous studies have documented that
piece-rates provide strong incentives for workers
to raise output and productivity (Lazear, 2000;
594 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Paarsch & Shearer, 2000; Shearer, 2004; Stiglitz,
1975). However, in team production, an output-tar-
get (or budget-based) scheme should be superior to
piece-rates in raising worker e?ort. The reasoning
behind this argument is that if every worker can
be punished with a low (penalty) wage when team
output or product quality falls below some target,
then su?cient incentives are generated for all work-
ers to work hard enough to hit those targets and to
monitor each other’s e?orts more closely to reduce
shirking, otherwise the entire team forfeits the
bonus. Simply put, the fact that a team could be
penalized for not meeting those targets becomes a
self-enforcing norm in a team. In contrast, with
piece-rates, workers know that they would be enti-
tled to receive some portion of the bonus as long
as the team generates output. Therefore, they may
not be as motivated to give their maximum e?ort
nor to police themselves as tightly, thus making
piece-rates more susceptible to the free-rider
problem.
10
Consistent with this view, Holmstrom (1982) pro-
posed the idea that an output-target (or budget-
based) scheme as a Nash equilibrium solution to
help mitigate the free-rider problem and to motivate
higher e?ort in his seminal paper on team-incentives.
He notes that the sharing of total output (most likely
a bonus) among team members will result in shirk-
ing. Incorporating the threat of a team-imposed pen-
alty for output below a target (the team shares less
than 100% bonus) would not yield an improvement
because team members would simply ignore the pen-
alty ex-post. Instead, incorporating a principal-
imposed penalty with a low (penalty) wage for out-
put below a target provides an incentive for workers
to work harder and to police themselves more strin-
gently. Empirical support for this argument has been
found by several studies which have documented an
advantage of output-target (or budget-based)
schemes over piece-rates in raising productivity in
teams (Fisher et al., 2003; Nalbantian & Schotter,
1997; Petersen, 1992).
Also, previous studies have documented that
piece-rates may discourage cooperation and often-
times may put workers (or teams) in competition
with one another (Brown & Phillips, 1986; Drago
& Garvey, 1998).
11
In contrast, other bonus schemes
such as gain-sharing bonuses should enhanced coop-
eration. Expanding on this point, the organizational
behavioral literature suggests that gain-sharing com-
pensation transforms a workforce, wherein workers
begin to interface in the whole organization and ulti-
mately leads to greater commitment and coopera-
tion to achieve performance goals Bullock and
Lawler, 1984; Schuster, 1983. ; Schuster, 1983). Sev-
eral ?eld studies have documented a rise in produc-
tivity as well as quality with gain-sharing through
heightened worker cooperation, information shar-
ing, and greater commitment from workers (Bullock
& Lawler, 1984; Schuster, 1984; Welbourne, Balkin,
& Gomez-Mejia, 1995). So, in this respect, the plant
‘gain-sharing’ bonus scheme should also lead to a
rise in productivity.
Lastly, recall that a primary di?erence between
the original and new incentive plan was larger vari-
ation in incentive pay. The new plan shifted a signif-
icant portion of a worker’s total earnings to the
combined performance of all teams in the factory.
In other words, workers could earn more in total
compensation, but their level of risk increased since
a larger amount of their earnings now were depen-
dent on the performance of all the teams, and there-
fore, could earn less pay as well. Based on the
conventional predictions of principal-agent theory
(Baiman, 1982; Holmstrom, 1979)
12
the notion that
individual incentive (risky) pay is associated with
positive e?ects on performance and/or worker pro-
ductivity is well documented in empirical studies
both in experimental and ?eld-settings (see Banker
et al., 1996b; Chow, 1983; Prendergast, 1999).
Hence, it is reasonable to assume that incentive
pay could as well drive strong incentive e?ects in
10
More formally, the fact that all workers bene?t from the e?ort
of fellow co-workers, team members may not be as motivated to
increase e?ort if increases in e?ort are re?ected by increase in pay
only on the order of 1/n (n = team size). All workers share in the
bene?t created from the extra e?ort of a particular worker.
Unless the marginal bene?t of an extra unit of e?ort exceeds the
marginal cost of working, the individual incentive to expend more
e?ort under team piece-rates is minimal.
11
As noted previously (refer to footnote 7) workers are often
reluctant to cooperate with each other when rewarded with piece-
rates for concerns that improvements in overall performance
could lead to a rise in performance standards. For instance, a rise
in ?rm-wide productivity could potentially lead to the ratcheting
performance standards in such a way that it lowers their bonuses
(Bandiera et al., 2005; Brown & Phillips, 1986).
12
Within the conventional framework of principal-agent theory,
the idea is that incentive (contingent) pay imposes more risk on
the workers. As a result, there should be more motivated to work
harder. This is also referred as risk-sharing in the contracting
literature. This issue has long been recognized in the contracting
theory in economics (Holmstrom, 1979) and accounting (Baiman,
1982). For further insight about the link between incentive pay
and economic performance, see the work of Prendergast (1999).
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 595
team production settings just as it has been demon-
strated in the case of individual incentives. Because
the possibility of earning lower pay is now at stake,
I hypothesized that the overall shift in incentive pay
under the new incentive plan should motivate work-
ers to work harder to reach the various performance
targets and these e?orts should lead to higher pro-
ductivity. Based on all the above arguments, I o?er
the following prediction:
H1. Ceteris paribus, productivity should rise after
the implementation of the new incentive plan.
Now, turning over to product quality, my predic-
tion is that the switch in incentives should also lead
to a rise in product quality. First, there is plenty of
empirical evidence documenting that the use of
piece-rates is associated with output of lower qual-
ity. Given the high emphasis that piece-rates place
on production, prior research found that workers
trade-o? output for quality under piece-rates
(Baker, 1992; Holmstrom & Milgrom, 1994; Lazear,
1986, 1991, 2000; Paarsch & Shearer, 2000; Stiglitz,
1975).
13
According to Lazear (2000, p. 1358) ‘‘one
defect of paying piece-rates is that quality may suf-
fer”. Expanding on this same point, Paarsch and
Shearer (2000, p. 61) note that, ‘‘piece-rates have
the advantage of providing incentives for workers
to work hard, but piece-rates also reduce the incen-
tive to the worker of producing quality output”.
Although the previous piece-rate contract
imposed a penalty for defective output, the penalty
was not as severe as the one imposed under the out-
put-target scheme. With piece-rates, teams were
penalized solely on the defective units found in a
particular production batch and could still receive
a bonus on the remaining ‘good’ output produced
during the day. As a result, the teams most likely
ignore the penalty ex-post and continue to overlook
quality. In contrast, under the output-target
scheme, the penalty for defects is much severe. If a
team does not reach the minimum threshold of
defects, the team loses the entire bonus regardless
of the quantity of output produced during the
day. So, the penalty function makes it more costly
for workers to miss the quality target. As a result,
teams should place more emphasis in reducing mis-
takes during assembly and also they should inspect
the quality of their output more carefully.
Furthermore, by invoking the same arguments
discussed above in the case of productivity, I also
predict that the adoption of the gain-sharing bonus
scheme and the shift in incentive pay should also
lead to a rise in product quality. Thus, based on
all the above arguments, I hypothesized that:
H2. Ceteris paribus, product quality should rise
after the implementation of the new incentive plan.
I now turn to discuss how the various organiza-
tional control changes complement the incentives.
Ichniowski and Shaw (2003) argue that incentive
pay will be more e?ective when ?rms also provide
other organizational initiatives (i.e. human resource
practices). In the case of team production, under-
taking cooperation and enticing workers to share
valuable information within and between teams
involves other mechanisms that would help facili-
tate communication and the exchange of informa-
tion between workers besides incentive pay. This
argument is consistent with the observation of Ran-
kin (2004), who asserts that cooperation in team
production settings requires a rich informational
environment. Also, Bandiera, Barankay, and Rasul
(2004) claim that one chief determinant for cooper-
ation is the ability of workers to monitor each
other’s performance. Therefore, if the production
setting lacks the infrastructure that enables workers
to communicate and interact with one another, and
to e?ectively monitor each other, cooperation
would be di?cult to achieve.
Additionally, a central observation of many theo-
retical studies is that worker monitoring can act as a
powerful mechanism to control the free-rider prob-
lem. But, the ability to monitor co-workers may be
greatly in?uenced by elements in job design and
workplace characteristics. As emphasize by Kandel
and Lazear (1992, p. 806), if workers do not have
the means to monitor each other e?ciently, or in
the absence of appropriate mechanisms to enforce
‘good behavior’, worker monitoring may fail in con-
trolling the free-rider problem.
As noted previously, the incentive plan was intro-
duced in combination with two initiatives designed
to facilitate intra- and inter-teams monitoring. So,
I expect such initiative to facilitate cooperation.
The reasoning behind this argument is that perfor-
mance feedback acts as a substitute for direct obser-
13
The idea that piece-rates lead to poor quality is not entirely
new; the problem lies when only some attributes of output can be
measured accurately, workers’ attention will be diverted to the
attributes that are rewarded in the compensation formula, such as
output. Additionally, as noted by Lazear (1991), much of this
problem has to do with full observability of output. For further
discussion as to how piece-rates could lead to lesser quality, see
Lazear (1986, 1991) and Stiglitz (1975).
596 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
vation and permits workers and teams to monitor
their progress and assess their own performance rel-
ative to their peers. With this information, teams
can seamlessly identify other teams that are per-
forming sub-optimally relative to the performance
standards set for the plant and o?er their help to
help them improve.
Furthermore, besides creating a richer informa-
tional environment, performance feedback should
also facilitate the rate of learning and strategy devel-
opment across production cells (Boning, Ichniow-
ski, & Shaw, 2007; Sprinkle, 2000). As noted by
Boning et al. (2007) as they examine the e?ects of
problem solving teams, group pay and productivity
in the steel industry, ‘‘problem solving e?ort refers
to developing and implementing ideas for improv-
ing production operations, and is based on the idea
that operators with hands-on experience on a pro-
duction line are in a unique position to devise rem-
edies for idiosyncratic problems. . .”. So, I expect
that the dissemination of performance information
should motivate more senior operators to transfer
valuable knowledge with co-workers. If this knowl-
edge is shared, productivity and quality improve-
ments should continue gradually after the
implementation of the plant bonus scheme.
Also, the body of the existing literature on incen-
tives for teams provides a general framework for
ways to minimize the free-rider problem in team
production, emphasizing that peer pressure can be
an e?ective control mechanism (Barron & Gjerde,
1997; Che & Yoo, 2002; Kandel & Lazear, 1992;
Macho-Stadler & Perez-Castrillo, 1993; Sewell,
1998). Thus, I posit that the public dissemination
of performance should also aid in controlling the
free-rider problem via peer pressure. For example,
with access to teams’ performance, teams can
appraise whether their peers are working with simi-
lar intensity to achieve the performance targets in
productivity and quality for the plant. Top perform-
ers could then exercise external pressure (e.g., public
shame, scrutiny) on slackers to raise their perfor-
mance to a level commensurate with the rest of
the plant.
Lastly, the old incentive plan lacked mechanisms
to properly identify or eliminate de?cient worker
behavior; so workers had little incentive to apply
pressure on co-workers. In contrast, the new incen-
tive system instituted a formal mechanism to punish
shirking and uncooperative behavior—the possibil-
ity of being terminated from the job. As a result, I
posit that the threat of dismissal is a powerful solu-
tion that may be levied against shirking and unco-
operativeness through peer pressure as well.
To summarize the above arguments, the new
incentive plan provided mechanisms to help facili-
tate cooperation, mutual monitoring, instigate peer
pressure, and promote the rate of learning. It also
provided workers with the means to punish de?cient
behavior (i.e., shirking, uncooperativeness). There-
fore, I hypothesized that these outcomes should
motivate teams to continue making gradual
improvements to productivity and product quality.
Hence, the following hypothesis is presented.
H3. Ceteris paribus, productivity and quality
improvements should continue and persist over
time following the implementation of the new
incentive plan.
My last set of predictions concerns absenteeism
and turnover. Recall that the plant bonus scheme
made it mandatory for workers to have near perfect
attendance to be entitled to receive a bonus. It may
be argued that in some sense the attendance bonus
and the provision under the plant bonus are substi-
tutes. Workers are penalized under either incentive
contract: missing a day of work precludes workers
from earning the bonus in either case. However,
the di?erence between the new incentive plan is that
the absence penalty function is steeper (more severe)
given the fact workers could lose a bigger paycheck
in the form of a larger quarterly bonus. Consistent
with this view, prior research has shown that atten-
dance bonus plans are more e?ective in curtailing
absenteeism when steeper penalties are imposed on
workers (Allen, 1981; Brown, Fakhfakh, & Ses-
sions, 1999). Based on this argument, I make the
following prediction:
H4. Ceteris paribus, the implementation of the new
incentive plan leads to a reduction in absenteeism.
Lastly, it is possible that the new incentive plan
could help in sorting out the workforce, and there-
fore, create a potential selection e?ect. That is, it
could have produced a selection e?ect in such a way
that the less productive workers leave the ?rm dislik-
ing the changes to the compensation plan and the
more productive remain or are attracted to the ?rm
as documented by several studies in the compensa-
tion literature (Banker, Lee, Potter, & Srinivasan,
2000; Chow, 1983; Waller & Chow, 1985). To inves-
tigate this possibility, I examine worker turnover be-
fore and after the incentive plan was adopted. No
speci?c prediction is made as to how the incentive
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 597
plan a?ects worker turnover. However, if the rate of
voluntary turnover is statistically higher in the early
months after adoption of the new incentive plan and
levels-o? in subsequent months relative to the old
incentive plan, this could possibly be an indication
of a selection e?ect in the workforce.
Data and methodology
Data and sample
The data were drawn from the three production
units at the factory.
14
The initial sample covers the
period of January 1999 through September 2003.
However, as it is characteristic of most ?eld studies
there were data limitations. Mainly, product defects
and several of the control variables were not fully
available for the entire period, thereby forming an
unbalanced panel of data. Uninterrupted time-series
data of all variables of interest is available from
September 1999 through September 2003. There-
fore, the study period covers this period. Also, be-
cause most data is reported on a monthly basis,
my main regressions are estimated using monthly
observations. In the case of productivity, an exten-
sion of the tests is supplemented with weekly data
in a reduced regression model. The ?nal sample con-
sists of a longitudinal cross sectional sample of 147
monthly observations; there are 49 observations for
each of the three production units.
Regression models and measurement of variables
To provide a thorough set of controls in my regres-
sions, I follow Hayes and Clark (1985) and include a
broad set of factors, which tend to a?ect labor pro-
ductivity in manufacturing operations. These include
production headcount, absenteeism, turnover,
worker training, overtime, engineering modi?cations
to the assembly process, plus other variables speci?-
cally related to the research site. Process engineers
were consulted to ensure that these variables ade-
quately addressed the corresponding production
technology. Because some of these factors undoubt-
edly a?ect other dimensions of manufacturing perfor-
mance, including product quality (see Banker, Field,
Schroeder, & Sinha, 1996a), several of these variables
are also used as controls on the regressions on prod-
uct quality. Further, to eliminate potential omitted
variables biases, I investigated whether any major
changes or relevant events occurred at the plant dur-
ing the study period. By the time the changes to the
compensation plan took e?ect the facility was fully
stable and mature and no other major events (e.g.,
management changes, adoption of new technology,
or equipment changes) other than the changes to
the incentive plan were introduced.
To assess the impact of the new incentive plan on
productivity, I estimate the following OLS regres-
sion model pooling observations from the three pro-
duction units:
PRODUCTIVITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
DEFECTS
it
þ b
4
HEADCOUNT
it
þ b
5
TURNOVER
it
þ b
6
ABSENTEEISM
it
þ b
7
OVERTIME
it
þ b
8
LAG TRAINING
itÀ1
þ b
9
ECO
it
þ b
10
RETAIL
it
þ b
11
INSTITUTIONAL
it
þ e
it
ðM1Þ
And, the following model is estimated on product
defects and material waste to measure the impact on
product quality:
QUALITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ b
8
RETAIL
it
þ b
9
INSTITUTIONAL
it
þ t
it
ðM2Þ
where subscript i refers to production unit i and t to
each month observation. An explanation of each
variable immediately follows. A similar but a more
parsimonious model with fewer control variables is
used to test the incentive plan e?ects on absenteeism
and turnover.
14
Data to calculate proxies for productivity, product quality,
and remaining variables of interest were hand-collected from
various departmental records, including accounting, production,
process-engineering, quality, and payroll. Additionally, I
obtained the opinions of management and production workers
regarding the impact of the incentive plan on plant performance
and held various discussions with process engineers to understand
the manufacturing process and to investigate other relevant
events that could have potentially confounded the impact of the
incentive plan on performance.
598 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Dependent variables
For consistency and to reduce potential biases in
the analysis, I use the same productivity and prod-
uct quality measures used by the plant as my main
proxies for productivity and quality. PRODUC-
TIVITY is de?ned as the ratio of standard labor
hours based on the actual output to the total
amount of labor hours spent in production. This
is equivalent to labor e?ciency variance expressed
in ratio form (Horngren, Datar, and Foster, 2002).
The metric excludes production hours unrelated to
the normal course of production, such as loss time
due to machinery breakdowns, material shortages,
and scheduling problems; these hours are excluded
from the denominator. Additionally, the ratio
excludes standard labor hours related to defective
units in the numerator; that is, labor hours worked
and earned but unaccounted for when units are
rejected for not complying with standards of
quality.
The proxy for product quality is the number of
product defects (DEFECTS) expressed as parts per
million (PPMs).
15
To normalize the data, this vari-
able is transformed by taking the natural log. The
second proxy for quality is labeled as WASTE and
it represents the dollar amount of material waste
de?ated by production volume. Just as in the case
of productivity, these are the chief measures used
by the factory to track product quality.
An important issue in the estimation of the
empirical model is whether the dependent variables
should be estimated as a level variable, or as a
change and/or percentage change variable. This
issue merits further discussion. As noted by Banker
et al. (2000, p. 82), accounting research has relied on
both, level and return or change models, to examine
the association between variables of interest and
?rm performance. The decision to use one measure
relative to the other depends on the research ques-
tion in place. In the context of compensation
research, the use of a level variable can be supported
if the treatment variable (i.e., pay) has a permanent
impact on performance measures. If the impact is
expected to be gradual over time, such as in the case
of this research site given that the changes to the
incentive plan occurred in stages, then the percent-
age change or di?erenced models could be more
suitable. Therefore, I present results based on three
speci?cations of each model: one using level vari-
ables, level changes, and percentage changes in pro-
ductivity, defects, and the remaining treatment
variables.
Main independent variables
Following a similar research approach to Banker
et al. (1996a), I regress each performance outcome
on a time trend to capture the mean changes in pro-
ductivity and product defects under the periods of
the old and new incentive plan. More precisely,
the variable OLDPLAN_TREND represent a linear
time trend for the period under the old incentive
plan (September 1999 to March 2000); it is mea-
sured as the number of months since the beginning
of the study period, September 1999, through
March 2000, and assumes a value of zero thereafter.
Likewise, NEWPLAN_TREND represent a linear
time trend for the period in which all elements in
the new incentive plan, the team bonus, the plant
bonuses, and the organizational control initiatives,
are operating altogether (July 2000 to September
2003); it is measured as the number of months since
July 2000 through September 2003, and assumes a
value of zero for the preceding months prior to the
introduction of the teambonus. Apositive and larger
coe?cient estimate on NEWPLAN_TRENDrelative
to the OLDPLAN_TREND coe?cient will indicate
that productivity is greater for the period of the new
incentive plan. Conversely, in the regressions on
product defects and material waste, a negative and
smaller estimate on NEWPLAN_TREND relative
to the OLDPLAN_TREND coe?cient would indi-
cate that product defects and waste is lower under
the new incentive plan.
It must be noted that the variable NEW-
PLAN_TREND tests for the hypothesized e?ects
of the incentive plan by measuring the performance
impact of the output-target and gain-sharing bonus
scheme, and the organizational initiatives, jointly.
The close window of separation between the
schemes’ implementation period makes it di?cult
to separate cleanly the performance impact associ-
ated across each incentive. For instance, potential
improvements to productivity and quality associ-
ated with the output-target scheme are re?ected in
future periods and such improvements may be
picked up by the gain-sharing scheme. However,
combining both incentives schemes and the organi-
zational control initiatives re?ects management’s
intention of treating all them as part of the same
incentive plan.
15
To arrive at this number, production batches are inspected at
random and defective units within a batch are rejected when the
batch fails to meet standards of quality.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 599
Control variables
The variables HEADCOUNT (the total number
of production workers), TURNOVER (the rate of
employee turnover) and ABSENTEEISM (the rate
of absenteeism) are used in both regressions to con-
trol for e?ects related to changes in the workforce.
HEADCOUNT controls for ?uctuations in produc-
tion headcount. While an increasing (decreasing)
trend in production headcount might yield an
increase (decrease) in labor productivity, no predic-
tion is made ex ante about the expected sign of this
coe?cient since a larger number of production
workers might yield an increase in total standard
hours (numerator) but also increase total produc-
tion hours (denominator) in the productivity ratio.
Both ABSENTEEISM and TURNOVER should
impact productivity and product quality adversely
due to a reduction in manpower. Therefore, I expect
a negative (positive) relationship between TURN-
OVER and labor productivity (defects) and the
same relationship with ABSENTEEISM.
The variable LAG_TRAINING is a 1 month lag
variable representing the total number of hours of
training received by production workers. The vari-
able includes both hours related to inductive train-
ing for new employees, as well as training hours
directed toward teaching current employees tech-
niques to make improvements to the manufacturing
process. The impact of training on productivity is
ambiguous. While one might expect employee train-
ing to impact positively labor productivity given
than it enables workers to acquire new knowledge
to improve their performance, the counter argument
is that hours spend in training is time diverted from
production. Also, a signi?cant amount of training
hours re?ect new employee training that may not
add incremental value. On the other hand, though
an immediate positive impact may not be realized,
these short-term losses in labor productivity may
be o?-set by future gains that are driven by new
knowledge at improving the manufacturing process.
Hence, given both arguments, I do not predict a pri-
ori the direction of such relationship.
OVERTIME represents total hours of overtime
and it controls for the e?ect of overtime in the pro-
ductivity model. ECO (the number of engineering
change orders) is incorporated on both regression
models to control for modi?cations in either prod-
uct design, material components, or in the assembly
process. Because these changes tend to improve
product functionality, quality, or reduce complexity
during assembly, the related coe?cient estimate
should be positively (negatively) associated with
productivity (defects). Also, many of the engineer-
ing changes originate on the production ?oor based
on a suggestion from a production worker. Thus, I
expect a signi?cant rise in the number of engineering
changes in the post-incentive plan implementation
period.
Because it is possible that the extent of product
quality could have a direct impact on labor produc-
tivity, I also included my main proxy for quality,
DEFECTS, in the estimation on productivity. The
rationale is that if fewer mistakes are made during
the assembly process, this translates into more pro-
ductive time for workers, which ultimately lead to
more output. Conversely, poor quality could have
an adverse impact on productivity due to the fact
that product defects tend to slow down a manufac-
turing process. For example, workers may divert
production time to reworking or inspecting units
(Cachon & Terwiesch, 2006).
16
Therefore, I predict
a negative relationship between defects and
productivity.
Based on the recommendations of quality engi-
neers, I also included an additional variable in the
regressions on defects and material waste to control
for the level of quality of material components used
in production. The variable DEFECTS_INCOM-
ING represents the amount of defective material
components detected during a random sample qual-
ity inspection as these materials arrive at the plant.
Since defective components are identi?ed prior to
assembly, in part this should help prevent product
defects and material spoilage. Hence, the estimated
coe?cient should be negative on both, the product
defects and material waste regressions.
Lastly, although the three production units share
similar production technology, due to the large mix
of products, the degree of assembly complexity
could vary among the units. Accordingly, to control
for di?erences in the assembly process that could
potentially impact productivity and quality, the
variables RETAIL, INSTITUTIONAL, and COM-
MERCIAL, represent dummy variables to control
for ?xed e?ects across the three production units.
17
16
According to the operations management literature, poor
quality is inversely related to how fast subassemblies ?ow in a
manufacturing process. Therefore, more defects, reworking units,
and waste generally slows down a process, and thus impacts
adversely labor productivity.
17
The dummy variable on the commercial unit (COMMER-
CIAL) is captured in the intercept of each regression model.
600 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Empirical results
Table 1, panel A, summarizes descriptive statis-
tics for the sample, showing means, medians, and
standard deviations for key variables of interest.
The mean (median) monthly productivity across
production units is 82% (88%), while the mean
(median) product defects is 1980 (279). A closer
look at the data indicates large variation for both
variables, with a standard deviation of 17% and
4302 for productivity and defects, respectively.
Mean and (median) for absenteeism and turnover
are 1.3% (0.90%) and 3.8% (2.1%), respectively.
Panel B reports tests of mean di?erences (Wilcoxon
Table 1
Descriptive statistics of variables of interest monthly observations for the period of September 1999 through September 2003
Variable Mean Standard deviation Minimum Median Maximum
Panel A: pooled sample (N = 147)
Labor productivity (%) 82.4 16.5 29.8 88.1 106.2
Product defects 1,980 4,302 0 279 20,693
Headcount 193 44 61 199 288
Turnover (%) 3.8 4.2 0 2.1 16.2
Absenteeism (%) 1.3 0.95 0 0.90 4.3
Overtime hours 2,022 3,167 0 857 20,191
Training hours 340 281 35 226 1103
Production volume (units) 989,197 432,672 127,679 932,414 2,148,570
Engineering change orders 2 2 0 1 10
Material defects incoming 4224 6132 0 2300 37,607
Material waste ($) 6348 3666 1070 5610 21,038
Standard labor hours 29,265 9462 2275 30,210 47,426
Production labor hours 35,342 9607 7624 35,894 61,481
Total labor cost ($) 170,295 115,606 33,119 127,631 516,072
Variable Median before Median after Di?erence
Panel B: di?erences in medians before and after changes to the incentive plan
Labor productivity (%) 59.3 89.9 30.6
***
Product defects 9244 186 À9058
***
Headcount 198 201 3
Turnover (%) 8.2 1.9 À6.3%
***
Absenteeism (%) 2.0 0.7 À1.3%
***
Overtime hours 2568 786 À1782
***
Training hours 107 248 141
**
Production volume (units) 953,186 988,324 35,138
Engineering change orders 1 2 1
Material defects incoming 6598 2035 À4563
***
Material waste ($) 3742 5583 1841
**
Standard labor hours 20,912 31,063 10,151
***
Production labor hours 31,207 36,129 4922
Total labor cost ($) 126,631 145,898 19,567
***
*,**,***
denotes statistical signi?cance at the 0.10%, 0.05%, and.01% levels, respectively. Descriptive statistics calculated using monthly
observations for each manufacturing unit for the period of September 1999 through September 2003. There are 49 monthly observations
for each production unit. Test of di?erences in median for the before and after period is performed using Wilcoxon rank sum test with
monthly observations for each production unit for the period of September 1999 to September 2003. The before period contains monthly
observations for the period of September 1999 through March 2000. The after period contains monthly observations from July 2000
through September 2003. Therefore, the test is performed eliminating the observations from April 2000 through June 2000, the period in
which the team-level (output-target) incentive scheme operates in isolation. Variables de?nitions: Labor productivity: total standard hours
divided by total production labor hours; Product defects: total sum of ?nished product defects (parts per million); Headcount: total number
of production workers; Turnover: total number of voluntary/involuntary resignations divided by the average production headcount;
absenteeism: total number of absences divided by the total number of worked days; overtime hours: total number of production overtime
hours training hours: total number of hours of worker training; volume: total units of output; engineering change orders: total number of
manufacturing process changes; material defects incoming: total number of material component defects during inspection upon arrival at
the factory; material waste: total amount of material waste in US dollars; standard labor hours: standard labor time per each unit of output
multiply by the number of units manufactured; production labor hours: total number of hours spent in production; total labor cost: total
direct labor cost, including salary and related cash bonuses.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 601
Signed-Rank test) on each variable for the periods
before and after the changes to the incentive plan
are introduced. Signi?cant changes are observed in
all key performance metrics: a mean rise of 30.6%
(p-value = .001) is observed in labor productivity,
and a mean decline of 9058 (p-value = .001) in prod-
uct defects. Similarly, both turnover and absentee-
ism exhibit a statistically signi?cant decline of
6.3% and 1.3%, respectively. These ?ndings provide
preliminary support of improvements in these out-
comes after the new incentive plan is introduced.
Table 2 reports Pearson correlations coe?cients
between the regression variables. An interesting pat-
tern is revealed in the strong negative correlation
(À.58) between productivity and product defects,
corroborating the notion that quality impacts pro-
ductivity. Productivity is also strongly negatively
correlated, as predicted, with turnover (À.71) and
absenteeism (À.77). Likewise, there is strong corre-
lation between product defects and turnover (.33),
absenteeism (.50), and defects incoming (À.70). A
high degree of correlation is also observed among
several of the control variables, which may indicate
a problem with multicollinearity. Therefore, I check
for the presence of it using variance in?ation factors
and condition index analyses and found no signi?-
cant threat of multicollinearity.
Regression results on productivity and quality
Hypotheses one and three predict that productiv-
ity should rise and improve over time, respectively,
after the implementation of the new incentive plan.
Table 3 presents regressions results on productivity
for the pooled sample (panel A) and by production
unit (panel B). The parameter of interests through-
out is b
1
and b
2
, which represents the respective time
trend for the old and new incentive plan, respec-
tively. Each coe?cient captures the change in pro-
ductivity over time. A comparison of these two
parameters will tell us whether the new incentive
plan is associated with any changes in productivity
across time. As shown, the coe?cient estimate
NEWPLAN_TREND is positive and statistically
signi?cant at the 1% level in the regression on level
productivity (b
2
= 0.003), change in productivity
(b
2
= 0.004), and percentage change in productivity
(b
2
= 0.006), whereas the OLDPLAN_TREND
coe?cient is negative and statistically signi?cant
on the level productivity model (b
1
= À0.119) and
insigni?cant in the change and percentage change
models. An F-test of di?erence among the two coef-
?cients show that the di?erence is signi?cant at the
1% level in the level model, and at the 5% level in
the percentage change model.
The explanatory power across the three models is
high, with adjusted R
2
of 70% in the case of the per-
centage change model and 85% in the level model.
Note further that the several of the factors included
as controls have the predicted sign and are signi?cant
at the 5% level or better. In particular, the coe?cient
estimates for TURNOVERand ABSENTEEISMare
both negative and statistically signi?cant. These ?nd-
ings suggest that absenteeism and turnover have an
adverse impact on productivity. Another interesting
result is that the coe?cient for DEFECTS is negative
and signi?cant (b
3
= À0.191, p-value < .05), corrob-
orating the fact that product defects have a negative
impact on productivity.
Results on productivity by production unit are
reported in panel B. Though the number of monthly
observations (sample size) dropped substantially,
the results are remarkably similar with those in the
pooled sample. The NEWPLAN_TREND coe?-
cient is positive and statistically signi?cant at the
conventional level for two of the three production
units. In contrast, the OLDPLAN_TREND coe?-
cient is either negative and signi?cant or statistically
insigni?cant. Likewise, b
2
is statistically and signi?-
cantly greater than b
1
at both models for the institu-
tional and retail units.
To ensure robustness of these results, I also esti-
mated a similar but a more parsimonious model
with fewer controls using weekly data for the same
time period. These results are not reported in tables;
however, they are qualitatively similar to the ones
reported on Table 3.
18
Taken all together, these ?ndings show that the
adoption of the new incentive plan led to a rise in
productivity and to gradual improvements in pro-
18
These results are not sensitive to changing the date of the time
trend variable for the new incentive plan, NEWPLAN_TREND.
Even when we allow the e?ect of the date to vary from July 1,
2000 back to April 1, 2000 the date of adoption of the output-
target scheme, the results are unchanged. Also, in a regression
model not reported in tables, I estimated this regression with
three time trend variables to represent the periods under piece-
rates (September 1999 to March 2000), the output-target scheme
(April 2000 to June 2000), and the period when all elements in the
incentive plan operate altogether (July 2000 and thereafter). This
speci?cation gives somewhat qualitatively similar results. How-
ever, because there are not su?cient time-series observations
surrounding the output-target scheme, the power of the statistics
tests are limited for the time trend coe?cient on the output-target
scheme.
602 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 2
Pearson correlation coe?cients of regression variables pooled sample (N = 147)
PRODUCTIVITY DEFECTS HEADCOUNT TURNOVER ABSENTEEISM OVERTIME TRAINING VOLUME ECO INCOMING WASTE
PRODUCTIVITY 1.00
DEFECTS À0.58 1.00
HEADCOUNT 0.21 À0.03 1.00
TURNOVER À0.71 0.33 0.09 1.00
ABSENTEEISM À0.77 0.50 À0.03 0.68 1.00
OVERTIME À0.54 0.26 0.29 0.54 0.56 1.00
TRAINING 0.22 À0.22 0.38 À0.21 À0.21 À0.04 1.00
VOLUME 0.35 À0.13 0.54 0.04 À0.19 0.02 À0.21 1.00
ECO 0.15 À0.22 0.09 À0.11 À0.02 À0.05 0.17 0.09 1.00
INCOMING À0.70 0.26 À0.25 0.59 0.52 0.35 À0.25 À0.11 0.13 1.00
WASTE À0.05 À0.16 0.30 0.15 0.15 0.14 À0.01 0.36 0.29 0.12 1.00
Pearson correlations are estimated using monthly observations from the three production units for the period of September 1999 through September 2003.
Variables de?nitions:
PRODUCTIVITY (labor productivity) = total sum of standard hours of output divided by total production labor hours;
DEFECTS (product defects) = total sum of ?nished product defects (parts per million);
HEADCOUNT (production headcount) = total number of production workers;
TURNOVER (turnover rate) = total number of voluntary/involuntary resignations divided by the average production headcount;
ABSENTEEISM (absenteeism rate) = total number of absences divided by the total number of worked days;
OVERTIME (overtime hours) = total number of production overtime hours;
TRAINING (employee training) = total number of hours of skilled-training;
VOLUME (production volume) = total units of output;
ECO (engineering change orders) = total number of engineering change orders for manufacturing process changes;
INCOMING (material defects incoming) = total number of material component defects during inspection upon arrival at the factory;
WASTE (material waste) = total amount of material waste in US dollars.
F
.
J
.
R
o
m
a
´
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
4
(
2
0
0
9
)
5
8
9
–
6
1
8
6
0
3
Table 3
Time-series regressions on labor productivity
Variable Productivity level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 0.854
***
0.258
***
0.439
***
(14.3) (4.42) (4.25)
OLDPLAN_TREND (b
1
) À0.119
***
À0.002 À0.046
**
(À4.19) (À1.43) (À1.96)
NEWPLAN_TREND (b
2
) 0.003
***
0.004
***
0.006
***
(3.06) (4.31) (4.06)
DEFECTS (b
3
) À0.191
**
À0.014
*
À0.026
(À2.14) (À1.74) (À1.56)
HEADCOUNT (b
4
) 0.0003
*
0.00 0.00
(1.94) (0.24) (0.24)
TURNOVER (b
5
) À0.983
***
À0.650
***
À0.931
**
(À4.12) (À2.82) (À2.32)
ABSENTEEISM (b
6
) À2.94
**
À2.19
**
À3.92
**
(À2.51) (À2.00) (À2.17)
OVERTIME (b
7
) À0.002
***
À0.001
**
À0.001
**
(À4.24) (À2.48) (À1.99)
LAG_TRAINING (b
8
) 0.0003 0.001 À0.004
(0.30) (0.18) (À0.30)
ECO (b
9
) 0.005
*
0.003 0.006
(1.81) (1.52) (1.56)
RETAIL (b
10
) 0.045
*
À0.117
***
À0.267
***
(1.86) (À4.36) (À5.65)
INSTITUTIONAL (b
11
) À0.18 À0.011 À0.015
(À0.75) (À0.43) (À0.22)
Adjusted R
2
0.85 0.72 0.70
Durbin–Watson statistic 1.88 1.99 2.02
N 147 147 147
Test of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.26 –
H
1
: b
1
– b
2
18.3
***
– 5.70
**
Variable Commercial unit Institutional unit Retail unit
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
Panel B: regressions by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.791
***
0.222
**
0.373
**
0.731
***
0.17 0.559
***
0.946
***
0248
***
0.322
***
(6.35) (2.08) (2.09) (6.87) (1.67) (3.78) (15.02) (3.50) (3.52)
OLDPLAN_TREND
(b
1
)
À0.004 0.006 0.013 À0.166
**
0.089 0.113 À0.055
**
À0.033
**
À0.045
**
(À0.65) (0.94) (1.13) (À2.34) (1.18) (0.48) (À4.53) (À2.49) (À2.58)
NEWPLAN_TREND
(b
2
)
0.013
***
0.012
***
0.021
***
0.006
**
0.005
**
0.007
**
0.003
**
0.004
**
0.005
**
(4.22) (4.56) (4.54) (2.39) (2.43) (2.06) (2.02) (2.41) (2.45)
DEFECTS (b
3
) 0.007 0.007 0.012 À0.016 À0.159 À0.033 À0.022
**
À0.014 À0.018
(0.58) (0.60) (0.57) (À1.43) (À1.40) (À1.45) (À2.18) (À1.36) (À1.33)
HEADCOUNT (b
4
) 0.001
**
0.001
***
0.002
***
0.006
*
0.001
*
0.0003 À0.00 À0.001
*
À0.001
*
(3.31) (3.35) (3.29) (1.79) (1.70) (0.64) (À0.66) (À1.88) (1.93)
TURNOVER (b
5
) À0.323 À0.349 À0.596 À0.599 À0.601 À0.323 0.221 0.520 0.720
(À0.90) (À0.99) (À1.00) (À1.46) (À1.44) (À0.43) (0.44) (1.04) (1.11)
ABSENTEEISM (b
6
) 0.256 0.218 0.241 À0.678 À0.598 À0.233 À4.67
**
À3.80 À5.02
(0.20) (0.17) (0.87) (À0.50) (À0.43) (À0.08) (À2.19) (À1.52) (À1.54)
OVERTIME (b
7
) 0.00 0.00 0.001 À0.001
*
À0.001 À0.004
***
À0.003
**
0.002
*
0.003
*
(0.99) (0.95) (0.87) (À1.71) (À1.17) (À2.90) (2.52) (1.77) (1.81)
LAG_TRAINING
(b
8
)
0.010 0.102 0.017 À0.058
**
À0.590
**
À0.165
***
0.012 À0.004 À0.006
(1.21) (1.27) (1.28) (À2.31) (À2.32) (À3.20) (0.55) (À0.21) (0.23)
ECO (b
9
) 0.001 0.001 0.01 0.007
*
0.007
*
0.009 0.021
***
0.0136
**
0.017
**
(0.32) (0.31) (0.32) (1.80) (1.68) (1.18) (4.26) (2.48) (2.49)
Adjusted R
2
0.51 0.53 0.54 0.64 0.56 0.59 0.95 0.85 0.86
Durbin–Watson
statistic
1.86 1.91 1.90 1.95 1.96 1.96 1.97 1.94 2.13
604 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
ductivity, and therefore, they provide empirical sup-
port to hypotheses one and three.
19
Results for the regressions on product defects
and material waste are summarized in Tables 4
and 5, respectively. As shown in Table 4, panel A,
the coe?cient OLDPLAN_TREND is positive
and statistically signi?cant in the defects level
model (b
1
= 0.251, p-value < .05) and positive and
insigni?cant in the change and percentage change
models, whereas NEWPLAN_TREND coe?cient
is negative and statistically signi?cant at the level
(b
2
= À0.052, p-value < .01), level change (b
2
=
À0.047, p-value < .01), and percentage change
(b
2
= À0.168, p-value < .01) models. A test of di?er-
ences between the two coe?cients indicates that the
NEWPLAN_TREND (b
2
= À0.52) coe?cient is sta-
tistically signi?cantly lower than the OLD-
PLAN_TREND coe?cient (b
1
= 0.251) in the
defects level model. However, note that there is no
di?erence between the two coe?cients in the
Table 3 (continued)
Variable Commercial unit Institutional unit Retail unit
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
N 49 49 49 49 49 49 49 49 49
Test of equality of main coe?cients (F-statistic)
H
0
:
b
1
= b
2
– 1.59 1.35 – – – – – –
H
1
:
b
1
– b
2
5.28
**
– – 5.83
**
4.37
**
4.8
**
21.2
***
6.91
***
7.30
***
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis.
All regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients
reports F-statistic.
This table presents OLS estimation results from regressing labor productivity, the ratio of earned hours to production hours, on a time trend
representing the period under each inventive plan, while controlling for general factors that could potentially a?ect labor productivity. The sample
consists of month observations from three production units for the period of September 1999 through September 2003. The following model is
estimated separately on productivity, change in productivity, and the percentage change in productivity by pooling data from the three production units:
PRODUCTIVITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
DEFECTS
it
þ b
4
HEADCOUNT
it
þ b
5
TURNOVER
it
þ b
6
ABSENTEEISM
it
þ b
7
OVERTIME
it
þ b
8
LAG TRAINING
it
þ b
9
ECO
it
þ b
10
RETAIL
it
þ b
11
INSTITUTIONAL
it
þ b
12
COMMERCIAL
it
þ e
it
And, the following model individually on each production unit:
PRODUCTIVITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
DEFECTS
it
þ b
4
HEADCOUNT
it
þ b
5
TURNOVER
it
þ b
6
ABSENTEEISM
it
þ b
7
OVERTIME
it
þ b
8
LAG TRAINING
it
þ b
9
ECO
it
þ t
it
Variables de?nitions: PRODUCTIVITY (level) = total standard hours of output divided by total production labor hours; PRODUCTIVITY (level
change) = change in productivity from1 month to the next; PRODUCTIVITY (%change) = percentage change in productivity from1 month to the
next; OLDPLAN_TREND = represents a time trend(t) for the period under the oldincentive planandit is measured as the number of months fromthe
beginning of the study period and before the introduction of the team bonus scheme, the period of September 1999 through March 2000. NEW-
PLAN_TREND = represents a time trend (t) for the period under the new incentive plan and it is measured as the number of months from the time in
which all changes to the incentive plan were implemented and thereafter, the period of July 2000 through September 2003. DEFECTS = the log of
?nished product defects (parts per million) adjusted for production volume; HEADCOUNT = number of production workers; TURNOVER
(turnover rate) = number of voluntary/involuntary resignations divided by the average production headcount; ABSENTEEISM (absenteeism
rate) = number of absences divided by the number of worked days in the month; OVERTIME = number of overtime hours scaled by headcount;
LAGTRAINING = prior month total hours of worker training scaled by headcount; ECO (engineering change orders) = number of manufacturing
process changes; RETAIL = a dummy variable set to 1 for observations from the retail manufacturing unit, zero otherwise; INSTITUTIONAL = a
dummy variable set to 1 for observations fromthe institutional manufacturing unit, zero otherwise; COMMERCIAL = a dummy variable set to 1 for
observations from the commercial manufacturing unit, zero otherwise. The COMMERCIAL dummy variable is set to the intercept.
19
For robustness, I alsoestimate analternative speci?cationof the
main model on productivity. Speci?cally, I regress each measure on
productivity (level, level change, and percentage change) perfor-
mance outcome ona time trend for eachof the three bonus schemes:
piece-rates, team-incentive, and the plant-incentive, in place of the
two time trends used currently to denote the periods under the old
and new incentive plan. Similarly, I performed tests of di?erences
across the three respective time trend coe?cients to detect di?er-
ences in performance across the three periods of interest. This
approach was followed in all three estimations. These set of results
are not currently reported in tables, but are available from the
author upon request. Overall, these results are qualitatively similar
to the ones reported in Table 3. In most cases, the coe?cient
estimates PLANT-INCENTIVE TREND, the time trend for the
period in which all elements of the new incentive plan operate fully,
is positive and statistically signi?cantly larger than the period in
which the team bonus operates in isolation and, in some cases, is
greater than the period under piece-rates.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 605
Table 4
Time-series regressions on product defects
Variable Log product defects (Parts per million)
Defects level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 2.22
***
À0.464 À0.165
(4.83) (1.03) (À1.09)
OLDPLAN_TREND (b
1
) 0.251
**
À0.022 0.001
(1.98) (À0.18) (0.30)
NEWPLAN_TREND (b
2
) À0.052
***
À0.047
***
À0.016
***
(À7.03) (À6.46) À(6.81)
HEADCOUNT (b
3
) 0.003
*
0.001 0.001
(1.74) (1.10) (1.29)
TURNOVER (b
4
) À1.84 À4.74
**
À1.38
*
(À0.76) (1.99) (À1.74)
ABSENTEEISM (b
5
) 21.5
*
20.2
*
5.50
(1.92) (1.83) (1.49)
ECO (b
6
) À0.0002 0.021 0.007
(À0.10) (0.74) (0.80)
DEFECTS_INCOMING (b
7
) 0.001 0.002
**
0.001
(1.03) (2.18) (1.45)
RETAIL (b
8
) À0.25
*
À0.003 À0.032
(À1.74) (À0.20) (À0.69)
INSTITUTIONAL (b
9
) 0.076 À0.74
***
À0.137
**
(0.38) (À4.08) (À2.23)
Adjusted R
2
0.64 0.61 0.56
Durbin–Watson statistic 1.97 1.96 1.96
N 147 147 147
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.40 1.90
H
1
: b
1
–b
2
5.70
**
– –
Variable Commercial unit Institutional unit Retail unit
Defects
level
Level
change
%
Change
Defects
level
Level
change
%
Change
Defects
level
Level
change
%
Change
Panel B: regressions by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.42 À0.89 À3.03 3.72
***
À0.18 À0.049 1.68 À0.69 À0.32
(1.00) (À1.52) (À1.48) (3.64) (À0.20) (À0.19) (1.56) (À0.73) (À0.89)
OLDPLAN_TREND (b
1
) 0.234
**
0.069 0.025 0.338 À0.297 À0.078 0.206 0.075 0.028
(2.67) (0.51) (0.53) (0.84) (À0.85) (À0.81) (0.81) (0.31) (0.29)
NEWPLAN_TREND (b
2
) À0.030
***
À0.051
***
À0.018
***
À0.052
**
À0.050
**
À0.140
**
À0.098
***
À0.084
***
À0.034
***
(À5.93) (À7.49) (À7.56) (À2.24) (À2.12) (À2.12) (À4.08) (À3.80) (À3.95)
HEADCOUNT (b
3
) 0.004
***
0.004 0.001 À0.003 À0.003 À.001 0.012
**
0.010
*
0.004
**
(3.46) (1.67) (1.64) (À0.75) (À0.67) (À0.66) (2.06) (1.92) (2.15)
TURNOVER (b
4
) 1.17 À1.11 À0.345 0.867 0.499 0.145 À10.7 À14.1
**
À5.58
**
(0.72) (À0.41) (À0.37) (0.15) (0.90) (0.90) (À1.60) (À2.18) (À2.22)
ABSENTEEISM (b
5
) 41.5
***
6.58 2.12 14.7 13.3 3.64 À25.8 À35.1 À12.5
(4.41) (0.50) (0.47) (0.70) (0.65) (0.63) (À0.75) (À1.06) (À0.98)
ECO (b
6
) 0.035 0.011 0.004 À0.135 À0.013 À0.003 À0.076 À0.028 À0.014
(1.60) (0.43) (0.43) (À0.21) (À0.20) (À0.21) (À1.29) (À0.48) (À0.50)
DEFECTS_INCOMING (b
7
) 0.006
*
0.001 0.001 0.00 0.00 0.00 À0.002 0.00 À0.005
(1.87) (1.15) (1.15) (0.54) (0.43) (0,41) (1.40) (À0.31) (À0.54)
Adjusted R
2
0.88 0.84 0.85 0.27 0.22 0.23 0.54 0.43 0.44
Durbin–Watson statistic 2.08 1.98 1.98 1.92 1.94 1.93 2.03 1.98 1.99
N 49 49 49 49 49 49 49 49 49
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.20 0.45 –** 0.48 0.42 – 2.54 2.24
H
1
: b
1
–b
2
8.66
***
– – 5.49
***
– – 8.91
***
– –
606 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 4 (continued)
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis. All
regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients reports
F-statistic.
This table presents OLS estimation results from regressing the log of ?nished product defects on a time trend representing the period under
each inventive plan, while controlling for general factors that could potentially a?ect product quality. The sample consists of month
observations from three production units for the period of September 1999 through September 2003. The following model is estimated
separately for level, level changes, and percentage change in product defects by pooling data from the three production units:
DEFECTS
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ b
8
RETAIL
it
þ b
9
INSTITUTIONAL
it
þ b
10
COMMERCIAL
it
þ e
it
And, the following model individually on each production unit:
DEFECTS
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ t
it
Variables de?nitions: DEFECTS (level) = log of total ?nished products defects (parts per million) scaled by production volume;
DEFECTS (level change) = change in products defects from 1 month to the next; DEFECTS (% change) = percentage change in product
defects from 1 month to the next; OLDPLAN_TREND = represents a time trend (t) for the period under the old incentive plan and it is
measured as the number of months from the beginning of the study period and before the introduction of the team bonus scheme, the
period of September 1999 through March 2000. NEWPLAN_TREND = represents a time trend (t) for the period under the new incentive
plan and it is measured as the number of months from the time in which all changes to the incentive plan were implemented and thereafter,
the period of July 2000 through September 2003; HEADCOUNT = number of production workers; TURNOVER (turnover
rate) = number of voluntary/involuntary resignations divided by the average production headcount; ABSENTEEISM (absenteeism
rate) = number of absences divided by the number of worked days in the month; ECO (engineering change orders) = number of manu-
facturing process changes; DEFECTS_INCOMING = total number of material component defects detected during an incoming
inspection upon arrival at the factory; RETAIL = a dummy variable set to 1 if retail manufacturing unit, zero otherwise; INSTITU-
TIONAL = a dummy variable set to 1 if institutional manufacturing unit, zero otherwise; COMMERCIAL = a dummy variable set to 1 if
commercial manufacturing unit, zero otherwise. The COMMERCIAL dummy variable is set to the intercept.
Table 5
Time-series regressions on material waste
Variable Material waste
Level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 0.012
***
À0.006
***
À1.82
***
(5.04) (À2.32) (À3.43)
OLDPLAN_TREND (b
1
) À0.004 0.0005 0.171
**
(À0.89) (1.18) (2.48)
NEWPLAN_TREND (b
2
) À0.016
***
À0.0002
***
À0.021
**
(À3.69) (À3.35) (À2.56)
HEADCOUNT (b
3
) À0.002
**
0.0001 0.006
***
(À2.03) (1.25) (3.09)
TURNOVER (b
4
) 0.003 À0.007 À0.18
(0.32) (À0.64) (À0.70)
ABSENTEEISM (b
5
) À0.020 0.076 9.40
(À0.43) (1.42) (0.77)
ECO (b
6
) 0.0002 0.00 0.021
(0.21) (0.72) (0.76)
RETAIL (b
7
) À0.001 0.0008 0.23
(À0.97) (0.72) (1.34)
INSTITUTIONAL (b
8
) 0.001 0.003
**
1.29
***
(0.14) (2.47) (6.04)
Adjusted R
2
0.21 0.22 0.41
Durbin–Watson statistic 1.79 2.06 1.99
N 147 147 147
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
1.49 –
H
1
: b
1
–b
2
4.33
**
– 6.77
***
(continued on next page)
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 607
changes or percentage change models. This could be
attributed to the fact that the impact on product
defects occurs almost immediately after the imposi-
tion of the quality quota and it is not as gradual as
in the case of productivity, so there is very little var-
iation in defects over time.
Table 5 (continued)
Variable Commercial unit Institutional unit Retail unit
Level Level
change
%
Change
Level Level
change
%
Change
Level Level
change
%
Change
Panel B: regressions by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.015
***
0.005 0.39 0.010
***
À0.003 0.095 0.003 À0.008
***
À2.82
***
(3.44) (1.32) (1.14) (4.79) (À1.42) (0.18) (1.48) (À3.99) (À4.61)
OLDPLAN_TREND
(b
1
)
0.001 0.002
**
0.14
**
0.003
***
À0.0005 0.086 À0.007 0.0001 0.40
**
(1.21) (2.97) (2.18) (3.67) (À0.88) (0.66) (À1.24) (0.33) (2.08)
NEWPLAN_TREND
(b
2
)
À0.002
*
À0.0001
*
À0.017
*
À0.002
***
À0.0002
***
À0.039
***
À0.009
*
À0.0009
**
À0.012
(À1.77) (À1.82) (À1.80) (À6.18) (À3.63) (À3.34) (À1.78) (À2.47) (À1.10)
HEADCOUNT (b
3
) À0.0002
*
À0.0003
**
À0.002
*
0.00 0.0003
**
0.006
**
0.00 0.0001 0.009
***
(À1.76) (À2.32) (À2.36) (0.55) (2.30) (2.35) (0.72) (1.45) (3.55)
TURNOVER (b
4
) À0.031
*
À0.031
*
À3.43
**
À0.025 À0.037
*
À6.78 À0.012 0.024
**
10.2
**
(1.77) (À1.98) (À2.21) (À1.49) (À1.95) (À1.47) (À0.78) (2.63) (2.70)
ABSENTEEISM (b
5
) À0.017 0.061 6.75 À0.11 0.041 À5.83 0.048 0.17
***
3.11
(À0.25) (1.01) (1.19) (À1.67) (0.50) (À0.29) (0.65) (2.81) (1.29)
ECO (b
6
) À0.00 À0.00 À0.00 0.001 0.00 0.096 À0.00 À0.001 À0.043
(À0.47) (À0.18) (0.10) (0.85) (0.39) (1.62) (À0.54) (À1.36) (À0.97)
Adjusted R
2
0.44 0.42 0.41 0.74 0.52 0.50 0.78 0.76 0.87
Durbin–Watson
statistic
1.70 1.65 1.96 1.94 1.91 1.98 1.95 2.01 1.75
N 49 49 49 49 49 49 49 49 49
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– – – 1.37 1.68 – 1.26 –
H
1
: b
1
–b
2
3.24
**
10.6
***
6.06
***
13.8
***
– – 3.06
**
– 4.04
***
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis.
All regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients
reports F-statistic.
This table presents OLS estimation results from regressing material waste on a time trend representing the period under each inventive
plan, while controlling for general factors that could potentially a?ect material waste. The sample consists of month observations from
three production units or the period of September 1999 through September 2003. The following model is estimated separately for level,
level changes, and percentage change in material waste by pooling data from the three production units:
WASTE
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ b
8
RETAIL
it
þ b
9
INSTITUTIONAL
it
þ b
10
COMMERCIAL
it
þ e
it
And, the following model individually on each production unit:
WASTE
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ t
it
Variables de?nitions: WASTE (level) = total amount of materials waste scaled by production volume; WASTE (level change) = change
in material waste from 1 month to the next; WASTE (% change) = percentage change in materials waste from 1 month to the next;
OLDPLAN_TREND = represents a time trend (t) for the period under the old incentive plan and it is measured as the number of months
from the beginning of the study period and before the introduction of the team bonus scheme, the period of September 1999 through
March 2000. NEWPLAN_TREND = represents a time trend (t) for the period under the new incentive plan and it is measured as the
number of months from the time in which all changes to the incentive plan were implemented and thereafter, the period of July 2000
through September 2003; HEADCOUNT = number of production workers; TURNOVER (turnover rate) = total number of voluntary/
involuntary resignations divided by the average production headcount); ABSENTEEISM (absenteeism rate) = number of absences
divided by the total number of worked days in the month; ECO (engineering change orders) = number of manufacturing process changes;
RETAIL = a dummy variable set to 1 if retail manufacturing unit, zero otherwise; INSTITUTIONAL = a dummy variable set to 1 if
institutional manufacturing unit, zero otherwise; COMMERCIAL = a dummy variable set to 1 if commercial manufacturing unit, zero
otherwise. The COMMERCIAL dummy variable is set to the intercept.
608 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 4, panel B, reports results of regressions by
each production unit. As shown, the results mirror
the same pattern of results from the pooled sample,
reinforcing that the conclusion that the new incen-
tive plan is associated with a decrease in products
defects.
Turning to the results on material waste in Table
5, we can observe that though the coe?cient esti-
mate for the time trend under the old compensation
plan is not signi?cant, the time trend for the new
plan’s time trend is negative and signi?cant at the
1% level across the pooled regressions, as well as
on the regressions by production unit, implying that
the amount of material waste decreased markedly
after the changes were made to the compensation
plan. Thus, these ?ndings and the ones on product
defects lend support to hypothesis two and partial
support to hypotheses three.
20
Overall, productivity and quality improvements
associated with the new incentive plan are sizeable
and fairly signi?cant as depicted in Fig. 2. As
observed, labor productivity begins to rise gradually
after the team bonus (output-target scheme) is intro-
duced, and this rise continues gradually for several
months until it reaches the upper 90–100 percentiles
through the end of the sample period. The same
broad pattern of improvements is observed for
product defects in panel B. However, quality
improvements are realized almost immediately once
the team bonus takes e?ect. Observe, for example,
the sharp decline in product defects in May 2000,
1 month after the quality quota is amended as part
of the team bonus. The steep drop in product
defects continues right up until the introduction of
the plant bonus in July 2000 and then levels o?
shortly after. This provides further evidence that
the imposition of the quality quota succeeded in
reducing the number of product defects.
In synthesis, all of the above results suggest
that the changes made to the original compensa-
tion plan, the adoption of team and the plant
bonus along with the various management con-
trol initiatives, led to a rise in productivity and
to a reduction in product defects and material
waste.
Regression results on absenteeism and turnover
To assess the impact of the e?ects of the new
incentive plan on absenteeism (Hypothesis 4), I
follow a similar approach and regress each variable
against a time trend for each, the old and new incen-
tive plan, and several controls, including the indus-
try’s rate of absenteeism and turnover in the region
where the plant operates. Results on absenteeism
are summarized in Table 6, panels A and B. As
shown, the coe?cient estimates for NEW-
PLAN_TREND is negative and signi?cant, whereas
the OLDPLAN_TREND coe?cient is positive and
signi?cant in the pooled and individual regressions
by production unit. These ?ndings indicate that
after controlling for the region’s industry rates of
absenteeism, the new incentive plan led to a reduc-
tion in absenteeism, and therefore, provides support
to Hypothesis 4.
Results on turnover are reported on Table 6,
panel A (columns 4–6), and on panel C. Interest-
ingly, in the pooled regressions, the coe?cient for
the OLDPLAN_TREND is positive and signi?cant
among the level, level changes, and percentage
change model. In contrast, the NEW-
PLAN_TREND coe?cient is negative and signi?-
cant among the three speci?cations. Similar results
are observed on the regressions by production unit.
Lastly, a test of di?erence between b
1
and b
2
on the
pooled as well as on the regressions by production
unit reveal that b
2
is signi?cantly lower than b
1
.
Taken together, these results indicate that the new
incentive plan led to a reduction in worker turnover
as well.
21
20
Just as in the case of productivity, to ensure robustness of my
results, I also estimate a similar model on both, product defects
and material waste. Speci?cally, I regress each performance
outcome (level, level change, and percentage change) on a time
trend for each of the three bonus schemes: piece-rates, team-
incentive, and the plant-incentive, in place of the two time trends
used currently to denote the periods under the old and new
incentive plan. Although the results are weaker to the ones
reported in tables due to the limited number of observations
surrounding the team-incentive, the general inference remains
unchanged. In most cases, the PLANT-INCENTIVE TREND
coe?cient is negative and signi?cantly smaller than the time trend
coe?cients for the periods under piece-rates and the team bonus
on the regressions on both regressions for product defects and
material waste.
21
To ensure robustness of these results, I followed a similar
approach to that of productivity and quality and regress
absenteeism and turnover on three time trend variables to
represent the periods under each of the three bonus schemes:
piece-rates, team-incentive, and the plant-incentive. These results
are not reported in tables, however, they are qualitatively similar
to the ones reported in Table 6.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 609
Discussion of the underlying factors that led to a rise
in productivity, quality and a reduction in absenteeism
and turnover
So far, the results from my regressions support
the central propositions that productivity, product
quality, absenteeism, and turnover improved after
the adoption of the new incentive plan. A follow
up question to address is which factors drive the
underlying performance improvements. Is it higher
e?ort, a selection e?ect, more peer pressure, or
heightened worker cooperation driving these
results? I recognize that given the available data, it
is impossible to empirically disentangle precisely
which factors drive the improvements. Nonetheless,
I address this question with qualitatively evidence
gathered from various interviews held with manag-
ers, production line supervisors, and assembly
January 1999 through September 2003
0%
20%
40%
60%
80%
100%
120%
J
a
n
-
9
9
M
a
r
-
9
9
M
a
y
-
9
9
J
u
l
-
9
9
S
e
p
-
9
9
N
o
v
-
9
9
J
a
n
-
0
0
M
a
r
-
0
0
M
a
y
-
0
0
J
u
l
-
0
0
S
e
p
-
0
0
N
o
v
-
0
0
J
a
n
-
0
1
M
a
r
-
0
1
M
a
y
-
0
1
J
u
l
-
0
1
S
e
p
-
0
1
N
o
v
-
0
1
J
a
n
-
0
2
M
a
r
-
0
2
M
a
y
-
0
2
J
u
l
-
0
2
S
e
p
-
0
2
N
o
v
-
0
2
J
a
n
-
0
3
M
a
r
-
0
3
M
a
y
-
0
3
J
u
l
-
0
3
S
e
p
-
0
3
Date
April 2000
Team output target
scheme
July 2000
Plant-wide bonus &
management control
initiatives
Previous incentive plan: piece-rates
New incentive plan: Team bonus, plant bonus, and control initiatives
September 1999 through September 2003
0
5,000
10,000
15,000
20,000
25,000
S
e
p
-
9
9
N
o
v
-
9
9
J
a
n
-
0
0
M
a
r
-
0
0
M
a
y
-
0
0
J
u
l
-
0
0
S
e
p
-
0
0
N
o
v
-
0
0
J
a
n
-
0
1
M
a
r
-
0
1
M
a
y
-
0
1
J
u
l
-
0
1
S
e
p
-
0
1
N
o
v
-
0
1
J
a
n
-
0
2
M
a
r
-
0
2
M
a
y
-
0
2
J
u
l
-
0
2
S
e
p
-
0
2
N
o
v
-
0
2
J
a
n
-
0
3
M
a
r
-
0
3
M
a
y
-
0
3
J
u
l
-
0
3
S
e
p
-
0
3
Date
Previous incentive plan: piece-rates
New incentive plan: team bonus, plant bonus, and control intitiatives
April 2000
Team output target
scheme
J
Plant-wide bonus &
management control
initiatives
uly 2000
a
b
Fig. 2. (a) Monthly labor productivity for entire plant. (b) Monthly ?nished product defects for entire plant.
610 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 6
Time-series regressions on absenteeism and turnover
Variable Absenteeism Turnover
Level Level change % Change Level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 0.018
***
À0.004 À0.22 0.023 À0.107
***
À1.41
***
(3.90) (À1.27) (À1.01) (1.15) (À5.38) (À3.77)
OLDPLAN_TREND (b
1
) 0.0027
***
0.0012
***
0.123
**
0.008
**
0.010
***
0.174
**
(2.75) (2.33) (2.30) (2.27) (2.62) (2.06)
NEWPLAN_TREND (b
2
) À0.0004
***
À0.0004
***
À0.021
***
À0.0002
***
À0.0016
***
À0.015
**
(À6.46) (À6.41) (À6.04) (À3.98) (À3.94) (À2.04)
HEADCOUNT (b
3
) 0.0001 0.0001 0.0001 0.0002
***
0.0001
**
0.003
**
(0.85) (0.70) (0.29) (2.99) (2.48) (2.10)
REGION_ABSENTEEISM (b
4
) 0.022 0.031 1.99 – – –
(0.76) (1.04) (1.10) – – –
REGION_TURNOVER (b
5
) – – – 0.001
***
0.227
**
4.83
***
– – – (3.17) (2.59) (2.78)
RETAIL (b
6
) À0.004
*
0.003 0.037 À0.021 0.004
***
0.112
(À1.85) (1.27) (0.32) (À1.46) (3.15) (0.56)
INSTITUTIONAL (b
7
) À0.002 0.002 0.139 À0.017 0.063
***
0.494
**
(À0.71) (1.18) (1.07) (1.18) (4.46) (2.23)
Adjusted R
2
0.37 0.33 0.34 0.48 0.44 0.27
Durbin–Watson statistic 1.94 1.97 1.96 1.97 1.98 1.94
N 147 147 147 147 147 147
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– – – – – –
H
1
: b
1
–b
2
10.2
***
7.18
***
7.00
***
7.98
***
9.73
***
5.28
***
Variable Commercial unit Institutional unit Retail unit
Level Level change % Change Level Level change % Change Level Level change % Change
Panel B: regressions on absenteeism by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.030
***
À0.004 À0.271 0.007 À0.0007 À0.57 0.010
***
À0.003 À0.383
(3.74) (À0.59) (À0.76) (1.07) (À1.30) (À1.61) (2.27) (À0.88) (À1.36)
OLDPLAN_TREND (b
1
) 0.005
***
0.0003 0.019 0.008
***
0.007
***
0.615
***
0.0015 0.0017
***
0.058
*
(2.83) (1.22) (0.26) (3.69) (3.66) (4.84) (1.37) (3.45) (1.74)
NEWPLAN_TREND (b
2
) À0.0005
***
À0.0004
***
À0.0234
***
À0.0005
***
À0.0004
***
À0.023
***
À0.0003
***
À0.0002
***
À0.021
***
(À5.66) (À4.00) (À4.07) (À3.34) (À3.44) (À3.33) (À5.95) (À4.07) (À5.65)
HEADCOUNT (b
3
) À0.0006 0.00 0.0004 0.0001
*
0.0005 0.003 0.0002 0.0001 0.001
(À1.60) (0.17) (0.25) (1.76) (1.54) (1.66) (1.16) (0.54) (0.99)
REGION_ABSENTEEISM (b
4
) 0.071 0.060 3.25 À0.011 0.010 0.983 À0.002 0.013 1.96
(0.99) (1.23) (1.38) (À0.17) (0.16) (0.24) (À0.80) (0.39) (0.86)
Adjusted R
2
0.32 0.36 0.37 0.46 0.47 0.56 0.57 0.56 0.60
(continued on next page)
F
.
J
.
R
o
m
a
´
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
4
(
2
0
0
9
)
5
8
9
–
6
1
8
6
1
1
Table 6 (continued)
Variable Absenteeism Turnover
Level Level change % Change Level Level change % Change
Durbin–Watson statistic 1.91 1.99 1.98 1.90 1.93 1.94 1.98 1.97 2.03
N 49 49 49 49 49 49 49 49 49
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.28 1.71 – – – – – –
H
1
: b
1
–b
2
9.80
***
– – 10.8
***
11.4
***
15.9
***
2.8
*
13.7
***
5.92
***
Variable Commercial unit Institutional unit Retail unit
Level Level change % Change Level Level change % Change Level Level change % Change
Panel C: regressions on turnover by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.158
***
0.022 0.186 0.012 À0.037 À0.23 À0.024 À0.094
***
À1.35
***
(4.07) (0.62) (0.64) (0.49) (À1.31) (À0.30) (À1.06) (À4.19) (À4.30)
OLDPLAN_TREND (b
1
) 0.021
**
0.017
**
0.139
**
0.023
***
0.020
*
0.097
***
0.009
*
0.021
***
0.32
***
(2.41) (2.08) (2.06) (2.40) (1.84) (4.76) (1.76) (3.86) (4.30)
NEWPLAN_TREND (b
2
) À0.0019
***
À0.002
***
À0.014
***
À0.002
***
À0.0017
***
À0.029
**
À0.001
***
À0.001
***
À0.019
***
(À4.58) (À4.62) (À4.63) (À4.26) (À4.35) (À2.11) (À3.87) (À4.30) (À4.25)
HEADCOUNT (b
3
) À0.0004
***
À0.0005
**
À0.003
**
0.0002 0.0001 À0.0008 0.0003
***
0.0003
***
0.004
***
(À2.71) (À2.65) (À2.66) (1.55) (1.25) (À0.17) (3.91) (3.22) (3.34)
REGION_TURNOVER (b
4
) 0.403 0.392 3.11 0.159 0.186 8.48
**
0.140 .166 2.36
(1.16) (1.19) (1.18) (0.48) (0.49) (2.46) (0.82) (0.97) (0.99)
Adjusted R
2
0.18 0.17 0.18 0.27 0.25 0.32 0.51 0.60 0.62
Durbin–Watson statistic 1.87 1.90 1.90 1.82 1.66
*
2.36 1.92 1.97 1.97
N
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– – – – – – – – –
H
1
: b
1
–b
2
7.02
***
5.37
**
5.28
**
6.72
***
4.12
**
22.8
***
4.43
***
17.3
***
21.3
***
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis.
All regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients reports F-statistic.
This table presents OLS estimation results from regressing absenteeism and worker turnover, respectively, on a time trend representing the period inventive plan, while controlling for
the regional rates of absenteeism and turnover of the maquiladora industry where the factory operates. The sample consists of month observations from three production units
covering the period of September 1999 through September 2003. The following models are estimated separately for level, level changes, and rate change in absenteeism and turnover,
respectively, pooling data from the three production units:
ð1Þ ABSENTEEISM
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
REGION ABSENTEEISM
it
þ b
5
RETAIL
it
þ b
6
INSTITUTIONAL
it
þ b
7
COMMERCIAL
it
þv
it
ð2Þ TURNOVER
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
REGION TURNOVER
it
þ b
5
RETAIL
it
þ b
6
INSTITUTIONAL
it
þ b
7
COMMERCIAL
it
þ s
it
6
1
2
F
.
J
.
R
o
m
a
´
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
4
(
2
0
0
9
)
5
8
9
–
6
1
8
workers who witnessed the introduction of the new
incentive plan ?rsthand.
Based on their opinions, I o?er the following
potential explanations. First, the perception among
workers is that the shift from piece-rates to the out-
put-target scheme induce them to work harder to
reach the output and quality quotas to earn a bonus.
However, a crucial mechanism that made the incen-
tive work was in providing teams with feedback on
their performance. This mechanism enable teams to
monitor their progress more e?ectively plus it enabled
them to exercise more control within their respective
production cell in such a way that they were better
able to coordinate their e?orts to reach the speci?ed
output and quality targets and to monitor each
other’s e?orts more rigorously to minimize shirking.
Second, management concurred that the plant-
wide incentive and the initiative to report feedback
on each team’s performance throughout the factory
has led to more cooperation and information sharing
throughout the plant. For example, production
supervisors acknowledged that such initiative has
promoted the rate of learning and information shar-
ing between the teams; stressing that it is not uncom-
mon for a particular production cell to communicate
with others upon discovering a method to speed-up
production, eliminate bottlenecks, or identify adefec-
tive component. Similarly, production teams meet on
a regular basis to discuss ways to improve productiv-
ity and quality. This behavior was uncommon under
the old incentive plan.
Third, workers noted that the levels of peer pres-
sure within each team rise as a result of the output-
target scheme and the accompanying mechanism
established to discipline shirkers. Several interviews
with production workers con?rm that this initiative
succeeded in raising the levels of peer pressure across
teams. This behavior also was not as prevalent under
the old incentive plan. Although piece-rates created
some friction between potential shirkers and high
performers, shirking was not formally disciplined in
the ?rm; thus, peer pressure was to some extent inef-
fective. In the words of a production worker: ‘‘. . .If a
member of a production cell is not pulling his weight
to meet the daily output quota, we demand extra
e?ort or otherwise we report his behavior with the
production supervisor. If this behavior persists, we
request his dismissal from the team”.
Last, it is possible that the new incentive plan
attracted workers with higher skill, producing a
sorting or selection e?ect. Therefore, a higher
quality of the workforce could have led to higher A
n
d
,
t
h
e
f
o
l
l
o
w
i
n
g
m
o
d
e
l
s
i
n
d
i
v
i
d
u
a
l
l
y
o
n
e
a
c
h
p
r
o
d
u
c
t
i
o
n
u
n
i
t
:
ð
1
Þ
A
B
S
E
N
T
E
E
I
S
M
i
t
¼
b
0
þ
b
1
O
L
D
P
L
A
N
T
R
E
N
D
i
t
þ
b
2
N
E
W
P
L
A
N
T
R
E
N
D
i
t
þ
b
3
H
E
A
D
C
O
U
N
T
i
t
þ
b
4
R
E
G
I
O
N
A
B
S
E
N
T
E
E
I
S
M
i
t
þ
v
i
t
ð
2
Þ
T
U
R
N
O
V
E
R
i
t
¼
b
0
þ
b
1
O
L
D
P
L
A
N
T
R
E
N
D
i
t
þ
b
2
N
E
W
P
L
A
N
T
R
E
N
D
i
t
þ
b
3
H
E
A
D
C
O
U
N
T
i
t
þ
b
4
R
E
G
I
O
N
T
U
R
N
O
V
E
R
i
t
þ
s
i
t
V
a
r
i
a
b
l
e
s
d
e
?
n
i
t
i
o
n
s
:
A
B
S
E
N
T
E
E
I
S
M
(
a
b
s
e
n
t
e
e
i
s
m
r
a
t
e
)
=
n
u
m
b
e
r
o
f
a
b
s
e
n
c
e
s
d
i
v
i
d
e
d
b
y
t
h
e
t
o
t
a
l
n
u
m
b
e
r
o
f
w
o
r
k
e
d
d
a
y
s
i
n
t
h
e
m
o
n
t
h
;
A
B
S
E
N
T
E
E
I
S
M
(
l
e
v
e
l
c
h
a
n
g
e
)
=
c
h
a
n
g
e
i
n
a
b
s
e
n
t
e
e
i
s
m
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
A
B
S
E
N
T
E
E
I
S
M
(
%
c
h
a
n
g
e
)
=
p
e
r
c
e
n
t
a
g
e
c
h
a
n
g
e
i
n
a
b
s
e
n
t
e
e
i
s
m
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
T
U
R
N
O
V
E
R
(
t
u
r
n
o
v
e
r
r
a
t
e
)
=
n
u
m
b
e
r
o
f
v
o
l
u
n
t
a
r
y
/
i
n
v
o
l
u
n
t
a
r
y
r
e
s
i
g
n
a
t
i
o
n
s
d
i
v
i
d
e
d
b
y
t
h
e
a
v
e
r
a
g
e
p
r
o
d
u
c
t
i
o
n
h
e
a
d
c
o
u
n
t
;
T
U
R
N
O
V
E
R
(
l
e
v
e
l
c
h
a
n
g
e
)
=
c
h
a
n
g
e
i
n
t
u
r
n
o
v
e
r
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
T
U
R
N
O
V
E
R
(
%
c
h
a
n
g
e
)
=
p
e
r
c
e
n
t
a
g
e
c
h
a
n
g
e
i
n
t
u
r
n
o
v
e
r
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
O
L
D
P
L
A
N
_
T
R
E
N
D
=
r
e
p
r
e
s
e
n
t
s
a
t
i
m
e
t
r
e
n
d
(
t
)
f
o
r
t
h
e
p
e
r
i
o
d
u
n
d
e
r
t
h
e
o
l
d
i
n
c
e
n
t
i
v
e
p
l
a
n
a
n
d
i
t
i
s
m
e
a
s
u
r
e
d
a
s
t
h
e
n
u
m
b
e
r
o
f
m
o
n
t
h
s
f
r
o
m
t
h
e
b
e
g
i
n
n
i
n
g
o
f
t
h
e
s
t
u
d
y
p
e
r
i
o
d
a
n
d
b
e
f
o
r
e
t
h
e
i
n
t
r
o
d
u
c
t
i
o
n
o
f
t
h
e
t
e
a
m
b
o
n
u
s
s
c
h
e
m
e
,
t
h
e
p
e
r
i
o
d
o
f
S
e
p
t
e
m
b
e
r
1
9
9
9
t
h
r
o
u
g
h
M
a
r
c
h
2
0
0
0
;
N
E
W
P
L
A
N
_
T
R
E
N
D
=
r
e
p
r
e
s
e
n
t
s
a
t
i
m
e
t
r
e
n
d
(
t
)
f
o
r
t
h
e
p
e
r
i
o
d
u
n
d
e
r
t
h
e
n
e
w
i
n
c
e
n
t
i
v
e
p
l
a
n
a
n
d
i
t
i
s
m
e
a
s
u
r
e
d
a
s
t
h
e
n
u
m
b
e
r
o
f
m
o
n
t
h
s
f
r
o
m
t
h
e
t
i
m
e
i
n
w
h
i
c
h
a
l
l
c
h
a
n
g
e
s
t
o
t
h
e
i
n
c
e
n
t
i
v
e
p
l
a
n
w
e
r
e
i
m
p
l
e
m
e
n
t
e
d
a
n
d
t
h
e
r
e
a
f
t
e
r
,
t
h
e
p
e
r
i
o
d
o
f
J
u
l
y
2
0
0
0
t
h
r
o
u
g
h
S
e
p
t
e
m
b
e
r
2
0
0
3
;
H
E
A
D
C
O
U
N
T
=
n
u
m
b
e
r
o
f
p
r
o
d
u
c
t
i
o
n
w
o
r
k
e
r
s
;
R
E
G
I
O
N
_
A
B
S
E
N
T
E
E
I
S
M
=
t
h
e
r
e
g
i
o
n
’
s
m
a
q
u
i
l
a
d
o
r
a
i
n
d
u
s
t
r
y
r
a
t
e
o
f
a
b
s
e
n
t
e
e
i
s
m
;
R
E
G
I
O
N
_
T
U
R
N
O
V
E
R
=
t
h
e
r
e
g
i
o
n
’
s
m
a
q
u
i
l
a
d
o
r
a
r
a
t
e
o
f
t
u
r
n
o
v
e
r
;
R
E
T
A
I
L
=
a
d
u
m
m
y
v
a
r
i
a
b
l
e
s
e
t
t
o
1
i
f
r
e
t
a
i
l
m
a
n
u
f
a
c
t
u
r
i
n
g
u
n
i
t
,
z
e
r
o
o
t
h
e
r
w
i
s
e
;
I
N
S
T
I
T
U
T
I
O
N
A
L
=
a
d
u
m
m
y
v
a
r
i
a
b
l
e
s
e
t
t
o
1
i
f
i
n
s
t
i
t
u
t
i
o
n
a
l
(
l
o
c
k
e
r
l
o
c
k
)
m
a
n
u
f
a
c
t
u
r
i
n
g
u
n
i
t
,
z
e
r
o
o
t
h
e
r
w
i
s
e
;
C
O
M
M
E
R
C
I
A
L
=
a
d
u
m
m
y
v
a
r
i
a
b
l
e
s
e
t
t
o
1
i
f
c
o
m
m
e
r
c
i
a
l
m
a
n
u
f
a
c
t
u
r
i
n
g
u
n
i
t
,
z
e
r
o
o
t
h
e
r
w
i
s
e
.
T
h
e
C
O
M
M
E
R
C
I
A
L
d
u
m
m
y
v
a
r
i
a
b
l
e
i
s
s
e
t
t
o
t
h
e
i
n
t
e
r
c
e
p
t
.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 613
productivity andbetter quality over time. Thoughit is
di?cult to make a direct test of that performance
improvements resulted from a selection e?ect given
that there is no data available on individual workers
productivity, production supervisors pointed out
that the incentive plan did have some e?ect in attract-
ing (repelling) more (less) productive employees.
As for the underlying reasons that led to a reduc-
tion in absenteeism, ?rst, it was con?rmed that
absenteeism was in part mitigated by the imposition
of a penalty in the plant bonus. The threat of being
excluded from participating in the bonus plan for
not keeping good attendance was indeed a powerful
mechanism in helping reduce absenteeism. Further-
more, workers noted that peer pressure within each
team also contributed to lowering absenteeism.
Absentees made it more di?cult for teams to reach
the day’s quotas; workers exerted pressure on co-
workers who regularly missed work by threatening
to dismiss them from the team.
Was the new incentive plan cost e?ective for the ?rm?
To this point, this study has argued that the new
incentive plan has led to various performance
improvements for this ?rm. A ?nal question to
address is whether the new incentive has been cost
e?ective. In other words, are the improvements in
productivity and quality, and the decrease in absen-
teeism and turnover economically important for the
?rm? To illustrate better, the factory experienced,
on average, a 31% rise in productivity and a 95%
reduction in product defects. However, it could very
well be the case that the incentive plan also raised
labor costs, o?-setting any performance improve-
ments and making the plan cost ine?ective.
To better understand the impact of the incentive
plan on production costs, I brought these results to
the attention of management. They indicated that
indeed the ?rm experienced a rise in labor costs
per worker as a result of having to pay higher
bonuses under the new plan; however, it was
emphasized that the rise in labor costs was o?-set
with higher productivity and better product quality.
For example, management estimates that the 31%
rise in productivity amounts to an additional output
between 5 and 7 million locks annually. It is impor-
tant to stress that this is not a one-time increase in
output that dissipates; the incremental volume
attributed to the gains in productivity is permanent
and carries over to subsequent years. Also, the costs
related to product defects, such as rework, plus
material waste costs after adjusting for production
volume, decrease substantially. In this regard, it is
also worthwhile mentioning that product defects
have been under control. For example, the plant
has had less than 300 defects (PPMs) per month
consistently since the incentive plan was introduced,
the conventional assessment of Six Sigma quality.
Finally, since turnover andabsenteeismdecreased,
this helped reduce training costs on new workers. In
sum, the plant’s bottomline improved despite the rise
in labor costs after the changes were made to the com-
pensation plan. If we account for all of these factors,
managers believed the improvements in manufactur-
ing performance easily o?-set the increase in direct
labor costs.
Conclusion
Using data from three production units of a large
manufacturing plant that employs production teams
in its assembly operations, this study investigates
how workers react to changes made to an existing
incentive plan for teams and whether the underlying
changes led to improvements in productivity and
product quality, and in reducing absenteeism and
turnover. The ?rm switched from an incentive plan
that relied on piece-rates to a two tier incentive plan
that rewarded individual team performance and
plant-wide performance. The new incentive plan
was introduced concurrently with several organiza-
tional control initiatives that aim to facilitate coop-
eration, reinforce monitoring, promote learning,
and encourage peer pressure between workers.
These include the distribution of performance
reports to teams, mechanisms to facilitate the
exchange of information to improve performance,
and rules to punish shirking and uncooperativeness.
I ?nd signi?cant improvements in worker pro-
ductivity and product quality as well as reductions
in worker absenteeism and turnover after the imple-
mentation of the incentive plan. These ?ndings
underscores an important point that has not been
emphasized in existing empirical studies of incentive
pay for teams: the need to introduce management
control and organizational changes in tandem with
incentive pay to capture greater incentive e?ects
from workers. In particular, organizational initia-
tives such as reporting performance to workers with
the application of formal rules to punish shirking
and uncooperative behavior are more e?ective moti-
vators than the e?ect generated by monetary incen-
tives alone.
614 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
In all fairness, some caveats apply. First, I recog-
nize that attempting to disentangle the e?ects on
productivity and defects empirically to each element
in the incentive plan presents limitations. Also, the
small window of separation between the implemen-
tation periods of the two incentives schemes pre-
cludes complete isolation of the performance
impact across the two. Finally, the traditional
caveat of data limitations with ?eld-based research
also applied in this paper, especially since the sam-
ple was truncated and could not account for all
observations in the pre incentive period.
In spite this, my results raised several new ques-
tions not previously address in the literature. Mainly,
my ?ndings that monetary incentives must be com-
plemented with other management control initiatives
when designing incentive plans for teams suggest that
?rms may want to consider adopting them jointly to
achieve better results. I suspect that this conclusion
holds more general; therefore, I leave it to future
research studies to examine the broader applicability
of these ?ndings. Perhaps experimental studies could
replicate the incentive plan implemented at this fac-
tory and see whether the same results hold.
Acknowledgements
I owe particular thanks to Jose Saralegui and the en-
tire management teamat the researchsite for their valu-
able support. This paper is based in part on my
dissertation at The University of Arizona. I gratefully
acknowledge the guidance of Leslie Eldenburg (Chair),
Alfonso Flores-Lagunes, Kaye Newberry, Je? Schatz-
berg, and William Waller. I would also like to thank
Ashiq Ali, David Larcker, Brian Rountree, Mark
Trombley, William Schwartz, Wim Van der Stede, an
anonymous referee, and workshop participants at the
University of Arizona, University of California Irvine,
EmoryUniversity, Rice University, andthe 2003Amer-
ican Accounting Association Annual Meeting for pro-
viding helpful comments on earlier drafts. Also, special
thanks to Michael Costa for his editorial assistance.
Appendix
Worker compensation under the original and new
incentive plan
Compensation under the original incentive plan
Each worker received a daily ?at wage of 70,
80, 91, or 101 Mexican pesos, depending on a
worker’s seniority and level of certi?cation. In
addition to the daily wage, each worker could
receive an attendance bonus of approximately
$115 pesos per week ($23 pesos per day). There
is also a group piece-rate bonus that ?uctuates
according to the output generated by each produc-
tion group. While some groups were more dynamic
than others, each worker earned on average a
piece-rate bonus of $600 pesos.
22
The average daily
compensation under the original incentive plan is
as follows:
Daily wage $90
Attendance bonus 23
Group-piece-rates 30
Total daily compensation $143 pesos
Compensation under the new incentive plan
Although, the base wage increased slightly dur-
ing the study period, the resulting incremental
change does not di?er signi?cantly from the base
wage under the old incentive plan.
23
Thus, for sim-
pli?cation and easy exposition, I am using the
daily ?at wage that was in place under the old
incentive plan. Under the team-level bonus, each
team member receives a daily bonus of $30 pesos
if the team meets the daily output and quality
quotas. Further, after reaching the output quota,
the incentive pays an additional bonus for each
unit of output above the day’s quota. As for the
plant bonus, each worker receives a cash bonus
equal to 6%, 8%, and 10% of the total accumu-
lated earnings during the quarter. The average
daily compensation under the new incentive plan
is as follows:
Daily wage $90
Output-target bonus 40
Gain-sharing 35
Total $165 pesos
22
I use an average of 22 working days in a month to estimate
the daily piece-rate bonus.
23
The daily ?at wage earned by each worker has been constant
throughout the study period. There have been two minor
increases to the minimum wage in the country, and the ?at wage
in the factory has increased in the same proportion.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 615
Mathematical and graphical representation of the
incentive plans
Original incentive plan
Under the original incentive plan, individual
compensation equals a daily ?at wage (a), an atten-
dance bonus
, and piece-rates (b). Thus, compen-
sation may be represented mathematically as
follows:
x ¼ a þ w þ bðxÞ
where x = number of units of output.
Average worker compensation is represented
graphically as follows:
I
+
?
?
? ?
New incentive plan
Team-incentive
The mathematical representation of the team-
incentive is expressed as follows:
x ¼ a þ RðXÞ À Pð/Þ
Let
rðxÞ ¼
c þ bðX
I
ÀX
T
Þ if x
i
> x
t
c if x
i
¼ x
t
0 if x
i
< x
t
0
B
@
pð/Þ ¼
rðxÞ if /
I
P/
Ã
; given x
i
0 if /
I
< /
Ã
; given x
i
where r(x) is the reward function, p(/) is a penalty
for failing to maintain the defects threshold, x total
compensation, a = a daily ?at wage, c = team bo-
nus for meeting the output-target, b = a piece-rate
factor for exceeding the output-target, x
t
= output-
target (in units of output), x
i
= actual output,
f
*
= defects threshold, f
i
= actual defects. The team
bonus takes the form of a ?at wage when a team
fails to reach either the output and quality targets.
The following are possible reward solutions when
a team exceeds the product defects threshold
(f
i
< f
*
):
ð1Þ x
i
> x
t
and /
i
< /
Ã
;
then x
i
¼ a
i
þ c
i
þ bðX
i
À X
t
Þ
ð2Þ x
i
¼ x
t
and /
i
< /
Ã
;
then x
i
¼ a
i
þ c
i
Eqs. (1) and (2) can be illustrated as follows:
i
i
<
*
I
+
I
+ (X
i
-X
t
)
i
+
i
i
X
t
X
j
?
?
?
?
?
?
?
? ?
If a team’s output (defects) falls (exceeds) the tar-
get, worker compensation equals a ?at wage (a):
ð3Þ x
i
> x
t
and /
i
P/
Ã
;
then x
i
¼ ½a
i
þ c
i
þ bðX
i
À X
t
Þ?
À ½c
i
þ bðX
i
À X
t
Þ? ¼ a
i
ð4Þ x
i
¼ x
t
and /
i
P/
Ã
;
then x
i
¼ ½a
i
þ c
i
? À c
i
¼ a
i
ð5Þ x
i
< x
t
and /
i
< /
Ã
;
then x
i
½a
i
þ c
i
?À ¼ a
i
Eqs. (3), (4), and (5) can be illustrated as follows:
I
i
*
[
i
+
i
]-
i
=
i
i i
[
i
+
i
+ (X
i
-X
t
)] – [
i
+ (X
i
-X
t
)] =
i
X
t
X
j
?
? ? ?
? ?
?
?
? ?
?
? ?
>_ ? ?
616 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Plant-incentive
The mathematical representation of the plant-
incentive is expressed as follows:
x ¼ c½X
t
Á
^
W? þ ð1 À cÞ½/
t
Á
^
W ?
X
t
¼
:06 if x
2
> x
a
Px
1
:08 if x
3
> x
a
Px
2
:10 if x
a
Px
3
0
B
@
/
t
¼
:06 if /
2
< /
a
6 /
1
:08 if /
3
< /
a
6 /
2
:10 if /
a
6 /
3
0
B
@
where
x = total bonus,
x
t
= labor productivity quarterly targets,
x
t
= {x
1
, x
2
, x
3
},
/
t
= quality (defects rate) quarterly targets,
/
t
= {/
1
, /
2
, /
3
},
c = reward weight,
x
a
= actual plant’s productivity in the quarter,
/
a
= actual plant’s product defects in the
quarter.
References
Alchian, A., & Demsetz, H. (1972). Production, information
costs, and economic organization. American Economic
Review, 62(5), 777–795.
Allen, S. (1981). Compensation, safety, and absenteeism: Evi-
dence from the paper industry. Industrial and Labor Relations
Review, 34(2), 207–218.
Baiman, S. (1982). Agency research in managerial accounting: A
survey. Journal of Accounting Literature, 1, 154–213.
Baker, G. (1992). Incentive contracts and performance measure-
ment. Journal of Political Economy, 100(3), 598–617.
Baker, G., Jensen, M., & Murphy, K. (1988). Compensation and
incentives: Practice vs. theory. Journal of Finance, 43, 593–616.
Bandiera, O., Barankay, I., & Rasul, I. (2004). Cooperation in
collective action. The Economics of Transition, 13(3),
473–498.
Bandiera, O., Barankay, I., & Rasul, I. (2005). Social preferences
and the response to incentives: Evidence from personnel data.
The Quarterly Journal of Economics, 120(3), 473–498.
Banker, R., Field, J., Schroeder, R., & Sinha, K. (1996a). Impact
of work teams on manufacturing performance: A longitudinal
?eld study. Academy of Management Journal, 39, 867–890.
Banker, R., Lee, S., & Potter, G. (1996b). A ?eld study of the
impact of a performance-based incentive plan. Journal of
Accounting and Economics, 21(2), 195–226.
Banker, R., Lee, S., Potter, G., & Srinivasan, D. (2000). An
empirical investigation of an incentive plan that includes
non?nancial performance measures. The Accounting Review,
75, 1–65.
Barron, J., & Gjerde, K. (1997). Peer pressure in an agency
relationship. Journal of Labor Economics, 15, 917–962.
Blinder, A. (1999). Paying for productivity: A look at the evidence
(2nd ed.). Washington, DC: The Brookings Institution.
Boning, B., Ichniowski, C., & Shaw, K. (2007). Opportunity
counts: Teams and the e?ectiveness of production incentives.
Journal of Labor Economics, 25(4), 613–650.
Bonner, S., & Sprinkle, G. (2002). The e?ects of monetary
incentives on e?ort and task performance: Theories, evidence,
and a framework for research. Accounting, Organizations, and
Society, 27, 303–345.
Brown, D., & Armstrong, M. (1999). Paying for contribution.
Dover, NH: Kogan Page.
Brown, S., Fakhfakh, F., & Sessions, J. (1999). absenteeism and
employee sharing: An empirical analysis based on French
panel data, 1981–1991. Industrial and Labor Relations Review,
52(2), 234–251.
Brown, P., & Phillips, P. (1986). The decline of the piece-rate
system in California canning: Technological innovation, labor
management, and union pressure, 1890–1947. The Business
History Review, 60(4), 564–601.
Bullock, R., & Lawler, E. (1984). Gainsharing: A few questions,
and few answers. Human Resources Management, 23(1),
23–40.
Cachon, G., & Terwiesch, C. (2006). Matching supply with
demand: An introduction to operations management (1st ed.).
McGraw Hill, Irwin.
Che, Y., & Yoo, S. (2002). Optimal incentives for teams.
American Economic Review, 91, 525–541.
Chow, C. (1983). The e?ects of job standard tightness and
compensation scheme on performance: An exploration of
linkages. The Accounting Review, 58, 667–685.
DeMatteo, J., Eby, L., & Sundstrom, E. (1998). Team-based
rewards: Current empirical evidence and directions for future
research. In B. Staw, L. Cummins, (Eds.), Research in
organizational behavior (Vol. 20, pp. 141–183).
Deutsch, M. (1949). An experimental study of the e?ects of
cooperation and competition upon group process. Human
Relations, 2, 129–152.
Deutsch, M. (1973). The resolution of con?ict: Constructive and
destructive processes (1st ed.). Yale University Press.
Drago, R., & Garvey, G. (1998). Incentives for helping on the
job: Theory and evidence. Journal of Labor Economics, 16(1),
1–25.
Fisher, G., Pe?er, S., & Sprinkle, G. (2003). Budget-based
contracts, budget levels, and group performance. Journal of
Management Accounting Research, 15, 51–74.
Hansen, D. (1997). Worker performance and group incentives: A
case study. Industrial and Labor Relations Review, 51, 37–49.
Hayes, R., & Clark, K. (1985). Exploring the sources of
productivity di?erences at the factory level. In K. B. Clark,
R. Hayes, & C. Lorenz (Eds.), The uneasy alliance: Managing
the productivity-technology dilemma. Boston: Harvard Busi-
ness School Press.
Holmstrom, B. (1979). Moral hazard and observability. The Bell
Journal of Economics, 30(1), 74–91.
Holmstrom, B. (1982). Moral hazard in teams. The Bell Journal
of Economics, 13(2), 324–340.
Holmstrom, B., & Milgrom, P. (1994). The ?rm as an incentive
system. American Economic Review, 84(4), 972–991.
Horngren, C., Datar, S., & Foster, G. (2002). Cost accounting: A
managerial emphasis (12th ed.). Prentice Hall.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 617
Ichniowski, C., & Shaw, K. (2003). Beyond incentive pay:
Insiders’ estimates of the value of complementary human
resource management practices. Journal of Economic Per-
spectives, 17(1), 155–180.
Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The e?ects of
human resource management practices on productivity: A
study of steel ?nishing lines. American Economic Review,
83(3), 291–313.
Indjejikian, R. (1999). Performance evaluation and compensation
research: An agency perspective. Accounting Horizons, 13,
147–157.
Kandel, E., & Lazear, E. (1992). Peer pressure and partnership.
Journal of Political Economy, 100(4), 801–817.
Knez, M., & Simester, D. (2001). Firm-wide incentives and
mutual monitoring at continental Airlines. Journal of Labor
Economics, 19(4), 743–772.
Lazear, E. (1986). Salaries and piece-rates. Journal of Business,
59(3), 405–431.
Lazear, E. (1991). Labor economics and the psychology of the
organizations. Journal of Economic Perspectives, 5(2), 89–110.
Lazear, E. (2000). Performance pay and productivity. The
American Economic Review, 90(5), 1346–1361.
Macho-Stadler, I., & Perez-Castrillo, J. (1993). Moral hazard
with several agents: The gains from cooperation. International
Journal of Industrial Organizations, 11, 73–100.
Milgrom, P., & Roberts, J. (1992). Economics, organizations, and
management (1st ed.). Upper Saddle River, NJ: Prentice Hall.
Milgrom, P., & Roberts, J. (1995). Complementarities and ?t:
Strategy, structure, and organizational change in manufac-
turing. Journal of Accounting and Economics, 19, 179–208.
Nalbantian, H., & Schotter, A. (1997). Productivity under group
incentives: An experimental study. The American Economic
Review, 87, 314–341.
Paarsch, H., & Shearer, B. (2000). Piece rates, ?xed wages, and
incentive e?ects: Statistical evidence from payroll records.
International Economic Review, 41(1), 59–92.
Petersen, T. (1992). Individual, collective, and systems rationality
in work groups: Dilemmas and market-type solutions. Amer-
ican Journal of Sociology, 98, 469–510.
Prendergast, C. (1999). The provision of incentives in ?rms.
Journal of Economic Literature, XXXVII, 7–63.
Rankin, F. (2004). Coordinating e?orts under team-based and
individual incentives: An experimental analysis. Contempo-
rary Accounting Research, 21, 191–221.
Schuster, M. (1983). Forty years on scalon plan research: A
review of the descriptive and empirical literature. In Interna-
tional yearbook of organizational democracy. Chichester: John
Wiley.
Schuster, M. (1984). The scanlon plan. A longitudinal analysis.
The Journal of Applied Behavioral Science, 20, 23–38.
Sewell, G. (1998). The discipline of teams: The control of
team-based industrial work through electronic and peer
surveillance. Administrative Science Quarterly, 43,
397–428.
Shearer, B. (2004). Piece rates, ?xed wages and incentives:
Evidence from a ?eld experiment. Review of Economic Studies,
71(2), 513–534.
Sprinkle, G. (2000). The e?ects of incentive contracts on learning
and performance. The Accounting Review, 75, 299–326.
Sprinkle, G. (2003). Perspectives on experimental research in
managerial accounting. Accounting, Organizations, and Soci-
ety, 28, 287–318.
Stiglitz, J. (1975). Incentives, risk, and information: Notes
towards a theory of hierarchy. The Bell Journal of Economics,
6(2), 552–579.
Waller, W., & Chow, C. (1985). The self-selection and e?ort
e?ects of standard-based employment contracts: A framework
and some empirical evidence. The Accounting Review, LX(3),
458–476.
Weitzman, M., & Kruse, D. (1999). Pro?t sharing and produc-
tivity. In A. Blinder (Ed.), Paying for productivity (2nd ed.).
Washington, DC: The Brookings Institution.
Welbourne, T., Balkin, D., & Gomez-Mejia, L. (1995). Gain-
sharing and mutual monitoring: A combined agency-organi-
zational justice interpretation. The Academy of Management
Journal, 38(3), 881–899.
Young, M. (1999). Field research methods in management
accounting. Accounting Horizons, 13(1), 76–84.
Young, M., Fisher, J., & Lindquist, T. (1993). The e?ects of
intergroup competition and intragroup cooperation on slack
and output in a manufacturing setting. The Accounting
Review, 68(3), 466–481.
618 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
doc_658502700.pdf
Using data from three production units of a large manufacturing plant that employs production teams in its assembly
operations, this paper examines how changes made to an existing team-based incentive plan affects labor productivity,
product quality, and worker absenteeism. The firm switched from a piece-rates contract and an attendance bonus and instituted
a two tier incentive plan comprising two different but complementary performance-based bonus schemes: one tier
based on individual team performance and the other tier on plant-wide performance. The incentives were introduced concurrently
with management control initiatives intended to facilitate cooperation and monitoring among the teams. I find
significant productivity gains and improvements in quality and absenteeism associated with the new incentive plan. These
findings underscore two important points that have not been emphasized in existing empirical studies of incentive pay:
incentive contracts for teams generate superior performance using combination of incentives, and the need to introduce
organizational changes to facilitate cooperation and peer monitoring in tandem with incentive pay to capture greater
incentive effects in team production.
An analysis of changes to a team-based incentive plan and
its e?ects on productivity, product quality, and absenteeism
Francisco J. Roma´n
*
Rawls College of Business, Texas Tech University, Department of Accounting, P.O. Box 42101, Lubbock, TX 79409, United States
Abstract
Using data from three production units of a large manufacturing plant that employs production teams in its assembly
operations, this paper examines how changes made to an existing team-based incentive plan a?ects labor productivity,
product quality, and worker absenteeism. The ?rm switched from a piece-rates contract and an attendance bonus and insti-
tuted a two tier incentive plan comprising two di?erent but complementary performance-based bonus schemes: one tier
based on individual team performance and the other tier on plant-wide performance. The incentives were introduced con-
currently with management control initiatives intended to facilitate cooperation and monitoring among the teams. I ?nd
signi?cant productivity gains and improvements in quality and absenteeism associated with the new incentive plan. These
?ndings underscore two important points that have not been emphasized in existing empirical studies of incentive pay:
incentive contracts for teams generate superior performance using combination of incentives, and the need to introduce
organizational changes to facilitate cooperation and peer monitoring in tandem with incentive pay to capture greater
incentive e?ects in team production.
Ó 2008 Elsevier Ltd. All rights reserved.
Introduction
In recent years, ?rms have increasingly turned to
group and team-based incentive compensation to
reward worker productivity (Blinder, 1999; Brown
& Armstrong, 1999; DeMatteo, Eby, & Sundstrom,
1998; Milgrom & Roberts, 1992; Weitzman &
Kruse, 1999).
1
Recent surveys underscore the
importance of team-incentive pay and incentive
plans for teams among ?rms. For example, Brown
and Armstrong (1999) assert that over 50% of major
US and European corporations employ some form
of group- and team-based bonus scheme in their
compensation plans.
Despite its importance, team pay has not received
the attention it should from accounting researchers
and their e?ects on performance are poorly under-
stood (Indjejikian, 1999; Sprinkle, 2003; Young,
1999).
2
Though related literature from other ?elds
0361-3682/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2008.08.004
*
Tel.: +1 806 742 3138; fax: +1 806 742 3182.
E-mail address: [email protected]
1
For further discussion on the role of team-based incentive
plans and group compensation, refer to Blinder (1999).
2
Although various experimental studies in accounting had
examined various issues concerning teams (see, for example,
Rankin, 2004; Young, Fisher, & Lindquist, 1993), and most
recently, other papers have addressed the implications of incen-
tive compensation on group performance (see, Fisher, Pe?er, &
Sprinkle, 2003), little empirical evidence exists documenting the
performance implications of incentive plans for teams in ?eld-
settings.
Available online at www.sciencedirect.com
Accounting, Organizations and Society 34 (2009) 589–618
www.elsevier.com/locate/aos
document improvements in productivity with the
use of ?rm-wide incentives, such as pro?t-sharing
and gain-sharing plans (see Hansen, 1997; Knez &
Simester, 2001; Weitzman & Kruse, 1999), there is
very little research that explicitly addresses the per-
formance e?ects of incentive plans speci?cally
designed for production teams using ?eld data.
Accordingly, the focus of this research study is to
shed new light about the e?ects of incentive plans
for teams in ?eld-settings. Drawing on data from
three production units of a large manufacturing ?rm
(a lock manufacturer) that employs production
teams in its assembly operations, the study investi-
gates how workers react to changes made to an exit-
ing incentive plan and whether the underlying
changes led to improvements in labor productivity
and product quality, and in reducing absenteeism.
More speci?cally, I compare changes in perfor-
mance in these outcomes after this ?rm switched
from an incentive plan that relied on piece-rates to
a two tier incentive plan that rewarded performance
at the team-level and at the plant-level. Under the
new bonus structure, workers were rewarded at
the team-level with a bonus when team output and
quality meets or exceeds a pre-speci?ed target, and
at the plant-level with a gain-sharing incentive plan
that pays quarterly bonuses when production teams
achieve productivity and quality targets for the
entire factory. Additionally, a central feature of
the new incentive plan is that it was implemented
in conjunction with a new management control sys-
tem, which comprised several initiatives intended to
facilitate cooperation, reinforce monitoring, and
encourage peer pressure between workers. These
include the distribution of performance reports to
teams, mechanisms to facilitate the exchange of
information to improve performance, and rules to
punish shirking and uncooperativeness.
This is a rich research setting to explore the
e?ects of team-based incentive plans on perfor-
mance. Of particular interest is the fact that this ?rm
not only made substantive changes to the bonus
structure to its existing compensation plan, but as
mentioned above, it adopted several initiatives in
tandem with the incentives seeking better results.
Given the inherent di?culties underlying team
pay, e.g., the free-rider problem, possible collusion
arrangements, investigating whether such manage-
ment control initiatives when o?ered jointly with
incentive pay are more e?ective in motivating work-
ers to raise performance is particularly important.
Also, this issue warrants further investigation given
the fact that changes in compensation systems sel-
dom occur without other major transformations
within the ?rm. For example, Ichniowski, Shaw,
and Prennushi (1997) and Ichniowski and Shaw
(2003) note that incentive compensation is usually
accompanied with human resources practices to eli-
cit optimal performance from the workforce. None-
theless, the empirical evidence as to how incentive
pay and organizational practices in?uence perfor-
mance is limited. Bonner and Sprinkle (2002) elabo-
rate further on this point when they argue that
research concerning the e?ects of monetary incen-
tives should provide more insight into the mecha-
nisms by which incentives in?uence performance.
Lastly, many ?rms nowadays introduce various
bonus schemes in combination aiming for better
results, but as pointed out by Sprinkle (2003) little
is known as to how combination of incentives a?ects
worker productivity.
It must be noted that my aim is not to test the
optimality of the incentive plan, or make a direct test
of theoretical predictions concerning team-incen-
tives of agency models (Alchian & Demsetz, 1972;
Holmstrom, 1982), or behavioral theories Deutsch,
1949, 1973). Instead, my goals is to broaden our
understanding of howteam-based incentive compen-
sations works in practice, and more importantly,
documents how ?rms can design more e?ective
incentive plans for teams. In this regard, the article
resembles a case study similar in scope to other stud-
ies which had examined the performance e?ects of
incentive pay in ?eld-settings (Banker, Lee, & Potter,
1996b; Hansen, 1997; Knez & Simester, 2001).
After controlling for numerous factors that are
known to in?uence productivity and product qual-
ity in a multivariate regression model, the results
show that by moving from piece-rates to the new
incentives and by adopting the organizational
changes, the factory had a 31% rise in productivity
and a 95% drop in product defects. Moreover, the
factory experienced a substantial decrease in worker
absenteeism. Although it is di?cult to causally link
these improvements speci?cally to higher e?ort,
more cooperation, improved monitoring, and peer
pressure (all outcomes are unobservable latent
constructs), qualitative evidence gathered from
management and workers attributes these improve-
ments to positive changes in worker behavior asso-
ciated with the incentive plan.
This research contributes to the literature in sev-
eral important ways. First, I provide exploratory
evidence on the interplay between incentive pay
590 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
and management control initiatives within ?rms.
These results emphasize a need to jointly consider
combination of incentive pay schemes along with
imposition of control mechanisms to facilitate com-
munication and monitoring between workers as we
design compensation systems for teams. These ?nd-
ings are in line with recent observations in the con-
tracting literature. For example, Bonner and
Sprinkle (2002) and Ichniowski and Shaw (2003)
calls for greater integration between incentive pay
and elements in job design when designing compen-
sation systems to enhance the incentive e?ects of
monetary incentives. Likewise, these ?ndings pro-
vide further evidence about the importance of creat-
ing complementarities between incentive pay and
organizational design as a viable strategy to enhance
organizational performance, coinciding with the
?ndings of theoretical studies (Baker, Jensen, &
Murphy, 1988; Holmstrom & Milgrom, 1994; Mil-
grom & Roberts, 1995). The study may also help
to guide ?rms in the design and implementation of
better incentive plans for teams.
The remainder of the paper is organized as fol-
lows: the next section discusses the research setting
and compares the original and new incentive plan.
Hypotheses development reviews the literature and
develops hypotheses. Data and methodology
describes the sample and the empirical methodol-
ogy. Results are discussed in Empirical results.
The paper concludes with a discussion of the main
?ndings and o?er suggestions for future research.
Research setting
The research site is a large manufacturing facility
operated by a division of a Fortune 500 US Com-
pany. The factory assembles a large assortment of
security metal locks and it is dominant world pro-
ducer of security locks.
3
The site consists of three
production units to serve distinct markets. The
retail unit assembles locks for retail customers, the
institutional unit combination locks for schools,
the military, and sports facilities, and the commer-
cial unit customizes locks according to detailed
customer speci?cations. The plant operates in a
non-unionized environment and employs approxi-
mately 800 production workers.
Before presenting the structure of the original and
new incentive plan, it is useful to describe the nature
of the production process to better understand the
motives of using team-based compensation.
Process technology and production teams
Relevant to our discussion is the fact that the
plant employs production teams and/or production
cells to perform all assembly operations. The vari-
ous assembly tasks required in the fabrication of
locks are performed in their entirety by these teams,
which range in size from 8 to 15 members.
4
All three
units share more or less the same production tech-
nology. The assembly process is primarily labor-
intensive and each member of a team is responsible
for completing a speci?c task. In synthesis, these in-
clude the perforation, attachment, and insertion of
various metal and rubber components inside the
body of a metal padlock, assisted by small machin-
ery and specialized tooling. One important aspect of
the manufacturing process is that these all assembly
tasks are complementary and sequentially interde-
pendent. In this process, each team is responsible
for monitoring quality and in executing machinery
setups between production runs. Teams works un-
der the direction of a production supervisor who
is responsible for monitoring the team’s output
and quality. Hence, given the nature of the assembly
process in which all assembly steps are highly com-
plementary, the ?rm has always used some form of
incentive bonus linked to the team’s performance in
its compensation plan.
Original incentive plan
The original compensation plan consisted of
three elements: base pay, an individual attendance
bonus, and a piece-rate bonus linked to the team’s
output.
5
Base pay consisted of a daily salary, which
varied according to seniority and skill-level. To
reduce absenteeism, which is problematic in the
maquiladora industry, workers received a monthly
attendance bonus. This portion of pay was mainly
3
The facility is located in Northern Mexico and it is part of the
maquiladora industry. Under this type of outsourcing production
arrangement, ?rms import duty-free materials and components
from the US into Mexico for use in the assembly of its products.
After assembly, the ?nished products are re-exported to the US
or to the plant’s parent country of origin.
4
At the time of the study, there were approximately 50
production teams in operation.
5
For ease in exposition, the main elements of the previous and
new incentive plan are summarized here. Further details of the
speci?cs under each plan are provided in the appendix.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 591
?xed and accounted for approximately 70% of a
worker’s total earnings. The remaining portion
was contingent upon each team’s performance.
Each team was rewarded on the basis of piece-rates
and worker pay was proportional to the number of
locks produced daily by each team; resulting cash
bonuses were calculated on a weekly basis and
shared equally among members of a team.
6
A follow up question is why did this ?rm decide to
make changes to the compensation plan? Manage-
ment pointed out several limitations with the existing
plan. First, the consensus was that piece-rates were
only partially e?ective in motivating teams to raise
productivity consistently across all teams in the fac-
tory. Therefore, productivity for the plant was less
than optimal. Although some teams were highly pro-
ductive, in particular teams consisting of more senior
workers were more cohesive and motivated to work
harder to earn higher bonuses, the same level of pro-
ductivity was di?cult to achieve consistently for
many teams. To illustrate, I refer the reader to
Fig. 1, which graphs the factory’s monthly productiv-
ity. Clearly, it can be seen that productivity in the
months preceding the changes to the incentive plan
was low; it averaged around 60%.
The problem with low productivity was further
eroded as a result of high turnover and absenteeism
at the plant. E?ort levels were generally sub-optimal
for teams experiencing high turnover. The frequent
arrival of new workers tended to retard production,
resulting in lower output levels and smaller bonuses
for these teams. As a result, workers were not moti-
vated to work as hard to raise output. Additionally,
it was evident hat the attendance bonus was not suc-
ceeding in reducing absenteeism. Besides harming
productivity, absenteeism also made it di?cult for
teams to adhere to a group norm to monitor and
enforce good behavior among themselves (reduce
shirking, enhance cooperation, etc.). Hence, it was
imperative to ?nd an alternative solution to reduc-
ing chronic levels of absenteeism and turnover.
Another problem experienced with piece-rates is
that it contributed to quality problems. Because
the piece-rate bonus was tied to output, teams
mainly focused at raising output at the expense of
quality and were not as motivated to inspect, pre-
vent, or reduce product defects. Although, quality
was closely monitored and management refused to
pay a bonus for defective units, the number of prod-
uct defects was exceptionally high. To illustrate,
right before the ?rm adopted the new incentive plan,
product defects, measured as parts per million
(PPMs), averaged around 12,000 defects per month.
Lastly, the existing compensation plan appeared
to provide few incentives for teams to cooperate
with one another and share information to improve
the plant’s performance. Several modes of coopera-
tion were seen as critical. First, it was important to
motivate teams to communicate with one another so
they could exchange information regarding process
improvements. For instance, seemingly trivial inno-
vations such opening combination locks faster to
test their functionality or inserting components in
a di?erent way can have a discernible impact on
labor productivity. Management wanted teams to
share this knowledge with their peers; however,
piece-rates provided no incentive for teams to do
so.
7
It was also imperative to motivate teams across
Study Period
Team-level incentive Plant-level incentive
Original Incentive Plan New Incentive Plan
Piece-rates Team output-target scheme Team incentive, plant incentive, mgmt. control initiatives
January 1999
April 2000 July 2000
September 2003
Fig. 1. Implementation timeline of incentive plan.
6
For purposes of the bonus calculation, the piece-rate factor
was tied to the total standard hours of output produced during
the day. Standard hours of output equal the number of locks
produced in a given day times the standard labor time based on-
time and motion calculations required to assemble a given lock
model.
7
While the motives as to why teams did not share information
with one another could vary, a valid explanation is that
oftentimes workers are reluctant to do so for concerns that this
could lead to a rise in performance standards. For instance, a rise
in ?rm-wide productivity could potentially lead to the ratcheting
performance standards in such a way that it lowers their bonuses.
Further discussion of this argument can be found in Bandiera,
Barankay, and Rasul (2005) and Brown and Phillips (1986).
592 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
the three production units to share valuable knowl-
edge to improve the plant’s performance in general.
Teams at each production unit had developed
unique core competencies. For instance, production
teams in the commercial unit were highly skilled in
executing rapid set ups between production runs.
New incentive plan
To address all of the above concerns, the ?rm
restructured the existing compensation plan. Under
the new plan the ?rm wanted to motivate workers to
improve productivity and quality not only through
greater e?ort levels but also through heightened
cooperation at all levels: within each team, between
teams, and across production units. Restructuring
compensation around a single incentive scheme,
such as the existing piece-rate contract, presented
limitations since it neglected inter-team coopera-
tion. Hence, in changing the structure of the com-
pensation package the ?rm decided to adopt a two
tier bonus plan to reward individual team perfor-
mance as well as the combined performance of all
teams.
Under the new pay structure, each worker main-
tained the same level of base pay, but the attendance
bonus and piece-rates were phased out. The piece-
rate contract was replaced by a team-level bonus
and the attendance bonus by a plant-level bonus.
Before describing in detail each bonus scheme, it is
important to highlight several important details
under the new incentive plan. First, the new plan
contain larger variation in incentive pay. Workers
could earn more in total compensation, but because
a larger portion of their earnings now dependent on
the combined performance of all the teams, the level
of risk increased as well, and consequently, workers
could earn less pay. Second, several organizational
control initiatives were instituted to reinforce the
plan. Last, the implementation dates of the team
and plant bonus varied. The team bonus was intro-
duced in April 2000 and the plant bonus in July
2000.
8
Fig. 1 provides a chronology on these events.
I now turn to discuss in explicit detail the bonus
schemes and the organizational control initiatives.
Team-incentive bonus (output-target pay scheme)
In sharp contrast to piece-rates, which rewarded
teams for any positive level of output, the new
team-incentive bonus (hereafter referred as an out-
put-target pay scheme) provides a daily ?xed cash
bonus to each worker only when the respective team
reaches an output quota (a ?xed quantity of locks)
and a quality quota (a minimum number of product
defects). To illustrate, at the start of each workday,
a production supervisor informs each team the
number of locks that must be assembled and a min-
imum threshold of acceptable defects based on the
mix of locks in each production batch, as well as
the supervisor’s expectation of the team’s productiv-
ity. Failure to meet both targets automatically dis-
quali?es the team from earning the bonus. Instead
each worker receives his/her daily wage. Moreover,
after reaching the day’s output quota, teams can
earn an additional bonus for each lock produced
in excess of the set target. The bonus for this extra
output is paid using the same piece-rate factor
employed under the old plan. The ?xed and variable
portions of pay make the bonus scheme a combina-
tion of a budget-based and linear (piece-rate) con-
tract.
9
This sort of bonus scheme has been
referred to as a budget-linear scheme in the compen-
sation literature in accounting (see Fisher et al.,
2003).
Plant-incentive bonus (gain-sharing pay scheme)
The plant-incentive bonus rewards all teams with
a quarterly cash bonus for meeting plant-wide quar-
terly performance targets on productivity and qual-
ity. It is structured in such way that all teams must
coordinate e?orts to achieve the underlying perfor-
mance targets. The amount of bonus is levered so
that the percentage of bonus earned increases as e?-
ciency levels improve. To illustrate, at the com-
mencement of each quarter, workers are presented
with three performance targets with varying degree
of di?culty for each of the two criteria. The amount
of bonus depends on attaining any one of three set
targets for the quarter. For example, if performance
targets for productivity are set at 70%, 80% and 90%
(similar targets are set for minimum number of
product defects), and the lowest target is reached,
8
Management opted to implement the changes to the incentive
plan gradually so its workforce could be more receptive and
productivity would not su?er as much by giving them more time
to adapt to the new system. However, workers were informed
several months in advanced about the changes that were going to
occur to the compensation plan.
9
Management pointed out that the intention of maintaining a
portion of the piece-rates contract under this budget-scheme was
to stimulate teams (especially the most dynamic) to raise output
during periods of high demand without ratcheting output
standards which can potentially demoralize and discourage
workers.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 593
then each worker receives a cash bonus of 6% of the
accumulated quarterly earnings. By reaching the
middle and upper targets, workers are entitled to
an 8% and 10% bonus, respectively. No bonus is
paid when the plant fails to meet the lowest set tar-
gets. As noted previously, the plant bonus replaced
an individual attendance bonus; therefore, an addi-
tional provision under the plant bonus is that work-
ers must maintain near perfect attendance during
the quarter to be entitled to the bonus.
Organizational control initiatives to reinforce the new
incentive plan
Several organizational control initiatives were
instituted as part of the new incentive plan to facil-
itate the exchange of information, cooperation
between teams, and monitoring within and between
teams. First, to strength the team bonus, each team
now is provided with feedback of their performance.
Speci?cally, teams are given a report that depicts the
actual number of locks and the number of product
defects accumulated throughout the day. The
intended purpose is twofold. Teams can assess
how well they are doing relative to the standards.
If the team trails the output and quality quotas,
workers can either intensify e?ort levels, or help
one another to meet their quotas. So, the intended
purpose is to stimulate e?ort and intra-team
cooperation.
Moreover, this mechanism could reduce the
extent of free-riding by way of mutual monitoring.
To illustrate consider this example. Suppose that a
particular production team reaches its output and
quality targets on a regular basis, and neither target
is ratcheted-up in subsequent periods. If on a given
day, this team signi?cantly trails the output-target,
absent external or extraordinary circumstances in
the assembly process (e.g., material shortages, bot-
tlenecks, etc.), it would indicate that the team’s
e?ort levels are not as intense as in previous periods,
signaling shirking. Hence, reporting each team’s
performance on a timely basis induces workers to
monitor each other’s e?orts more rigorously, allow-
ing them in this way to minimize shirking.
The second organizational initiative consisted of
providing feedback on performance to the entire
factory. A weekly report is generated showing the
performance of each team on several key metrics:
productivity, product defects, material waste, the
number of production orders completed on-time,
absenteeism, and turnover. This information is dis-
seminated publicly to all teams. The intended pur-
pose is to help promote cooperation and
information sharing across the three units. For
example, low-performing teams could identify bet-
ter performers and seek help so their performance
can improve. Similarly, high performers can identify
low performers and volunteer to help them improve.
In addition, the distribution of performance reports
could facilitate monitoring among teams as a way to
control the free-rider problem by way of peer pres-
sure as discuss in more detail in the hypotheses
section.
A third and ?nal mechanism was instituted to
punish de?cient worker behavior, such as shirking
or uncooperativeness. Based on a team’s suggestion,
workers performing sub-optimally with the rest of
the team can be dismissed after repeated warnings
and eventually be laid-o?. Previously, shirking was
not disciplined. A potential e?ect of this mechanism
is that it could also help to instigate peer pressure
within each team.
Hypotheses development
This section develops hypotheses concerning the
sensitivity of worker productivity and product qual-
ity to the new incentive plan and describes how the
various elements on the plan could have led to
improvements in productivity and product quality,
as well as a reduction in worker absenteeism. An
important observation is that given the fact that
the new incentive plan comprised a combination
of incentive schemes which were introduced at dif-
ferent dates, plus the plan was implemented in con-
junction with the other initiatives of organizational
control, it makes it nearly impossible to make test-
able predictions on the e?ects on productivity and
quality attributed speci?cally to each element in
the incentive plan. Therefore, my predictions are
based on how the various elements in the plan could
have impacted productivity and quality as a whole
rather than individually. My predictions are guided
by studies in the compensation literature in account-
ing and labor economics, as well as in organiza-
tional behavior.
A starting point in our analysis is the switch from
piece-rates to the output-target scheme. Building on
prior research ?ndings, one might expect that work-
ers would be motivated by an underlying piece-rates
contract and work hard to earn large bonuses. For
example, numerous studies have documented that
piece-rates provide strong incentives for workers
to raise output and productivity (Lazear, 2000;
594 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Paarsch & Shearer, 2000; Shearer, 2004; Stiglitz,
1975). However, in team production, an output-tar-
get (or budget-based) scheme should be superior to
piece-rates in raising worker e?ort. The reasoning
behind this argument is that if every worker can
be punished with a low (penalty) wage when team
output or product quality falls below some target,
then su?cient incentives are generated for all work-
ers to work hard enough to hit those targets and to
monitor each other’s e?orts more closely to reduce
shirking, otherwise the entire team forfeits the
bonus. Simply put, the fact that a team could be
penalized for not meeting those targets becomes a
self-enforcing norm in a team. In contrast, with
piece-rates, workers know that they would be enti-
tled to receive some portion of the bonus as long
as the team generates output. Therefore, they may
not be as motivated to give their maximum e?ort
nor to police themselves as tightly, thus making
piece-rates more susceptible to the free-rider
problem.
10
Consistent with this view, Holmstrom (1982) pro-
posed the idea that an output-target (or budget-
based) scheme as a Nash equilibrium solution to
help mitigate the free-rider problem and to motivate
higher e?ort in his seminal paper on team-incentives.
He notes that the sharing of total output (most likely
a bonus) among team members will result in shirk-
ing. Incorporating the threat of a team-imposed pen-
alty for output below a target (the team shares less
than 100% bonus) would not yield an improvement
because team members would simply ignore the pen-
alty ex-post. Instead, incorporating a principal-
imposed penalty with a low (penalty) wage for out-
put below a target provides an incentive for workers
to work harder and to police themselves more strin-
gently. Empirical support for this argument has been
found by several studies which have documented an
advantage of output-target (or budget-based)
schemes over piece-rates in raising productivity in
teams (Fisher et al., 2003; Nalbantian & Schotter,
1997; Petersen, 1992).
Also, previous studies have documented that
piece-rates may discourage cooperation and often-
times may put workers (or teams) in competition
with one another (Brown & Phillips, 1986; Drago
& Garvey, 1998).
11
In contrast, other bonus schemes
such as gain-sharing bonuses should enhanced coop-
eration. Expanding on this point, the organizational
behavioral literature suggests that gain-sharing com-
pensation transforms a workforce, wherein workers
begin to interface in the whole organization and ulti-
mately leads to greater commitment and coopera-
tion to achieve performance goals Bullock and
Lawler, 1984; Schuster, 1983. ; Schuster, 1983). Sev-
eral ?eld studies have documented a rise in produc-
tivity as well as quality with gain-sharing through
heightened worker cooperation, information shar-
ing, and greater commitment from workers (Bullock
& Lawler, 1984; Schuster, 1984; Welbourne, Balkin,
& Gomez-Mejia, 1995). So, in this respect, the plant
‘gain-sharing’ bonus scheme should also lead to a
rise in productivity.
Lastly, recall that a primary di?erence between
the original and new incentive plan was larger vari-
ation in incentive pay. The new plan shifted a signif-
icant portion of a worker’s total earnings to the
combined performance of all teams in the factory.
In other words, workers could earn more in total
compensation, but their level of risk increased since
a larger amount of their earnings now were depen-
dent on the performance of all the teams, and there-
fore, could earn less pay as well. Based on the
conventional predictions of principal-agent theory
(Baiman, 1982; Holmstrom, 1979)
12
the notion that
individual incentive (risky) pay is associated with
positive e?ects on performance and/or worker pro-
ductivity is well documented in empirical studies
both in experimental and ?eld-settings (see Banker
et al., 1996b; Chow, 1983; Prendergast, 1999).
Hence, it is reasonable to assume that incentive
pay could as well drive strong incentive e?ects in
10
More formally, the fact that all workers bene?t from the e?ort
of fellow co-workers, team members may not be as motivated to
increase e?ort if increases in e?ort are re?ected by increase in pay
only on the order of 1/n (n = team size). All workers share in the
bene?t created from the extra e?ort of a particular worker.
Unless the marginal bene?t of an extra unit of e?ort exceeds the
marginal cost of working, the individual incentive to expend more
e?ort under team piece-rates is minimal.
11
As noted previously (refer to footnote 7) workers are often
reluctant to cooperate with each other when rewarded with piece-
rates for concerns that improvements in overall performance
could lead to a rise in performance standards. For instance, a rise
in ?rm-wide productivity could potentially lead to the ratcheting
performance standards in such a way that it lowers their bonuses
(Bandiera et al., 2005; Brown & Phillips, 1986).
12
Within the conventional framework of principal-agent theory,
the idea is that incentive (contingent) pay imposes more risk on
the workers. As a result, there should be more motivated to work
harder. This is also referred as risk-sharing in the contracting
literature. This issue has long been recognized in the contracting
theory in economics (Holmstrom, 1979) and accounting (Baiman,
1982). For further insight about the link between incentive pay
and economic performance, see the work of Prendergast (1999).
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 595
team production settings just as it has been demon-
strated in the case of individual incentives. Because
the possibility of earning lower pay is now at stake,
I hypothesized that the overall shift in incentive pay
under the new incentive plan should motivate work-
ers to work harder to reach the various performance
targets and these e?orts should lead to higher pro-
ductivity. Based on all the above arguments, I o?er
the following prediction:
H1. Ceteris paribus, productivity should rise after
the implementation of the new incentive plan.
Now, turning over to product quality, my predic-
tion is that the switch in incentives should also lead
to a rise in product quality. First, there is plenty of
empirical evidence documenting that the use of
piece-rates is associated with output of lower qual-
ity. Given the high emphasis that piece-rates place
on production, prior research found that workers
trade-o? output for quality under piece-rates
(Baker, 1992; Holmstrom & Milgrom, 1994; Lazear,
1986, 1991, 2000; Paarsch & Shearer, 2000; Stiglitz,
1975).
13
According to Lazear (2000, p. 1358) ‘‘one
defect of paying piece-rates is that quality may suf-
fer”. Expanding on this same point, Paarsch and
Shearer (2000, p. 61) note that, ‘‘piece-rates have
the advantage of providing incentives for workers
to work hard, but piece-rates also reduce the incen-
tive to the worker of producing quality output”.
Although the previous piece-rate contract
imposed a penalty for defective output, the penalty
was not as severe as the one imposed under the out-
put-target scheme. With piece-rates, teams were
penalized solely on the defective units found in a
particular production batch and could still receive
a bonus on the remaining ‘good’ output produced
during the day. As a result, the teams most likely
ignore the penalty ex-post and continue to overlook
quality. In contrast, under the output-target
scheme, the penalty for defects is much severe. If a
team does not reach the minimum threshold of
defects, the team loses the entire bonus regardless
of the quantity of output produced during the
day. So, the penalty function makes it more costly
for workers to miss the quality target. As a result,
teams should place more emphasis in reducing mis-
takes during assembly and also they should inspect
the quality of their output more carefully.
Furthermore, by invoking the same arguments
discussed above in the case of productivity, I also
predict that the adoption of the gain-sharing bonus
scheme and the shift in incentive pay should also
lead to a rise in product quality. Thus, based on
all the above arguments, I hypothesized that:
H2. Ceteris paribus, product quality should rise
after the implementation of the new incentive plan.
I now turn to discuss how the various organiza-
tional control changes complement the incentives.
Ichniowski and Shaw (2003) argue that incentive
pay will be more e?ective when ?rms also provide
other organizational initiatives (i.e. human resource
practices). In the case of team production, under-
taking cooperation and enticing workers to share
valuable information within and between teams
involves other mechanisms that would help facili-
tate communication and the exchange of informa-
tion between workers besides incentive pay. This
argument is consistent with the observation of Ran-
kin (2004), who asserts that cooperation in team
production settings requires a rich informational
environment. Also, Bandiera, Barankay, and Rasul
(2004) claim that one chief determinant for cooper-
ation is the ability of workers to monitor each
other’s performance. Therefore, if the production
setting lacks the infrastructure that enables workers
to communicate and interact with one another, and
to e?ectively monitor each other, cooperation
would be di?cult to achieve.
Additionally, a central observation of many theo-
retical studies is that worker monitoring can act as a
powerful mechanism to control the free-rider prob-
lem. But, the ability to monitor co-workers may be
greatly in?uenced by elements in job design and
workplace characteristics. As emphasize by Kandel
and Lazear (1992, p. 806), if workers do not have
the means to monitor each other e?ciently, or in
the absence of appropriate mechanisms to enforce
‘good behavior’, worker monitoring may fail in con-
trolling the free-rider problem.
As noted previously, the incentive plan was intro-
duced in combination with two initiatives designed
to facilitate intra- and inter-teams monitoring. So,
I expect such initiative to facilitate cooperation.
The reasoning behind this argument is that perfor-
mance feedback acts as a substitute for direct obser-
13
The idea that piece-rates lead to poor quality is not entirely
new; the problem lies when only some attributes of output can be
measured accurately, workers’ attention will be diverted to the
attributes that are rewarded in the compensation formula, such as
output. Additionally, as noted by Lazear (1991), much of this
problem has to do with full observability of output. For further
discussion as to how piece-rates could lead to lesser quality, see
Lazear (1986, 1991) and Stiglitz (1975).
596 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
vation and permits workers and teams to monitor
their progress and assess their own performance rel-
ative to their peers. With this information, teams
can seamlessly identify other teams that are per-
forming sub-optimally relative to the performance
standards set for the plant and o?er their help to
help them improve.
Furthermore, besides creating a richer informa-
tional environment, performance feedback should
also facilitate the rate of learning and strategy devel-
opment across production cells (Boning, Ichniow-
ski, & Shaw, 2007; Sprinkle, 2000). As noted by
Boning et al. (2007) as they examine the e?ects of
problem solving teams, group pay and productivity
in the steel industry, ‘‘problem solving e?ort refers
to developing and implementing ideas for improv-
ing production operations, and is based on the idea
that operators with hands-on experience on a pro-
duction line are in a unique position to devise rem-
edies for idiosyncratic problems. . .”. So, I expect
that the dissemination of performance information
should motivate more senior operators to transfer
valuable knowledge with co-workers. If this knowl-
edge is shared, productivity and quality improve-
ments should continue gradually after the
implementation of the plant bonus scheme.
Also, the body of the existing literature on incen-
tives for teams provides a general framework for
ways to minimize the free-rider problem in team
production, emphasizing that peer pressure can be
an e?ective control mechanism (Barron & Gjerde,
1997; Che & Yoo, 2002; Kandel & Lazear, 1992;
Macho-Stadler & Perez-Castrillo, 1993; Sewell,
1998). Thus, I posit that the public dissemination
of performance should also aid in controlling the
free-rider problem via peer pressure. For example,
with access to teams’ performance, teams can
appraise whether their peers are working with simi-
lar intensity to achieve the performance targets in
productivity and quality for the plant. Top perform-
ers could then exercise external pressure (e.g., public
shame, scrutiny) on slackers to raise their perfor-
mance to a level commensurate with the rest of
the plant.
Lastly, the old incentive plan lacked mechanisms
to properly identify or eliminate de?cient worker
behavior; so workers had little incentive to apply
pressure on co-workers. In contrast, the new incen-
tive system instituted a formal mechanism to punish
shirking and uncooperative behavior—the possibil-
ity of being terminated from the job. As a result, I
posit that the threat of dismissal is a powerful solu-
tion that may be levied against shirking and unco-
operativeness through peer pressure as well.
To summarize the above arguments, the new
incentive plan provided mechanisms to help facili-
tate cooperation, mutual monitoring, instigate peer
pressure, and promote the rate of learning. It also
provided workers with the means to punish de?cient
behavior (i.e., shirking, uncooperativeness). There-
fore, I hypothesized that these outcomes should
motivate teams to continue making gradual
improvements to productivity and product quality.
Hence, the following hypothesis is presented.
H3. Ceteris paribus, productivity and quality
improvements should continue and persist over
time following the implementation of the new
incentive plan.
My last set of predictions concerns absenteeism
and turnover. Recall that the plant bonus scheme
made it mandatory for workers to have near perfect
attendance to be entitled to receive a bonus. It may
be argued that in some sense the attendance bonus
and the provision under the plant bonus are substi-
tutes. Workers are penalized under either incentive
contract: missing a day of work precludes workers
from earning the bonus in either case. However,
the di?erence between the new incentive plan is that
the absence penalty function is steeper (more severe)
given the fact workers could lose a bigger paycheck
in the form of a larger quarterly bonus. Consistent
with this view, prior research has shown that atten-
dance bonus plans are more e?ective in curtailing
absenteeism when steeper penalties are imposed on
workers (Allen, 1981; Brown, Fakhfakh, & Ses-
sions, 1999). Based on this argument, I make the
following prediction:
H4. Ceteris paribus, the implementation of the new
incentive plan leads to a reduction in absenteeism.
Lastly, it is possible that the new incentive plan
could help in sorting out the workforce, and there-
fore, create a potential selection e?ect. That is, it
could have produced a selection e?ect in such a way
that the less productive workers leave the ?rm dislik-
ing the changes to the compensation plan and the
more productive remain or are attracted to the ?rm
as documented by several studies in the compensa-
tion literature (Banker, Lee, Potter, & Srinivasan,
2000; Chow, 1983; Waller & Chow, 1985). To inves-
tigate this possibility, I examine worker turnover be-
fore and after the incentive plan was adopted. No
speci?c prediction is made as to how the incentive
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 597
plan a?ects worker turnover. However, if the rate of
voluntary turnover is statistically higher in the early
months after adoption of the new incentive plan and
levels-o? in subsequent months relative to the old
incentive plan, this could possibly be an indication
of a selection e?ect in the workforce.
Data and methodology
Data and sample
The data were drawn from the three production
units at the factory.
14
The initial sample covers the
period of January 1999 through September 2003.
However, as it is characteristic of most ?eld studies
there were data limitations. Mainly, product defects
and several of the control variables were not fully
available for the entire period, thereby forming an
unbalanced panel of data. Uninterrupted time-series
data of all variables of interest is available from
September 1999 through September 2003. There-
fore, the study period covers this period. Also, be-
cause most data is reported on a monthly basis,
my main regressions are estimated using monthly
observations. In the case of productivity, an exten-
sion of the tests is supplemented with weekly data
in a reduced regression model. The ?nal sample con-
sists of a longitudinal cross sectional sample of 147
monthly observations; there are 49 observations for
each of the three production units.
Regression models and measurement of variables
To provide a thorough set of controls in my regres-
sions, I follow Hayes and Clark (1985) and include a
broad set of factors, which tend to a?ect labor pro-
ductivity in manufacturing operations. These include
production headcount, absenteeism, turnover,
worker training, overtime, engineering modi?cations
to the assembly process, plus other variables speci?-
cally related to the research site. Process engineers
were consulted to ensure that these variables ade-
quately addressed the corresponding production
technology. Because some of these factors undoubt-
edly a?ect other dimensions of manufacturing perfor-
mance, including product quality (see Banker, Field,
Schroeder, & Sinha, 1996a), several of these variables
are also used as controls on the regressions on prod-
uct quality. Further, to eliminate potential omitted
variables biases, I investigated whether any major
changes or relevant events occurred at the plant dur-
ing the study period. By the time the changes to the
compensation plan took e?ect the facility was fully
stable and mature and no other major events (e.g.,
management changes, adoption of new technology,
or equipment changes) other than the changes to
the incentive plan were introduced.
To assess the impact of the new incentive plan on
productivity, I estimate the following OLS regres-
sion model pooling observations from the three pro-
duction units:
PRODUCTIVITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
DEFECTS
it
þ b
4
HEADCOUNT
it
þ b
5
TURNOVER
it
þ b
6
ABSENTEEISM
it
þ b
7
OVERTIME
it
þ b
8
LAG TRAINING
itÀ1
þ b
9
ECO
it
þ b
10
RETAIL
it
þ b
11
INSTITUTIONAL
it
þ e
it
ðM1Þ
And, the following model is estimated on product
defects and material waste to measure the impact on
product quality:
QUALITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ b
8
RETAIL
it
þ b
9
INSTITUTIONAL
it
þ t
it
ðM2Þ
where subscript i refers to production unit i and t to
each month observation. An explanation of each
variable immediately follows. A similar but a more
parsimonious model with fewer control variables is
used to test the incentive plan e?ects on absenteeism
and turnover.
14
Data to calculate proxies for productivity, product quality,
and remaining variables of interest were hand-collected from
various departmental records, including accounting, production,
process-engineering, quality, and payroll. Additionally, I
obtained the opinions of management and production workers
regarding the impact of the incentive plan on plant performance
and held various discussions with process engineers to understand
the manufacturing process and to investigate other relevant
events that could have potentially confounded the impact of the
incentive plan on performance.
598 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Dependent variables
For consistency and to reduce potential biases in
the analysis, I use the same productivity and prod-
uct quality measures used by the plant as my main
proxies for productivity and quality. PRODUC-
TIVITY is de?ned as the ratio of standard labor
hours based on the actual output to the total
amount of labor hours spent in production. This
is equivalent to labor e?ciency variance expressed
in ratio form (Horngren, Datar, and Foster, 2002).
The metric excludes production hours unrelated to
the normal course of production, such as loss time
due to machinery breakdowns, material shortages,
and scheduling problems; these hours are excluded
from the denominator. Additionally, the ratio
excludes standard labor hours related to defective
units in the numerator; that is, labor hours worked
and earned but unaccounted for when units are
rejected for not complying with standards of
quality.
The proxy for product quality is the number of
product defects (DEFECTS) expressed as parts per
million (PPMs).
15
To normalize the data, this vari-
able is transformed by taking the natural log. The
second proxy for quality is labeled as WASTE and
it represents the dollar amount of material waste
de?ated by production volume. Just as in the case
of productivity, these are the chief measures used
by the factory to track product quality.
An important issue in the estimation of the
empirical model is whether the dependent variables
should be estimated as a level variable, or as a
change and/or percentage change variable. This
issue merits further discussion. As noted by Banker
et al. (2000, p. 82), accounting research has relied on
both, level and return or change models, to examine
the association between variables of interest and
?rm performance. The decision to use one measure
relative to the other depends on the research ques-
tion in place. In the context of compensation
research, the use of a level variable can be supported
if the treatment variable (i.e., pay) has a permanent
impact on performance measures. If the impact is
expected to be gradual over time, such as in the case
of this research site given that the changes to the
incentive plan occurred in stages, then the percent-
age change or di?erenced models could be more
suitable. Therefore, I present results based on three
speci?cations of each model: one using level vari-
ables, level changes, and percentage changes in pro-
ductivity, defects, and the remaining treatment
variables.
Main independent variables
Following a similar research approach to Banker
et al. (1996a), I regress each performance outcome
on a time trend to capture the mean changes in pro-
ductivity and product defects under the periods of
the old and new incentive plan. More precisely,
the variable OLDPLAN_TREND represent a linear
time trend for the period under the old incentive
plan (September 1999 to March 2000); it is mea-
sured as the number of months since the beginning
of the study period, September 1999, through
March 2000, and assumes a value of zero thereafter.
Likewise, NEWPLAN_TREND represent a linear
time trend for the period in which all elements in
the new incentive plan, the team bonus, the plant
bonuses, and the organizational control initiatives,
are operating altogether (July 2000 to September
2003); it is measured as the number of months since
July 2000 through September 2003, and assumes a
value of zero for the preceding months prior to the
introduction of the teambonus. Apositive and larger
coe?cient estimate on NEWPLAN_TRENDrelative
to the OLDPLAN_TREND coe?cient will indicate
that productivity is greater for the period of the new
incentive plan. Conversely, in the regressions on
product defects and material waste, a negative and
smaller estimate on NEWPLAN_TREND relative
to the OLDPLAN_TREND coe?cient would indi-
cate that product defects and waste is lower under
the new incentive plan.
It must be noted that the variable NEW-
PLAN_TREND tests for the hypothesized e?ects
of the incentive plan by measuring the performance
impact of the output-target and gain-sharing bonus
scheme, and the organizational initiatives, jointly.
The close window of separation between the
schemes’ implementation period makes it di?cult
to separate cleanly the performance impact associ-
ated across each incentive. For instance, potential
improvements to productivity and quality associ-
ated with the output-target scheme are re?ected in
future periods and such improvements may be
picked up by the gain-sharing scheme. However,
combining both incentives schemes and the organi-
zational control initiatives re?ects management’s
intention of treating all them as part of the same
incentive plan.
15
To arrive at this number, production batches are inspected at
random and defective units within a batch are rejected when the
batch fails to meet standards of quality.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 599
Control variables
The variables HEADCOUNT (the total number
of production workers), TURNOVER (the rate of
employee turnover) and ABSENTEEISM (the rate
of absenteeism) are used in both regressions to con-
trol for e?ects related to changes in the workforce.
HEADCOUNT controls for ?uctuations in produc-
tion headcount. While an increasing (decreasing)
trend in production headcount might yield an
increase (decrease) in labor productivity, no predic-
tion is made ex ante about the expected sign of this
coe?cient since a larger number of production
workers might yield an increase in total standard
hours (numerator) but also increase total produc-
tion hours (denominator) in the productivity ratio.
Both ABSENTEEISM and TURNOVER should
impact productivity and product quality adversely
due to a reduction in manpower. Therefore, I expect
a negative (positive) relationship between TURN-
OVER and labor productivity (defects) and the
same relationship with ABSENTEEISM.
The variable LAG_TRAINING is a 1 month lag
variable representing the total number of hours of
training received by production workers. The vari-
able includes both hours related to inductive train-
ing for new employees, as well as training hours
directed toward teaching current employees tech-
niques to make improvements to the manufacturing
process. The impact of training on productivity is
ambiguous. While one might expect employee train-
ing to impact positively labor productivity given
than it enables workers to acquire new knowledge
to improve their performance, the counter argument
is that hours spend in training is time diverted from
production. Also, a signi?cant amount of training
hours re?ect new employee training that may not
add incremental value. On the other hand, though
an immediate positive impact may not be realized,
these short-term losses in labor productivity may
be o?-set by future gains that are driven by new
knowledge at improving the manufacturing process.
Hence, given both arguments, I do not predict a pri-
ori the direction of such relationship.
OVERTIME represents total hours of overtime
and it controls for the e?ect of overtime in the pro-
ductivity model. ECO (the number of engineering
change orders) is incorporated on both regression
models to control for modi?cations in either prod-
uct design, material components, or in the assembly
process. Because these changes tend to improve
product functionality, quality, or reduce complexity
during assembly, the related coe?cient estimate
should be positively (negatively) associated with
productivity (defects). Also, many of the engineer-
ing changes originate on the production ?oor based
on a suggestion from a production worker. Thus, I
expect a signi?cant rise in the number of engineering
changes in the post-incentive plan implementation
period.
Because it is possible that the extent of product
quality could have a direct impact on labor produc-
tivity, I also included my main proxy for quality,
DEFECTS, in the estimation on productivity. The
rationale is that if fewer mistakes are made during
the assembly process, this translates into more pro-
ductive time for workers, which ultimately lead to
more output. Conversely, poor quality could have
an adverse impact on productivity due to the fact
that product defects tend to slow down a manufac-
turing process. For example, workers may divert
production time to reworking or inspecting units
(Cachon & Terwiesch, 2006).
16
Therefore, I predict
a negative relationship between defects and
productivity.
Based on the recommendations of quality engi-
neers, I also included an additional variable in the
regressions on defects and material waste to control
for the level of quality of material components used
in production. The variable DEFECTS_INCOM-
ING represents the amount of defective material
components detected during a random sample qual-
ity inspection as these materials arrive at the plant.
Since defective components are identi?ed prior to
assembly, in part this should help prevent product
defects and material spoilage. Hence, the estimated
coe?cient should be negative on both, the product
defects and material waste regressions.
Lastly, although the three production units share
similar production technology, due to the large mix
of products, the degree of assembly complexity
could vary among the units. Accordingly, to control
for di?erences in the assembly process that could
potentially impact productivity and quality, the
variables RETAIL, INSTITUTIONAL, and COM-
MERCIAL, represent dummy variables to control
for ?xed e?ects across the three production units.
17
16
According to the operations management literature, poor
quality is inversely related to how fast subassemblies ?ow in a
manufacturing process. Therefore, more defects, reworking units,
and waste generally slows down a process, and thus impacts
adversely labor productivity.
17
The dummy variable on the commercial unit (COMMER-
CIAL) is captured in the intercept of each regression model.
600 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Empirical results
Table 1, panel A, summarizes descriptive statis-
tics for the sample, showing means, medians, and
standard deviations for key variables of interest.
The mean (median) monthly productivity across
production units is 82% (88%), while the mean
(median) product defects is 1980 (279). A closer
look at the data indicates large variation for both
variables, with a standard deviation of 17% and
4302 for productivity and defects, respectively.
Mean and (median) for absenteeism and turnover
are 1.3% (0.90%) and 3.8% (2.1%), respectively.
Panel B reports tests of mean di?erences (Wilcoxon
Table 1
Descriptive statistics of variables of interest monthly observations for the period of September 1999 through September 2003
Variable Mean Standard deviation Minimum Median Maximum
Panel A: pooled sample (N = 147)
Labor productivity (%) 82.4 16.5 29.8 88.1 106.2
Product defects 1,980 4,302 0 279 20,693
Headcount 193 44 61 199 288
Turnover (%) 3.8 4.2 0 2.1 16.2
Absenteeism (%) 1.3 0.95 0 0.90 4.3
Overtime hours 2,022 3,167 0 857 20,191
Training hours 340 281 35 226 1103
Production volume (units) 989,197 432,672 127,679 932,414 2,148,570
Engineering change orders 2 2 0 1 10
Material defects incoming 4224 6132 0 2300 37,607
Material waste ($) 6348 3666 1070 5610 21,038
Standard labor hours 29,265 9462 2275 30,210 47,426
Production labor hours 35,342 9607 7624 35,894 61,481
Total labor cost ($) 170,295 115,606 33,119 127,631 516,072
Variable Median before Median after Di?erence
Panel B: di?erences in medians before and after changes to the incentive plan
Labor productivity (%) 59.3 89.9 30.6
***
Product defects 9244 186 À9058
***
Headcount 198 201 3
Turnover (%) 8.2 1.9 À6.3%
***
Absenteeism (%) 2.0 0.7 À1.3%
***
Overtime hours 2568 786 À1782
***
Training hours 107 248 141
**
Production volume (units) 953,186 988,324 35,138
Engineering change orders 1 2 1
Material defects incoming 6598 2035 À4563
***
Material waste ($) 3742 5583 1841
**
Standard labor hours 20,912 31,063 10,151
***
Production labor hours 31,207 36,129 4922
Total labor cost ($) 126,631 145,898 19,567
***
*,**,***
denotes statistical signi?cance at the 0.10%, 0.05%, and.01% levels, respectively. Descriptive statistics calculated using monthly
observations for each manufacturing unit for the period of September 1999 through September 2003. There are 49 monthly observations
for each production unit. Test of di?erences in median for the before and after period is performed using Wilcoxon rank sum test with
monthly observations for each production unit for the period of September 1999 to September 2003. The before period contains monthly
observations for the period of September 1999 through March 2000. The after period contains monthly observations from July 2000
through September 2003. Therefore, the test is performed eliminating the observations from April 2000 through June 2000, the period in
which the team-level (output-target) incentive scheme operates in isolation. Variables de?nitions: Labor productivity: total standard hours
divided by total production labor hours; Product defects: total sum of ?nished product defects (parts per million); Headcount: total number
of production workers; Turnover: total number of voluntary/involuntary resignations divided by the average production headcount;
absenteeism: total number of absences divided by the total number of worked days; overtime hours: total number of production overtime
hours training hours: total number of hours of worker training; volume: total units of output; engineering change orders: total number of
manufacturing process changes; material defects incoming: total number of material component defects during inspection upon arrival at
the factory; material waste: total amount of material waste in US dollars; standard labor hours: standard labor time per each unit of output
multiply by the number of units manufactured; production labor hours: total number of hours spent in production; total labor cost: total
direct labor cost, including salary and related cash bonuses.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 601
Signed-Rank test) on each variable for the periods
before and after the changes to the incentive plan
are introduced. Signi?cant changes are observed in
all key performance metrics: a mean rise of 30.6%
(p-value = .001) is observed in labor productivity,
and a mean decline of 9058 (p-value = .001) in prod-
uct defects. Similarly, both turnover and absentee-
ism exhibit a statistically signi?cant decline of
6.3% and 1.3%, respectively. These ?ndings provide
preliminary support of improvements in these out-
comes after the new incentive plan is introduced.
Table 2 reports Pearson correlations coe?cients
between the regression variables. An interesting pat-
tern is revealed in the strong negative correlation
(À.58) between productivity and product defects,
corroborating the notion that quality impacts pro-
ductivity. Productivity is also strongly negatively
correlated, as predicted, with turnover (À.71) and
absenteeism (À.77). Likewise, there is strong corre-
lation between product defects and turnover (.33),
absenteeism (.50), and defects incoming (À.70). A
high degree of correlation is also observed among
several of the control variables, which may indicate
a problem with multicollinearity. Therefore, I check
for the presence of it using variance in?ation factors
and condition index analyses and found no signi?-
cant threat of multicollinearity.
Regression results on productivity and quality
Hypotheses one and three predict that productiv-
ity should rise and improve over time, respectively,
after the implementation of the new incentive plan.
Table 3 presents regressions results on productivity
for the pooled sample (panel A) and by production
unit (panel B). The parameter of interests through-
out is b
1
and b
2
, which represents the respective time
trend for the old and new incentive plan, respec-
tively. Each coe?cient captures the change in pro-
ductivity over time. A comparison of these two
parameters will tell us whether the new incentive
plan is associated with any changes in productivity
across time. As shown, the coe?cient estimate
NEWPLAN_TREND is positive and statistically
signi?cant at the 1% level in the regression on level
productivity (b
2
= 0.003), change in productivity
(b
2
= 0.004), and percentage change in productivity
(b
2
= 0.006), whereas the OLDPLAN_TREND
coe?cient is negative and statistically signi?cant
on the level productivity model (b
1
= À0.119) and
insigni?cant in the change and percentage change
models. An F-test of di?erence among the two coef-
?cients show that the di?erence is signi?cant at the
1% level in the level model, and at the 5% level in
the percentage change model.
The explanatory power across the three models is
high, with adjusted R
2
of 70% in the case of the per-
centage change model and 85% in the level model.
Note further that the several of the factors included
as controls have the predicted sign and are signi?cant
at the 5% level or better. In particular, the coe?cient
estimates for TURNOVERand ABSENTEEISMare
both negative and statistically signi?cant. These ?nd-
ings suggest that absenteeism and turnover have an
adverse impact on productivity. Another interesting
result is that the coe?cient for DEFECTS is negative
and signi?cant (b
3
= À0.191, p-value < .05), corrob-
orating the fact that product defects have a negative
impact on productivity.
Results on productivity by production unit are
reported in panel B. Though the number of monthly
observations (sample size) dropped substantially,
the results are remarkably similar with those in the
pooled sample. The NEWPLAN_TREND coe?-
cient is positive and statistically signi?cant at the
conventional level for two of the three production
units. In contrast, the OLDPLAN_TREND coe?-
cient is either negative and signi?cant or statistically
insigni?cant. Likewise, b
2
is statistically and signi?-
cantly greater than b
1
at both models for the institu-
tional and retail units.
To ensure robustness of these results, I also esti-
mated a similar but a more parsimonious model
with fewer controls using weekly data for the same
time period. These results are not reported in tables;
however, they are qualitatively similar to the ones
reported on Table 3.
18
Taken all together, these ?ndings show that the
adoption of the new incentive plan led to a rise in
productivity and to gradual improvements in pro-
18
These results are not sensitive to changing the date of the time
trend variable for the new incentive plan, NEWPLAN_TREND.
Even when we allow the e?ect of the date to vary from July 1,
2000 back to April 1, 2000 the date of adoption of the output-
target scheme, the results are unchanged. Also, in a regression
model not reported in tables, I estimated this regression with
three time trend variables to represent the periods under piece-
rates (September 1999 to March 2000), the output-target scheme
(April 2000 to June 2000), and the period when all elements in the
incentive plan operate altogether (July 2000 and thereafter). This
speci?cation gives somewhat qualitatively similar results. How-
ever, because there are not su?cient time-series observations
surrounding the output-target scheme, the power of the statistics
tests are limited for the time trend coe?cient on the output-target
scheme.
602 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 2
Pearson correlation coe?cients of regression variables pooled sample (N = 147)
PRODUCTIVITY DEFECTS HEADCOUNT TURNOVER ABSENTEEISM OVERTIME TRAINING VOLUME ECO INCOMING WASTE
PRODUCTIVITY 1.00
DEFECTS À0.58 1.00
HEADCOUNT 0.21 À0.03 1.00
TURNOVER À0.71 0.33 0.09 1.00
ABSENTEEISM À0.77 0.50 À0.03 0.68 1.00
OVERTIME À0.54 0.26 0.29 0.54 0.56 1.00
TRAINING 0.22 À0.22 0.38 À0.21 À0.21 À0.04 1.00
VOLUME 0.35 À0.13 0.54 0.04 À0.19 0.02 À0.21 1.00
ECO 0.15 À0.22 0.09 À0.11 À0.02 À0.05 0.17 0.09 1.00
INCOMING À0.70 0.26 À0.25 0.59 0.52 0.35 À0.25 À0.11 0.13 1.00
WASTE À0.05 À0.16 0.30 0.15 0.15 0.14 À0.01 0.36 0.29 0.12 1.00
Pearson correlations are estimated using monthly observations from the three production units for the period of September 1999 through September 2003.
Variables de?nitions:
PRODUCTIVITY (labor productivity) = total sum of standard hours of output divided by total production labor hours;
DEFECTS (product defects) = total sum of ?nished product defects (parts per million);
HEADCOUNT (production headcount) = total number of production workers;
TURNOVER (turnover rate) = total number of voluntary/involuntary resignations divided by the average production headcount;
ABSENTEEISM (absenteeism rate) = total number of absences divided by the total number of worked days;
OVERTIME (overtime hours) = total number of production overtime hours;
TRAINING (employee training) = total number of hours of skilled-training;
VOLUME (production volume) = total units of output;
ECO (engineering change orders) = total number of engineering change orders for manufacturing process changes;
INCOMING (material defects incoming) = total number of material component defects during inspection upon arrival at the factory;
WASTE (material waste) = total amount of material waste in US dollars.
F
.
J
.
R
o
m
a
´
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
4
(
2
0
0
9
)
5
8
9
–
6
1
8
6
0
3
Table 3
Time-series regressions on labor productivity
Variable Productivity level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 0.854
***
0.258
***
0.439
***
(14.3) (4.42) (4.25)
OLDPLAN_TREND (b
1
) À0.119
***
À0.002 À0.046
**
(À4.19) (À1.43) (À1.96)
NEWPLAN_TREND (b
2
) 0.003
***
0.004
***
0.006
***
(3.06) (4.31) (4.06)
DEFECTS (b
3
) À0.191
**
À0.014
*
À0.026
(À2.14) (À1.74) (À1.56)
HEADCOUNT (b
4
) 0.0003
*
0.00 0.00
(1.94) (0.24) (0.24)
TURNOVER (b
5
) À0.983
***
À0.650
***
À0.931
**
(À4.12) (À2.82) (À2.32)
ABSENTEEISM (b
6
) À2.94
**
À2.19
**
À3.92
**
(À2.51) (À2.00) (À2.17)
OVERTIME (b
7
) À0.002
***
À0.001
**
À0.001
**
(À4.24) (À2.48) (À1.99)
LAG_TRAINING (b
8
) 0.0003 0.001 À0.004
(0.30) (0.18) (À0.30)
ECO (b
9
) 0.005
*
0.003 0.006
(1.81) (1.52) (1.56)
RETAIL (b
10
) 0.045
*
À0.117
***
À0.267
***
(1.86) (À4.36) (À5.65)
INSTITUTIONAL (b
11
) À0.18 À0.011 À0.015
(À0.75) (À0.43) (À0.22)
Adjusted R
2
0.85 0.72 0.70
Durbin–Watson statistic 1.88 1.99 2.02
N 147 147 147
Test of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.26 –
H
1
: b
1
– b
2
18.3
***
– 5.70
**
Variable Commercial unit Institutional unit Retail unit
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
Panel B: regressions by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.791
***
0.222
**
0.373
**
0.731
***
0.17 0.559
***
0.946
***
0248
***
0.322
***
(6.35) (2.08) (2.09) (6.87) (1.67) (3.78) (15.02) (3.50) (3.52)
OLDPLAN_TREND
(b
1
)
À0.004 0.006 0.013 À0.166
**
0.089 0.113 À0.055
**
À0.033
**
À0.045
**
(À0.65) (0.94) (1.13) (À2.34) (1.18) (0.48) (À4.53) (À2.49) (À2.58)
NEWPLAN_TREND
(b
2
)
0.013
***
0.012
***
0.021
***
0.006
**
0.005
**
0.007
**
0.003
**
0.004
**
0.005
**
(4.22) (4.56) (4.54) (2.39) (2.43) (2.06) (2.02) (2.41) (2.45)
DEFECTS (b
3
) 0.007 0.007 0.012 À0.016 À0.159 À0.033 À0.022
**
À0.014 À0.018
(0.58) (0.60) (0.57) (À1.43) (À1.40) (À1.45) (À2.18) (À1.36) (À1.33)
HEADCOUNT (b
4
) 0.001
**
0.001
***
0.002
***
0.006
*
0.001
*
0.0003 À0.00 À0.001
*
À0.001
*
(3.31) (3.35) (3.29) (1.79) (1.70) (0.64) (À0.66) (À1.88) (1.93)
TURNOVER (b
5
) À0.323 À0.349 À0.596 À0.599 À0.601 À0.323 0.221 0.520 0.720
(À0.90) (À0.99) (À1.00) (À1.46) (À1.44) (À0.43) (0.44) (1.04) (1.11)
ABSENTEEISM (b
6
) 0.256 0.218 0.241 À0.678 À0.598 À0.233 À4.67
**
À3.80 À5.02
(0.20) (0.17) (0.87) (À0.50) (À0.43) (À0.08) (À2.19) (À1.52) (À1.54)
OVERTIME (b
7
) 0.00 0.00 0.001 À0.001
*
À0.001 À0.004
***
À0.003
**
0.002
*
0.003
*
(0.99) (0.95) (0.87) (À1.71) (À1.17) (À2.90) (2.52) (1.77) (1.81)
LAG_TRAINING
(b
8
)
0.010 0.102 0.017 À0.058
**
À0.590
**
À0.165
***
0.012 À0.004 À0.006
(1.21) (1.27) (1.28) (À2.31) (À2.32) (À3.20) (0.55) (À0.21) (0.23)
ECO (b
9
) 0.001 0.001 0.01 0.007
*
0.007
*
0.009 0.021
***
0.0136
**
0.017
**
(0.32) (0.31) (0.32) (1.80) (1.68) (1.18) (4.26) (2.48) (2.49)
Adjusted R
2
0.51 0.53 0.54 0.64 0.56 0.59 0.95 0.85 0.86
Durbin–Watson
statistic
1.86 1.91 1.90 1.95 1.96 1.96 1.97 1.94 2.13
604 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
ductivity, and therefore, they provide empirical sup-
port to hypotheses one and three.
19
Results for the regressions on product defects
and material waste are summarized in Tables 4
and 5, respectively. As shown in Table 4, panel A,
the coe?cient OLDPLAN_TREND is positive
and statistically signi?cant in the defects level
model (b
1
= 0.251, p-value < .05) and positive and
insigni?cant in the change and percentage change
models, whereas NEWPLAN_TREND coe?cient
is negative and statistically signi?cant at the level
(b
2
= À0.052, p-value < .01), level change (b
2
=
À0.047, p-value < .01), and percentage change
(b
2
= À0.168, p-value < .01) models. A test of di?er-
ences between the two coe?cients indicates that the
NEWPLAN_TREND (b
2
= À0.52) coe?cient is sta-
tistically signi?cantly lower than the OLD-
PLAN_TREND coe?cient (b
1
= 0.251) in the
defects level model. However, note that there is no
di?erence between the two coe?cients in the
Table 3 (continued)
Variable Commercial unit Institutional unit Retail unit
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
Productivity
level
Level
change
%
Change
N 49 49 49 49 49 49 49 49 49
Test of equality of main coe?cients (F-statistic)
H
0
:
b
1
= b
2
– 1.59 1.35 – – – – – –
H
1
:
b
1
– b
2
5.28
**
– – 5.83
**
4.37
**
4.8
**
21.2
***
6.91
***
7.30
***
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis.
All regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients
reports F-statistic.
This table presents OLS estimation results from regressing labor productivity, the ratio of earned hours to production hours, on a time trend
representing the period under each inventive plan, while controlling for general factors that could potentially a?ect labor productivity. The sample
consists of month observations from three production units for the period of September 1999 through September 2003. The following model is
estimated separately on productivity, change in productivity, and the percentage change in productivity by pooling data from the three production units:
PRODUCTIVITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
DEFECTS
it
þ b
4
HEADCOUNT
it
þ b
5
TURNOVER
it
þ b
6
ABSENTEEISM
it
þ b
7
OVERTIME
it
þ b
8
LAG TRAINING
it
þ b
9
ECO
it
þ b
10
RETAIL
it
þ b
11
INSTITUTIONAL
it
þ b
12
COMMERCIAL
it
þ e
it
And, the following model individually on each production unit:
PRODUCTIVITY
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
DEFECTS
it
þ b
4
HEADCOUNT
it
þ b
5
TURNOVER
it
þ b
6
ABSENTEEISM
it
þ b
7
OVERTIME
it
þ b
8
LAG TRAINING
it
þ b
9
ECO
it
þ t
it
Variables de?nitions: PRODUCTIVITY (level) = total standard hours of output divided by total production labor hours; PRODUCTIVITY (level
change) = change in productivity from1 month to the next; PRODUCTIVITY (%change) = percentage change in productivity from1 month to the
next; OLDPLAN_TREND = represents a time trend(t) for the period under the oldincentive planandit is measured as the number of months fromthe
beginning of the study period and before the introduction of the team bonus scheme, the period of September 1999 through March 2000. NEW-
PLAN_TREND = represents a time trend (t) for the period under the new incentive plan and it is measured as the number of months from the time in
which all changes to the incentive plan were implemented and thereafter, the period of July 2000 through September 2003. DEFECTS = the log of
?nished product defects (parts per million) adjusted for production volume; HEADCOUNT = number of production workers; TURNOVER
(turnover rate) = number of voluntary/involuntary resignations divided by the average production headcount; ABSENTEEISM (absenteeism
rate) = number of absences divided by the number of worked days in the month; OVERTIME = number of overtime hours scaled by headcount;
LAGTRAINING = prior month total hours of worker training scaled by headcount; ECO (engineering change orders) = number of manufacturing
process changes; RETAIL = a dummy variable set to 1 for observations from the retail manufacturing unit, zero otherwise; INSTITUTIONAL = a
dummy variable set to 1 for observations fromthe institutional manufacturing unit, zero otherwise; COMMERCIAL = a dummy variable set to 1 for
observations from the commercial manufacturing unit, zero otherwise. The COMMERCIAL dummy variable is set to the intercept.
19
For robustness, I alsoestimate analternative speci?cationof the
main model on productivity. Speci?cally, I regress each measure on
productivity (level, level change, and percentage change) perfor-
mance outcome ona time trend for eachof the three bonus schemes:
piece-rates, team-incentive, and the plant-incentive, in place of the
two time trends used currently to denote the periods under the old
and new incentive plan. Similarly, I performed tests of di?erences
across the three respective time trend coe?cients to detect di?er-
ences in performance across the three periods of interest. This
approach was followed in all three estimations. These set of results
are not currently reported in tables, but are available from the
author upon request. Overall, these results are qualitatively similar
to the ones reported in Table 3. In most cases, the coe?cient
estimates PLANT-INCENTIVE TREND, the time trend for the
period in which all elements of the new incentive plan operate fully,
is positive and statistically signi?cantly larger than the period in
which the team bonus operates in isolation and, in some cases, is
greater than the period under piece-rates.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 605
Table 4
Time-series regressions on product defects
Variable Log product defects (Parts per million)
Defects level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 2.22
***
À0.464 À0.165
(4.83) (1.03) (À1.09)
OLDPLAN_TREND (b
1
) 0.251
**
À0.022 0.001
(1.98) (À0.18) (0.30)
NEWPLAN_TREND (b
2
) À0.052
***
À0.047
***
À0.016
***
(À7.03) (À6.46) À(6.81)
HEADCOUNT (b
3
) 0.003
*
0.001 0.001
(1.74) (1.10) (1.29)
TURNOVER (b
4
) À1.84 À4.74
**
À1.38
*
(À0.76) (1.99) (À1.74)
ABSENTEEISM (b
5
) 21.5
*
20.2
*
5.50
(1.92) (1.83) (1.49)
ECO (b
6
) À0.0002 0.021 0.007
(À0.10) (0.74) (0.80)
DEFECTS_INCOMING (b
7
) 0.001 0.002
**
0.001
(1.03) (2.18) (1.45)
RETAIL (b
8
) À0.25
*
À0.003 À0.032
(À1.74) (À0.20) (À0.69)
INSTITUTIONAL (b
9
) 0.076 À0.74
***
À0.137
**
(0.38) (À4.08) (À2.23)
Adjusted R
2
0.64 0.61 0.56
Durbin–Watson statistic 1.97 1.96 1.96
N 147 147 147
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.40 1.90
H
1
: b
1
–b
2
5.70
**
– –
Variable Commercial unit Institutional unit Retail unit
Defects
level
Level
change
%
Change
Defects
level
Level
change
%
Change
Defects
level
Level
change
%
Change
Panel B: regressions by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.42 À0.89 À3.03 3.72
***
À0.18 À0.049 1.68 À0.69 À0.32
(1.00) (À1.52) (À1.48) (3.64) (À0.20) (À0.19) (1.56) (À0.73) (À0.89)
OLDPLAN_TREND (b
1
) 0.234
**
0.069 0.025 0.338 À0.297 À0.078 0.206 0.075 0.028
(2.67) (0.51) (0.53) (0.84) (À0.85) (À0.81) (0.81) (0.31) (0.29)
NEWPLAN_TREND (b
2
) À0.030
***
À0.051
***
À0.018
***
À0.052
**
À0.050
**
À0.140
**
À0.098
***
À0.084
***
À0.034
***
(À5.93) (À7.49) (À7.56) (À2.24) (À2.12) (À2.12) (À4.08) (À3.80) (À3.95)
HEADCOUNT (b
3
) 0.004
***
0.004 0.001 À0.003 À0.003 À.001 0.012
**
0.010
*
0.004
**
(3.46) (1.67) (1.64) (À0.75) (À0.67) (À0.66) (2.06) (1.92) (2.15)
TURNOVER (b
4
) 1.17 À1.11 À0.345 0.867 0.499 0.145 À10.7 À14.1
**
À5.58
**
(0.72) (À0.41) (À0.37) (0.15) (0.90) (0.90) (À1.60) (À2.18) (À2.22)
ABSENTEEISM (b
5
) 41.5
***
6.58 2.12 14.7 13.3 3.64 À25.8 À35.1 À12.5
(4.41) (0.50) (0.47) (0.70) (0.65) (0.63) (À0.75) (À1.06) (À0.98)
ECO (b
6
) 0.035 0.011 0.004 À0.135 À0.013 À0.003 À0.076 À0.028 À0.014
(1.60) (0.43) (0.43) (À0.21) (À0.20) (À0.21) (À1.29) (À0.48) (À0.50)
DEFECTS_INCOMING (b
7
) 0.006
*
0.001 0.001 0.00 0.00 0.00 À0.002 0.00 À0.005
(1.87) (1.15) (1.15) (0.54) (0.43) (0,41) (1.40) (À0.31) (À0.54)
Adjusted R
2
0.88 0.84 0.85 0.27 0.22 0.23 0.54 0.43 0.44
Durbin–Watson statistic 2.08 1.98 1.98 1.92 1.94 1.93 2.03 1.98 1.99
N 49 49 49 49 49 49 49 49 49
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.20 0.45 –** 0.48 0.42 – 2.54 2.24
H
1
: b
1
–b
2
8.66
***
– – 5.49
***
– – 8.91
***
– –
606 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 4 (continued)
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis. All
regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients reports
F-statistic.
This table presents OLS estimation results from regressing the log of ?nished product defects on a time trend representing the period under
each inventive plan, while controlling for general factors that could potentially a?ect product quality. The sample consists of month
observations from three production units for the period of September 1999 through September 2003. The following model is estimated
separately for level, level changes, and percentage change in product defects by pooling data from the three production units:
DEFECTS
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ b
8
RETAIL
it
þ b
9
INSTITUTIONAL
it
þ b
10
COMMERCIAL
it
þ e
it
And, the following model individually on each production unit:
DEFECTS
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ t
it
Variables de?nitions: DEFECTS (level) = log of total ?nished products defects (parts per million) scaled by production volume;
DEFECTS (level change) = change in products defects from 1 month to the next; DEFECTS (% change) = percentage change in product
defects from 1 month to the next; OLDPLAN_TREND = represents a time trend (t) for the period under the old incentive plan and it is
measured as the number of months from the beginning of the study period and before the introduction of the team bonus scheme, the
period of September 1999 through March 2000. NEWPLAN_TREND = represents a time trend (t) for the period under the new incentive
plan and it is measured as the number of months from the time in which all changes to the incentive plan were implemented and thereafter,
the period of July 2000 through September 2003; HEADCOUNT = number of production workers; TURNOVER (turnover
rate) = number of voluntary/involuntary resignations divided by the average production headcount; ABSENTEEISM (absenteeism
rate) = number of absences divided by the number of worked days in the month; ECO (engineering change orders) = number of manu-
facturing process changes; DEFECTS_INCOMING = total number of material component defects detected during an incoming
inspection upon arrival at the factory; RETAIL = a dummy variable set to 1 if retail manufacturing unit, zero otherwise; INSTITU-
TIONAL = a dummy variable set to 1 if institutional manufacturing unit, zero otherwise; COMMERCIAL = a dummy variable set to 1 if
commercial manufacturing unit, zero otherwise. The COMMERCIAL dummy variable is set to the intercept.
Table 5
Time-series regressions on material waste
Variable Material waste
Level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 0.012
***
À0.006
***
À1.82
***
(5.04) (À2.32) (À3.43)
OLDPLAN_TREND (b
1
) À0.004 0.0005 0.171
**
(À0.89) (1.18) (2.48)
NEWPLAN_TREND (b
2
) À0.016
***
À0.0002
***
À0.021
**
(À3.69) (À3.35) (À2.56)
HEADCOUNT (b
3
) À0.002
**
0.0001 0.006
***
(À2.03) (1.25) (3.09)
TURNOVER (b
4
) 0.003 À0.007 À0.18
(0.32) (À0.64) (À0.70)
ABSENTEEISM (b
5
) À0.020 0.076 9.40
(À0.43) (1.42) (0.77)
ECO (b
6
) 0.0002 0.00 0.021
(0.21) (0.72) (0.76)
RETAIL (b
7
) À0.001 0.0008 0.23
(À0.97) (0.72) (1.34)
INSTITUTIONAL (b
8
) 0.001 0.003
**
1.29
***
(0.14) (2.47) (6.04)
Adjusted R
2
0.21 0.22 0.41
Durbin–Watson statistic 1.79 2.06 1.99
N 147 147 147
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
1.49 –
H
1
: b
1
–b
2
4.33
**
– 6.77
***
(continued on next page)
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 607
changes or percentage change models. This could be
attributed to the fact that the impact on product
defects occurs almost immediately after the imposi-
tion of the quality quota and it is not as gradual as
in the case of productivity, so there is very little var-
iation in defects over time.
Table 5 (continued)
Variable Commercial unit Institutional unit Retail unit
Level Level
change
%
Change
Level Level
change
%
Change
Level Level
change
%
Change
Panel B: regressions by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.015
***
0.005 0.39 0.010
***
À0.003 0.095 0.003 À0.008
***
À2.82
***
(3.44) (1.32) (1.14) (4.79) (À1.42) (0.18) (1.48) (À3.99) (À4.61)
OLDPLAN_TREND
(b
1
)
0.001 0.002
**
0.14
**
0.003
***
À0.0005 0.086 À0.007 0.0001 0.40
**
(1.21) (2.97) (2.18) (3.67) (À0.88) (0.66) (À1.24) (0.33) (2.08)
NEWPLAN_TREND
(b
2
)
À0.002
*
À0.0001
*
À0.017
*
À0.002
***
À0.0002
***
À0.039
***
À0.009
*
À0.0009
**
À0.012
(À1.77) (À1.82) (À1.80) (À6.18) (À3.63) (À3.34) (À1.78) (À2.47) (À1.10)
HEADCOUNT (b
3
) À0.0002
*
À0.0003
**
À0.002
*
0.00 0.0003
**
0.006
**
0.00 0.0001 0.009
***
(À1.76) (À2.32) (À2.36) (0.55) (2.30) (2.35) (0.72) (1.45) (3.55)
TURNOVER (b
4
) À0.031
*
À0.031
*
À3.43
**
À0.025 À0.037
*
À6.78 À0.012 0.024
**
10.2
**
(1.77) (À1.98) (À2.21) (À1.49) (À1.95) (À1.47) (À0.78) (2.63) (2.70)
ABSENTEEISM (b
5
) À0.017 0.061 6.75 À0.11 0.041 À5.83 0.048 0.17
***
3.11
(À0.25) (1.01) (1.19) (À1.67) (0.50) (À0.29) (0.65) (2.81) (1.29)
ECO (b
6
) À0.00 À0.00 À0.00 0.001 0.00 0.096 À0.00 À0.001 À0.043
(À0.47) (À0.18) (0.10) (0.85) (0.39) (1.62) (À0.54) (À1.36) (À0.97)
Adjusted R
2
0.44 0.42 0.41 0.74 0.52 0.50 0.78 0.76 0.87
Durbin–Watson
statistic
1.70 1.65 1.96 1.94 1.91 1.98 1.95 2.01 1.75
N 49 49 49 49 49 49 49 49 49
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– – – 1.37 1.68 – 1.26 –
H
1
: b
1
–b
2
3.24
**
10.6
***
6.06
***
13.8
***
– – 3.06
**
– 4.04
***
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis.
All regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients
reports F-statistic.
This table presents OLS estimation results from regressing material waste on a time trend representing the period under each inventive
plan, while controlling for general factors that could potentially a?ect material waste. The sample consists of month observations from
three production units or the period of September 1999 through September 2003. The following model is estimated separately for level,
level changes, and percentage change in material waste by pooling data from the three production units:
WASTE
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ b
8
RETAIL
it
þ b
9
INSTITUTIONAL
it
þ b
10
COMMERCIAL
it
þ e
it
And, the following model individually on each production unit:
WASTE
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
TURNOVER
it
þ b
5
ABSENTEEISM
it
þ b
6
ECO
it
þ b
7
DEFECTS INCOMING
it
þ t
it
Variables de?nitions: WASTE (level) = total amount of materials waste scaled by production volume; WASTE (level change) = change
in material waste from 1 month to the next; WASTE (% change) = percentage change in materials waste from 1 month to the next;
OLDPLAN_TREND = represents a time trend (t) for the period under the old incentive plan and it is measured as the number of months
from the beginning of the study period and before the introduction of the team bonus scheme, the period of September 1999 through
March 2000. NEWPLAN_TREND = represents a time trend (t) for the period under the new incentive plan and it is measured as the
number of months from the time in which all changes to the incentive plan were implemented and thereafter, the period of July 2000
through September 2003; HEADCOUNT = number of production workers; TURNOVER (turnover rate) = total number of voluntary/
involuntary resignations divided by the average production headcount); ABSENTEEISM (absenteeism rate) = number of absences
divided by the total number of worked days in the month; ECO (engineering change orders) = number of manufacturing process changes;
RETAIL = a dummy variable set to 1 if retail manufacturing unit, zero otherwise; INSTITUTIONAL = a dummy variable set to 1 if
institutional manufacturing unit, zero otherwise; COMMERCIAL = a dummy variable set to 1 if commercial manufacturing unit, zero
otherwise. The COMMERCIAL dummy variable is set to the intercept.
608 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 4, panel B, reports results of regressions by
each production unit. As shown, the results mirror
the same pattern of results from the pooled sample,
reinforcing that the conclusion that the new incen-
tive plan is associated with a decrease in products
defects.
Turning to the results on material waste in Table
5, we can observe that though the coe?cient esti-
mate for the time trend under the old compensation
plan is not signi?cant, the time trend for the new
plan’s time trend is negative and signi?cant at the
1% level across the pooled regressions, as well as
on the regressions by production unit, implying that
the amount of material waste decreased markedly
after the changes were made to the compensation
plan. Thus, these ?ndings and the ones on product
defects lend support to hypothesis two and partial
support to hypotheses three.
20
Overall, productivity and quality improvements
associated with the new incentive plan are sizeable
and fairly signi?cant as depicted in Fig. 2. As
observed, labor productivity begins to rise gradually
after the team bonus (output-target scheme) is intro-
duced, and this rise continues gradually for several
months until it reaches the upper 90–100 percentiles
through the end of the sample period. The same
broad pattern of improvements is observed for
product defects in panel B. However, quality
improvements are realized almost immediately once
the team bonus takes e?ect. Observe, for example,
the sharp decline in product defects in May 2000,
1 month after the quality quota is amended as part
of the team bonus. The steep drop in product
defects continues right up until the introduction of
the plant bonus in July 2000 and then levels o?
shortly after. This provides further evidence that
the imposition of the quality quota succeeded in
reducing the number of product defects.
In synthesis, all of the above results suggest
that the changes made to the original compensa-
tion plan, the adoption of team and the plant
bonus along with the various management con-
trol initiatives, led to a rise in productivity and
to a reduction in product defects and material
waste.
Regression results on absenteeism and turnover
To assess the impact of the e?ects of the new
incentive plan on absenteeism (Hypothesis 4), I
follow a similar approach and regress each variable
against a time trend for each, the old and new incen-
tive plan, and several controls, including the indus-
try’s rate of absenteeism and turnover in the region
where the plant operates. Results on absenteeism
are summarized in Table 6, panels A and B. As
shown, the coe?cient estimates for NEW-
PLAN_TREND is negative and signi?cant, whereas
the OLDPLAN_TREND coe?cient is positive and
signi?cant in the pooled and individual regressions
by production unit. These ?ndings indicate that
after controlling for the region’s industry rates of
absenteeism, the new incentive plan led to a reduc-
tion in absenteeism, and therefore, provides support
to Hypothesis 4.
Results on turnover are reported on Table 6,
panel A (columns 4–6), and on panel C. Interest-
ingly, in the pooled regressions, the coe?cient for
the OLDPLAN_TREND is positive and signi?cant
among the level, level changes, and percentage
change model. In contrast, the NEW-
PLAN_TREND coe?cient is negative and signi?-
cant among the three speci?cations. Similar results
are observed on the regressions by production unit.
Lastly, a test of di?erence between b
1
and b
2
on the
pooled as well as on the regressions by production
unit reveal that b
2
is signi?cantly lower than b
1
.
Taken together, these results indicate that the new
incentive plan led to a reduction in worker turnover
as well.
21
20
Just as in the case of productivity, to ensure robustness of my
results, I also estimate a similar model on both, product defects
and material waste. Speci?cally, I regress each performance
outcome (level, level change, and percentage change) on a time
trend for each of the three bonus schemes: piece-rates, team-
incentive, and the plant-incentive, in place of the two time trends
used currently to denote the periods under the old and new
incentive plan. Although the results are weaker to the ones
reported in tables due to the limited number of observations
surrounding the team-incentive, the general inference remains
unchanged. In most cases, the PLANT-INCENTIVE TREND
coe?cient is negative and signi?cantly smaller than the time trend
coe?cients for the periods under piece-rates and the team bonus
on the regressions on both regressions for product defects and
material waste.
21
To ensure robustness of these results, I followed a similar
approach to that of productivity and quality and regress
absenteeism and turnover on three time trend variables to
represent the periods under each of the three bonus schemes:
piece-rates, team-incentive, and the plant-incentive. These results
are not reported in tables, however, they are qualitatively similar
to the ones reported in Table 6.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 609
Discussion of the underlying factors that led to a rise
in productivity, quality and a reduction in absenteeism
and turnover
So far, the results from my regressions support
the central propositions that productivity, product
quality, absenteeism, and turnover improved after
the adoption of the new incentive plan. A follow
up question to address is which factors drive the
underlying performance improvements. Is it higher
e?ort, a selection e?ect, more peer pressure, or
heightened worker cooperation driving these
results? I recognize that given the available data, it
is impossible to empirically disentangle precisely
which factors drive the improvements. Nonetheless,
I address this question with qualitatively evidence
gathered from various interviews held with manag-
ers, production line supervisors, and assembly
January 1999 through September 2003
0%
20%
40%
60%
80%
100%
120%
J
a
n
-
9
9
M
a
r
-
9
9
M
a
y
-
9
9
J
u
l
-
9
9
S
e
p
-
9
9
N
o
v
-
9
9
J
a
n
-
0
0
M
a
r
-
0
0
M
a
y
-
0
0
J
u
l
-
0
0
S
e
p
-
0
0
N
o
v
-
0
0
J
a
n
-
0
1
M
a
r
-
0
1
M
a
y
-
0
1
J
u
l
-
0
1
S
e
p
-
0
1
N
o
v
-
0
1
J
a
n
-
0
2
M
a
r
-
0
2
M
a
y
-
0
2
J
u
l
-
0
2
S
e
p
-
0
2
N
o
v
-
0
2
J
a
n
-
0
3
M
a
r
-
0
3
M
a
y
-
0
3
J
u
l
-
0
3
S
e
p
-
0
3
Date
April 2000
Team output target
scheme
July 2000
Plant-wide bonus &
management control
initiatives
Previous incentive plan: piece-rates
New incentive plan: Team bonus, plant bonus, and control initiatives
September 1999 through September 2003
0
5,000
10,000
15,000
20,000
25,000
S
e
p
-
9
9
N
o
v
-
9
9
J
a
n
-
0
0
M
a
r
-
0
0
M
a
y
-
0
0
J
u
l
-
0
0
S
e
p
-
0
0
N
o
v
-
0
0
J
a
n
-
0
1
M
a
r
-
0
1
M
a
y
-
0
1
J
u
l
-
0
1
S
e
p
-
0
1
N
o
v
-
0
1
J
a
n
-
0
2
M
a
r
-
0
2
M
a
y
-
0
2
J
u
l
-
0
2
S
e
p
-
0
2
N
o
v
-
0
2
J
a
n
-
0
3
M
a
r
-
0
3
M
a
y
-
0
3
J
u
l
-
0
3
S
e
p
-
0
3
Date
Previous incentive plan: piece-rates
New incentive plan: team bonus, plant bonus, and control intitiatives
April 2000
Team output target
scheme
J
Plant-wide bonus &
management control
initiatives
uly 2000
a
b
Fig. 2. (a) Monthly labor productivity for entire plant. (b) Monthly ?nished product defects for entire plant.
610 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Table 6
Time-series regressions on absenteeism and turnover
Variable Absenteeism Turnover
Level Level change % Change Level Level change % Change
Panel A: pooled sample (monthly data)
INTERCEPT (b
0
) 0.018
***
À0.004 À0.22 0.023 À0.107
***
À1.41
***
(3.90) (À1.27) (À1.01) (1.15) (À5.38) (À3.77)
OLDPLAN_TREND (b
1
) 0.0027
***
0.0012
***
0.123
**
0.008
**
0.010
***
0.174
**
(2.75) (2.33) (2.30) (2.27) (2.62) (2.06)
NEWPLAN_TREND (b
2
) À0.0004
***
À0.0004
***
À0.021
***
À0.0002
***
À0.0016
***
À0.015
**
(À6.46) (À6.41) (À6.04) (À3.98) (À3.94) (À2.04)
HEADCOUNT (b
3
) 0.0001 0.0001 0.0001 0.0002
***
0.0001
**
0.003
**
(0.85) (0.70) (0.29) (2.99) (2.48) (2.10)
REGION_ABSENTEEISM (b
4
) 0.022 0.031 1.99 – – –
(0.76) (1.04) (1.10) – – –
REGION_TURNOVER (b
5
) – – – 0.001
***
0.227
**
4.83
***
– – – (3.17) (2.59) (2.78)
RETAIL (b
6
) À0.004
*
0.003 0.037 À0.021 0.004
***
0.112
(À1.85) (1.27) (0.32) (À1.46) (3.15) (0.56)
INSTITUTIONAL (b
7
) À0.002 0.002 0.139 À0.017 0.063
***
0.494
**
(À0.71) (1.18) (1.07) (1.18) (4.46) (2.23)
Adjusted R
2
0.37 0.33 0.34 0.48 0.44 0.27
Durbin–Watson statistic 1.94 1.97 1.96 1.97 1.98 1.94
N 147 147 147 147 147 147
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– – – – – –
H
1
: b
1
–b
2
10.2
***
7.18
***
7.00
***
7.98
***
9.73
***
5.28
***
Variable Commercial unit Institutional unit Retail unit
Level Level change % Change Level Level change % Change Level Level change % Change
Panel B: regressions on absenteeism by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.030
***
À0.004 À0.271 0.007 À0.0007 À0.57 0.010
***
À0.003 À0.383
(3.74) (À0.59) (À0.76) (1.07) (À1.30) (À1.61) (2.27) (À0.88) (À1.36)
OLDPLAN_TREND (b
1
) 0.005
***
0.0003 0.019 0.008
***
0.007
***
0.615
***
0.0015 0.0017
***
0.058
*
(2.83) (1.22) (0.26) (3.69) (3.66) (4.84) (1.37) (3.45) (1.74)
NEWPLAN_TREND (b
2
) À0.0005
***
À0.0004
***
À0.0234
***
À0.0005
***
À0.0004
***
À0.023
***
À0.0003
***
À0.0002
***
À0.021
***
(À5.66) (À4.00) (À4.07) (À3.34) (À3.44) (À3.33) (À5.95) (À4.07) (À5.65)
HEADCOUNT (b
3
) À0.0006 0.00 0.0004 0.0001
*
0.0005 0.003 0.0002 0.0001 0.001
(À1.60) (0.17) (0.25) (1.76) (1.54) (1.66) (1.16) (0.54) (0.99)
REGION_ABSENTEEISM (b
4
) 0.071 0.060 3.25 À0.011 0.010 0.983 À0.002 0.013 1.96
(0.99) (1.23) (1.38) (À0.17) (0.16) (0.24) (À0.80) (0.39) (0.86)
Adjusted R
2
0.32 0.36 0.37 0.46 0.47 0.56 0.57 0.56 0.60
(continued on next page)
F
.
J
.
R
o
m
a
´
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
4
(
2
0
0
9
)
5
8
9
–
6
1
8
6
1
1
Table 6 (continued)
Variable Absenteeism Turnover
Level Level change % Change Level Level change % Change
Durbin–Watson statistic 1.91 1.99 1.98 1.90 1.93 1.94 1.98 1.97 2.03
N 49 49 49 49 49 49 49 49 49
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– 1.28 1.71 – – – – – –
H
1
: b
1
–b
2
9.80
***
– – 10.8
***
11.4
***
15.9
***
2.8
*
13.7
***
5.92
***
Variable Commercial unit Institutional unit Retail unit
Level Level change % Change Level Level change % Change Level Level change % Change
Panel C: regressions on turnover by manufacturing unit (monthly data)
INTERCEPT (b
0
) 0.158
***
0.022 0.186 0.012 À0.037 À0.23 À0.024 À0.094
***
À1.35
***
(4.07) (0.62) (0.64) (0.49) (À1.31) (À0.30) (À1.06) (À4.19) (À4.30)
OLDPLAN_TREND (b
1
) 0.021
**
0.017
**
0.139
**
0.023
***
0.020
*
0.097
***
0.009
*
0.021
***
0.32
***
(2.41) (2.08) (2.06) (2.40) (1.84) (4.76) (1.76) (3.86) (4.30)
NEWPLAN_TREND (b
2
) À0.0019
***
À0.002
***
À0.014
***
À0.002
***
À0.0017
***
À0.029
**
À0.001
***
À0.001
***
À0.019
***
(À4.58) (À4.62) (À4.63) (À4.26) (À4.35) (À2.11) (À3.87) (À4.30) (À4.25)
HEADCOUNT (b
3
) À0.0004
***
À0.0005
**
À0.003
**
0.0002 0.0001 À0.0008 0.0003
***
0.0003
***
0.004
***
(À2.71) (À2.65) (À2.66) (1.55) (1.25) (À0.17) (3.91) (3.22) (3.34)
REGION_TURNOVER (b
4
) 0.403 0.392 3.11 0.159 0.186 8.48
**
0.140 .166 2.36
(1.16) (1.19) (1.18) (0.48) (0.49) (2.46) (0.82) (0.97) (0.99)
Adjusted R
2
0.18 0.17 0.18 0.27 0.25 0.32 0.51 0.60 0.62
Durbin–Watson statistic 1.87 1.90 1.90 1.82 1.66
*
2.36 1.92 1.97 1.97
N
Tests of equality of main coe?cients (F-statistic)
H
0
: b
1
= b
2
– – – – – – – – –
H
1
: b
1
–b
2
7.02
***
5.37
**
5.28
**
6.72
***
4.12
**
22.8
***
4.43
***
17.3
***
21.3
***
***,**,*
denotes statistical signi?cance at the 1%, 5%, and 10% levels in a two-tailed test, respectively; t-statistics reported in parenthesis.
All regressions report Yule-Walker estimates (Prais and Winsten) to adjust for autocorrelation. Test of equality of main coe?cients reports F-statistic.
This table presents OLS estimation results from regressing absenteeism and worker turnover, respectively, on a time trend representing the period inventive plan, while controlling for
the regional rates of absenteeism and turnover of the maquiladora industry where the factory operates. The sample consists of month observations from three production units
covering the period of September 1999 through September 2003. The following models are estimated separately for level, level changes, and rate change in absenteeism and turnover,
respectively, pooling data from the three production units:
ð1Þ ABSENTEEISM
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
REGION ABSENTEEISM
it
þ b
5
RETAIL
it
þ b
6
INSTITUTIONAL
it
þ b
7
COMMERCIAL
it
þv
it
ð2Þ TURNOVER
it
¼ b
0
þ b
1
OLDPLAN TREND
it
þ b
2
NEWPLAN TREND
it
þ b
3
HEADCOUNT
it
þ b
4
REGION TURNOVER
it
þ b
5
RETAIL
it
þ b
6
INSTITUTIONAL
it
þ b
7
COMMERCIAL
it
þ s
it
6
1
2
F
.
J
.
R
o
m
a
´
n
/
A
c
c
o
u
n
t
i
n
g
,
O
r
g
a
n
i
z
a
t
i
o
n
s
a
n
d
S
o
c
i
e
t
y
3
4
(
2
0
0
9
)
5
8
9
–
6
1
8
workers who witnessed the introduction of the new
incentive plan ?rsthand.
Based on their opinions, I o?er the following
potential explanations. First, the perception among
workers is that the shift from piece-rates to the out-
put-target scheme induce them to work harder to
reach the output and quality quotas to earn a bonus.
However, a crucial mechanism that made the incen-
tive work was in providing teams with feedback on
their performance. This mechanism enable teams to
monitor their progress more e?ectively plus it enabled
them to exercise more control within their respective
production cell in such a way that they were better
able to coordinate their e?orts to reach the speci?ed
output and quality targets and to monitor each
other’s e?orts more rigorously to minimize shirking.
Second, management concurred that the plant-
wide incentive and the initiative to report feedback
on each team’s performance throughout the factory
has led to more cooperation and information sharing
throughout the plant. For example, production
supervisors acknowledged that such initiative has
promoted the rate of learning and information shar-
ing between the teams; stressing that it is not uncom-
mon for a particular production cell to communicate
with others upon discovering a method to speed-up
production, eliminate bottlenecks, or identify adefec-
tive component. Similarly, production teams meet on
a regular basis to discuss ways to improve productiv-
ity and quality. This behavior was uncommon under
the old incentive plan.
Third, workers noted that the levels of peer pres-
sure within each team rise as a result of the output-
target scheme and the accompanying mechanism
established to discipline shirkers. Several interviews
with production workers con?rm that this initiative
succeeded in raising the levels of peer pressure across
teams. This behavior also was not as prevalent under
the old incentive plan. Although piece-rates created
some friction between potential shirkers and high
performers, shirking was not formally disciplined in
the ?rm; thus, peer pressure was to some extent inef-
fective. In the words of a production worker: ‘‘. . .If a
member of a production cell is not pulling his weight
to meet the daily output quota, we demand extra
e?ort or otherwise we report his behavior with the
production supervisor. If this behavior persists, we
request his dismissal from the team”.
Last, it is possible that the new incentive plan
attracted workers with higher skill, producing a
sorting or selection e?ect. Therefore, a higher
quality of the workforce could have led to higher A
n
d
,
t
h
e
f
o
l
l
o
w
i
n
g
m
o
d
e
l
s
i
n
d
i
v
i
d
u
a
l
l
y
o
n
e
a
c
h
p
r
o
d
u
c
t
i
o
n
u
n
i
t
:
ð
1
Þ
A
B
S
E
N
T
E
E
I
S
M
i
t
¼
b
0
þ
b
1
O
L
D
P
L
A
N
T
R
E
N
D
i
t
þ
b
2
N
E
W
P
L
A
N
T
R
E
N
D
i
t
þ
b
3
H
E
A
D
C
O
U
N
T
i
t
þ
b
4
R
E
G
I
O
N
A
B
S
E
N
T
E
E
I
S
M
i
t
þ
v
i
t
ð
2
Þ
T
U
R
N
O
V
E
R
i
t
¼
b
0
þ
b
1
O
L
D
P
L
A
N
T
R
E
N
D
i
t
þ
b
2
N
E
W
P
L
A
N
T
R
E
N
D
i
t
þ
b
3
H
E
A
D
C
O
U
N
T
i
t
þ
b
4
R
E
G
I
O
N
T
U
R
N
O
V
E
R
i
t
þ
s
i
t
V
a
r
i
a
b
l
e
s
d
e
?
n
i
t
i
o
n
s
:
A
B
S
E
N
T
E
E
I
S
M
(
a
b
s
e
n
t
e
e
i
s
m
r
a
t
e
)
=
n
u
m
b
e
r
o
f
a
b
s
e
n
c
e
s
d
i
v
i
d
e
d
b
y
t
h
e
t
o
t
a
l
n
u
m
b
e
r
o
f
w
o
r
k
e
d
d
a
y
s
i
n
t
h
e
m
o
n
t
h
;
A
B
S
E
N
T
E
E
I
S
M
(
l
e
v
e
l
c
h
a
n
g
e
)
=
c
h
a
n
g
e
i
n
a
b
s
e
n
t
e
e
i
s
m
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
A
B
S
E
N
T
E
E
I
S
M
(
%
c
h
a
n
g
e
)
=
p
e
r
c
e
n
t
a
g
e
c
h
a
n
g
e
i
n
a
b
s
e
n
t
e
e
i
s
m
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
T
U
R
N
O
V
E
R
(
t
u
r
n
o
v
e
r
r
a
t
e
)
=
n
u
m
b
e
r
o
f
v
o
l
u
n
t
a
r
y
/
i
n
v
o
l
u
n
t
a
r
y
r
e
s
i
g
n
a
t
i
o
n
s
d
i
v
i
d
e
d
b
y
t
h
e
a
v
e
r
a
g
e
p
r
o
d
u
c
t
i
o
n
h
e
a
d
c
o
u
n
t
;
T
U
R
N
O
V
E
R
(
l
e
v
e
l
c
h
a
n
g
e
)
=
c
h
a
n
g
e
i
n
t
u
r
n
o
v
e
r
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
T
U
R
N
O
V
E
R
(
%
c
h
a
n
g
e
)
=
p
e
r
c
e
n
t
a
g
e
c
h
a
n
g
e
i
n
t
u
r
n
o
v
e
r
r
a
t
e
f
r
o
m
1
m
o
n
t
h
t
o
t
h
e
n
e
x
t
;
O
L
D
P
L
A
N
_
T
R
E
N
D
=
r
e
p
r
e
s
e
n
t
s
a
t
i
m
e
t
r
e
n
d
(
t
)
f
o
r
t
h
e
p
e
r
i
o
d
u
n
d
e
r
t
h
e
o
l
d
i
n
c
e
n
t
i
v
e
p
l
a
n
a
n
d
i
t
i
s
m
e
a
s
u
r
e
d
a
s
t
h
e
n
u
m
b
e
r
o
f
m
o
n
t
h
s
f
r
o
m
t
h
e
b
e
g
i
n
n
i
n
g
o
f
t
h
e
s
t
u
d
y
p
e
r
i
o
d
a
n
d
b
e
f
o
r
e
t
h
e
i
n
t
r
o
d
u
c
t
i
o
n
o
f
t
h
e
t
e
a
m
b
o
n
u
s
s
c
h
e
m
e
,
t
h
e
p
e
r
i
o
d
o
f
S
e
p
t
e
m
b
e
r
1
9
9
9
t
h
r
o
u
g
h
M
a
r
c
h
2
0
0
0
;
N
E
W
P
L
A
N
_
T
R
E
N
D
=
r
e
p
r
e
s
e
n
t
s
a
t
i
m
e
t
r
e
n
d
(
t
)
f
o
r
t
h
e
p
e
r
i
o
d
u
n
d
e
r
t
h
e
n
e
w
i
n
c
e
n
t
i
v
e
p
l
a
n
a
n
d
i
t
i
s
m
e
a
s
u
r
e
d
a
s
t
h
e
n
u
m
b
e
r
o
f
m
o
n
t
h
s
f
r
o
m
t
h
e
t
i
m
e
i
n
w
h
i
c
h
a
l
l
c
h
a
n
g
e
s
t
o
t
h
e
i
n
c
e
n
t
i
v
e
p
l
a
n
w
e
r
e
i
m
p
l
e
m
e
n
t
e
d
a
n
d
t
h
e
r
e
a
f
t
e
r
,
t
h
e
p
e
r
i
o
d
o
f
J
u
l
y
2
0
0
0
t
h
r
o
u
g
h
S
e
p
t
e
m
b
e
r
2
0
0
3
;
H
E
A
D
C
O
U
N
T
=
n
u
m
b
e
r
o
f
p
r
o
d
u
c
t
i
o
n
w
o
r
k
e
r
s
;
R
E
G
I
O
N
_
A
B
S
E
N
T
E
E
I
S
M
=
t
h
e
r
e
g
i
o
n
’
s
m
a
q
u
i
l
a
d
o
r
a
i
n
d
u
s
t
r
y
r
a
t
e
o
f
a
b
s
e
n
t
e
e
i
s
m
;
R
E
G
I
O
N
_
T
U
R
N
O
V
E
R
=
t
h
e
r
e
g
i
o
n
’
s
m
a
q
u
i
l
a
d
o
r
a
r
a
t
e
o
f
t
u
r
n
o
v
e
r
;
R
E
T
A
I
L
=
a
d
u
m
m
y
v
a
r
i
a
b
l
e
s
e
t
t
o
1
i
f
r
e
t
a
i
l
m
a
n
u
f
a
c
t
u
r
i
n
g
u
n
i
t
,
z
e
r
o
o
t
h
e
r
w
i
s
e
;
I
N
S
T
I
T
U
T
I
O
N
A
L
=
a
d
u
m
m
y
v
a
r
i
a
b
l
e
s
e
t
t
o
1
i
f
i
n
s
t
i
t
u
t
i
o
n
a
l
(
l
o
c
k
e
r
l
o
c
k
)
m
a
n
u
f
a
c
t
u
r
i
n
g
u
n
i
t
,
z
e
r
o
o
t
h
e
r
w
i
s
e
;
C
O
M
M
E
R
C
I
A
L
=
a
d
u
m
m
y
v
a
r
i
a
b
l
e
s
e
t
t
o
1
i
f
c
o
m
m
e
r
c
i
a
l
m
a
n
u
f
a
c
t
u
r
i
n
g
u
n
i
t
,
z
e
r
o
o
t
h
e
r
w
i
s
e
.
T
h
e
C
O
M
M
E
R
C
I
A
L
d
u
m
m
y
v
a
r
i
a
b
l
e
i
s
s
e
t
t
o
t
h
e
i
n
t
e
r
c
e
p
t
.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 613
productivity andbetter quality over time. Thoughit is
di?cult to make a direct test of that performance
improvements resulted from a selection e?ect given
that there is no data available on individual workers
productivity, production supervisors pointed out
that the incentive plan did have some e?ect in attract-
ing (repelling) more (less) productive employees.
As for the underlying reasons that led to a reduc-
tion in absenteeism, ?rst, it was con?rmed that
absenteeism was in part mitigated by the imposition
of a penalty in the plant bonus. The threat of being
excluded from participating in the bonus plan for
not keeping good attendance was indeed a powerful
mechanism in helping reduce absenteeism. Further-
more, workers noted that peer pressure within each
team also contributed to lowering absenteeism.
Absentees made it more di?cult for teams to reach
the day’s quotas; workers exerted pressure on co-
workers who regularly missed work by threatening
to dismiss them from the team.
Was the new incentive plan cost e?ective for the ?rm?
To this point, this study has argued that the new
incentive plan has led to various performance
improvements for this ?rm. A ?nal question to
address is whether the new incentive has been cost
e?ective. In other words, are the improvements in
productivity and quality, and the decrease in absen-
teeism and turnover economically important for the
?rm? To illustrate better, the factory experienced,
on average, a 31% rise in productivity and a 95%
reduction in product defects. However, it could very
well be the case that the incentive plan also raised
labor costs, o?-setting any performance improve-
ments and making the plan cost ine?ective.
To better understand the impact of the incentive
plan on production costs, I brought these results to
the attention of management. They indicated that
indeed the ?rm experienced a rise in labor costs
per worker as a result of having to pay higher
bonuses under the new plan; however, it was
emphasized that the rise in labor costs was o?-set
with higher productivity and better product quality.
For example, management estimates that the 31%
rise in productivity amounts to an additional output
between 5 and 7 million locks annually. It is impor-
tant to stress that this is not a one-time increase in
output that dissipates; the incremental volume
attributed to the gains in productivity is permanent
and carries over to subsequent years. Also, the costs
related to product defects, such as rework, plus
material waste costs after adjusting for production
volume, decrease substantially. In this regard, it is
also worthwhile mentioning that product defects
have been under control. For example, the plant
has had less than 300 defects (PPMs) per month
consistently since the incentive plan was introduced,
the conventional assessment of Six Sigma quality.
Finally, since turnover andabsenteeismdecreased,
this helped reduce training costs on new workers. In
sum, the plant’s bottomline improved despite the rise
in labor costs after the changes were made to the com-
pensation plan. If we account for all of these factors,
managers believed the improvements in manufactur-
ing performance easily o?-set the increase in direct
labor costs.
Conclusion
Using data from three production units of a large
manufacturing plant that employs production teams
in its assembly operations, this study investigates
how workers react to changes made to an existing
incentive plan for teams and whether the underlying
changes led to improvements in productivity and
product quality, and in reducing absenteeism and
turnover. The ?rm switched from an incentive plan
that relied on piece-rates to a two tier incentive plan
that rewarded individual team performance and
plant-wide performance. The new incentive plan
was introduced concurrently with several organiza-
tional control initiatives that aim to facilitate coop-
eration, reinforce monitoring, promote learning,
and encourage peer pressure between workers.
These include the distribution of performance
reports to teams, mechanisms to facilitate the
exchange of information to improve performance,
and rules to punish shirking and uncooperativeness.
I ?nd signi?cant improvements in worker pro-
ductivity and product quality as well as reductions
in worker absenteeism and turnover after the imple-
mentation of the incentive plan. These ?ndings
underscores an important point that has not been
emphasized in existing empirical studies of incentive
pay for teams: the need to introduce management
control and organizational changes in tandem with
incentive pay to capture greater incentive e?ects
from workers. In particular, organizational initia-
tives such as reporting performance to workers with
the application of formal rules to punish shirking
and uncooperative behavior are more e?ective moti-
vators than the e?ect generated by monetary incen-
tives alone.
614 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
In all fairness, some caveats apply. First, I recog-
nize that attempting to disentangle the e?ects on
productivity and defects empirically to each element
in the incentive plan presents limitations. Also, the
small window of separation between the implemen-
tation periods of the two incentives schemes pre-
cludes complete isolation of the performance
impact across the two. Finally, the traditional
caveat of data limitations with ?eld-based research
also applied in this paper, especially since the sam-
ple was truncated and could not account for all
observations in the pre incentive period.
In spite this, my results raised several new ques-
tions not previously address in the literature. Mainly,
my ?ndings that monetary incentives must be com-
plemented with other management control initiatives
when designing incentive plans for teams suggest that
?rms may want to consider adopting them jointly to
achieve better results. I suspect that this conclusion
holds more general; therefore, I leave it to future
research studies to examine the broader applicability
of these ?ndings. Perhaps experimental studies could
replicate the incentive plan implemented at this fac-
tory and see whether the same results hold.
Acknowledgements
I owe particular thanks to Jose Saralegui and the en-
tire management teamat the researchsite for their valu-
able support. This paper is based in part on my
dissertation at The University of Arizona. I gratefully
acknowledge the guidance of Leslie Eldenburg (Chair),
Alfonso Flores-Lagunes, Kaye Newberry, Je? Schatz-
berg, and William Waller. I would also like to thank
Ashiq Ali, David Larcker, Brian Rountree, Mark
Trombley, William Schwartz, Wim Van der Stede, an
anonymous referee, and workshop participants at the
University of Arizona, University of California Irvine,
EmoryUniversity, Rice University, andthe 2003Amer-
ican Accounting Association Annual Meeting for pro-
viding helpful comments on earlier drafts. Also, special
thanks to Michael Costa for his editorial assistance.
Appendix
Worker compensation under the original and new
incentive plan
Compensation under the original incentive plan
Each worker received a daily ?at wage of 70,
80, 91, or 101 Mexican pesos, depending on a
worker’s seniority and level of certi?cation. In
addition to the daily wage, each worker could
receive an attendance bonus of approximately
$115 pesos per week ($23 pesos per day). There
is also a group piece-rate bonus that ?uctuates
according to the output generated by each produc-
tion group. While some groups were more dynamic
than others, each worker earned on average a
piece-rate bonus of $600 pesos.
22
The average daily
compensation under the original incentive plan is
as follows:
Daily wage $90
Attendance bonus 23
Group-piece-rates 30
Total daily compensation $143 pesos
Compensation under the new incentive plan
Although, the base wage increased slightly dur-
ing the study period, the resulting incremental
change does not di?er signi?cantly from the base
wage under the old incentive plan.
23
Thus, for sim-
pli?cation and easy exposition, I am using the
daily ?at wage that was in place under the old
incentive plan. Under the team-level bonus, each
team member receives a daily bonus of $30 pesos
if the team meets the daily output and quality
quotas. Further, after reaching the output quota,
the incentive pays an additional bonus for each
unit of output above the day’s quota. As for the
plant bonus, each worker receives a cash bonus
equal to 6%, 8%, and 10% of the total accumu-
lated earnings during the quarter. The average
daily compensation under the new incentive plan
is as follows:
Daily wage $90
Output-target bonus 40
Gain-sharing 35
Total $165 pesos
22
I use an average of 22 working days in a month to estimate
the daily piece-rate bonus.
23
The daily ?at wage earned by each worker has been constant
throughout the study period. There have been two minor
increases to the minimum wage in the country, and the ?at wage
in the factory has increased in the same proportion.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 615
Mathematical and graphical representation of the
incentive plans
Original incentive plan
Under the original incentive plan, individual
compensation equals a daily ?at wage (a), an atten-
dance bonus
, and piece-rates (b). Thus, compen-sation may be represented mathematically as
follows:
x ¼ a þ w þ bðxÞ
where x = number of units of output.
Average worker compensation is represented
graphically as follows:
I
+
?
?
? ?
New incentive plan
Team-incentive
The mathematical representation of the team-
incentive is expressed as follows:
x ¼ a þ RðXÞ À Pð/Þ
Let
rðxÞ ¼
c þ bðX
I
ÀX
T
Þ if x
i
> x
t
c if x
i
¼ x
t
0 if x
i
< x
t
0
B
@
pð/Þ ¼
rðxÞ if /
I
P/
Ã
; given x
i
0 if /
I
< /
Ã
; given x
i
where r(x) is the reward function, p(/) is a penalty
for failing to maintain the defects threshold, x total
compensation, a = a daily ?at wage, c = team bo-
nus for meeting the output-target, b = a piece-rate
factor for exceeding the output-target, x
t
= output-
target (in units of output), x
i
= actual output,
f
*
= defects threshold, f
i
= actual defects. The team
bonus takes the form of a ?at wage when a team
fails to reach either the output and quality targets.
The following are possible reward solutions when
a team exceeds the product defects threshold
(f
i
< f
*
):
ð1Þ x
i
> x
t
and /
i
< /
Ã
;
then x
i
¼ a
i
þ c
i
þ bðX
i
À X
t
Þ
ð2Þ x
i
¼ x
t
and /
i
< /
Ã
;
then x
i
¼ a
i
þ c
i
Eqs. (1) and (2) can be illustrated as follows:
i
i
<
*
I
+
I
+ (X
i
-X
t
)
i
+
i
i
X
t
X
j
?
?
?
?
?
?
?
? ?
If a team’s output (defects) falls (exceeds) the tar-
get, worker compensation equals a ?at wage (a):
ð3Þ x
i
> x
t
and /
i
P/
Ã
;
then x
i
¼ ½a
i
þ c
i
þ bðX
i
À X
t
Þ?
À ½c
i
þ bðX
i
À X
t
Þ? ¼ a
i
ð4Þ x
i
¼ x
t
and /
i
P/
Ã
;
then x
i
¼ ½a
i
þ c
i
? À c
i
¼ a
i
ð5Þ x
i
< x
t
and /
i
< /
Ã
;
then x
i
½a
i
þ c
i
?À ¼ a
i
Eqs. (3), (4), and (5) can be illustrated as follows:
I
i
*
[
i
+
i
]-
i
=
i
i i
[
i
+
i
+ (X
i
-X
t
)] – [
i
+ (X
i
-X
t
)] =
i
X
t
X
j
?
? ? ?
? ?
?
?
? ?
?
? ?
>_ ? ?
616 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
Plant-incentive
The mathematical representation of the plant-
incentive is expressed as follows:
x ¼ c½X
t
Á
^
W? þ ð1 À cÞ½/
t
Á
^
W ?
X
t
¼
:06 if x
2
> x
a
Px
1
:08 if x
3
> x
a
Px
2
:10 if x
a
Px
3
0
B
@
/
t
¼
:06 if /
2
< /
a
6 /
1
:08 if /
3
< /
a
6 /
2
:10 if /
a
6 /
3
0
B
@
where
x = total bonus,
x
t
= labor productivity quarterly targets,
x
t
= {x
1
, x
2
, x
3
},
/
t
= quality (defects rate) quarterly targets,
/
t
= {/
1
, /
2
, /
3
},
c = reward weight,
x
a
= actual plant’s productivity in the quarter,
/
a
= actual plant’s product defects in the
quarter.
References
Alchian, A., & Demsetz, H. (1972). Production, information
costs, and economic organization. American Economic
Review, 62(5), 777–795.
Allen, S. (1981). Compensation, safety, and absenteeism: Evi-
dence from the paper industry. Industrial and Labor Relations
Review, 34(2), 207–218.
Baiman, S. (1982). Agency research in managerial accounting: A
survey. Journal of Accounting Literature, 1, 154–213.
Baker, G. (1992). Incentive contracts and performance measure-
ment. Journal of Political Economy, 100(3), 598–617.
Baker, G., Jensen, M., & Murphy, K. (1988). Compensation and
incentives: Practice vs. theory. Journal of Finance, 43, 593–616.
Bandiera, O., Barankay, I., & Rasul, I. (2004). Cooperation in
collective action. The Economics of Transition, 13(3),
473–498.
Bandiera, O., Barankay, I., & Rasul, I. (2005). Social preferences
and the response to incentives: Evidence from personnel data.
The Quarterly Journal of Economics, 120(3), 473–498.
Banker, R., Field, J., Schroeder, R., & Sinha, K. (1996a). Impact
of work teams on manufacturing performance: A longitudinal
?eld study. Academy of Management Journal, 39, 867–890.
Banker, R., Lee, S., & Potter, G. (1996b). A ?eld study of the
impact of a performance-based incentive plan. Journal of
Accounting and Economics, 21(2), 195–226.
Banker, R., Lee, S., Potter, G., & Srinivasan, D. (2000). An
empirical investigation of an incentive plan that includes
non?nancial performance measures. The Accounting Review,
75, 1–65.
Barron, J., & Gjerde, K. (1997). Peer pressure in an agency
relationship. Journal of Labor Economics, 15, 917–962.
Blinder, A. (1999). Paying for productivity: A look at the evidence
(2nd ed.). Washington, DC: The Brookings Institution.
Boning, B., Ichniowski, C., & Shaw, K. (2007). Opportunity
counts: Teams and the e?ectiveness of production incentives.
Journal of Labor Economics, 25(4), 613–650.
Bonner, S., & Sprinkle, G. (2002). The e?ects of monetary
incentives on e?ort and task performance: Theories, evidence,
and a framework for research. Accounting, Organizations, and
Society, 27, 303–345.
Brown, D., & Armstrong, M. (1999). Paying for contribution.
Dover, NH: Kogan Page.
Brown, S., Fakhfakh, F., & Sessions, J. (1999). absenteeism and
employee sharing: An empirical analysis based on French
panel data, 1981–1991. Industrial and Labor Relations Review,
52(2), 234–251.
Brown, P., & Phillips, P. (1986). The decline of the piece-rate
system in California canning: Technological innovation, labor
management, and union pressure, 1890–1947. The Business
History Review, 60(4), 564–601.
Bullock, R., & Lawler, E. (1984). Gainsharing: A few questions,
and few answers. Human Resources Management, 23(1),
23–40.
Cachon, G., & Terwiesch, C. (2006). Matching supply with
demand: An introduction to operations management (1st ed.).
McGraw Hill, Irwin.
Che, Y., & Yoo, S. (2002). Optimal incentives for teams.
American Economic Review, 91, 525–541.
Chow, C. (1983). The e?ects of job standard tightness and
compensation scheme on performance: An exploration of
linkages. The Accounting Review, 58, 667–685.
DeMatteo, J., Eby, L., & Sundstrom, E. (1998). Team-based
rewards: Current empirical evidence and directions for future
research. In B. Staw, L. Cummins, (Eds.), Research in
organizational behavior (Vol. 20, pp. 141–183).
Deutsch, M. (1949). An experimental study of the e?ects of
cooperation and competition upon group process. Human
Relations, 2, 129–152.
Deutsch, M. (1973). The resolution of con?ict: Constructive and
destructive processes (1st ed.). Yale University Press.
Drago, R., & Garvey, G. (1998). Incentives for helping on the
job: Theory and evidence. Journal of Labor Economics, 16(1),
1–25.
Fisher, G., Pe?er, S., & Sprinkle, G. (2003). Budget-based
contracts, budget levels, and group performance. Journal of
Management Accounting Research, 15, 51–74.
Hansen, D. (1997). Worker performance and group incentives: A
case study. Industrial and Labor Relations Review, 51, 37–49.
Hayes, R., & Clark, K. (1985). Exploring the sources of
productivity di?erences at the factory level. In K. B. Clark,
R. Hayes, & C. Lorenz (Eds.), The uneasy alliance: Managing
the productivity-technology dilemma. Boston: Harvard Busi-
ness School Press.
Holmstrom, B. (1979). Moral hazard and observability. The Bell
Journal of Economics, 30(1), 74–91.
Holmstrom, B. (1982). Moral hazard in teams. The Bell Journal
of Economics, 13(2), 324–340.
Holmstrom, B., & Milgrom, P. (1994). The ?rm as an incentive
system. American Economic Review, 84(4), 972–991.
Horngren, C., Datar, S., & Foster, G. (2002). Cost accounting: A
managerial emphasis (12th ed.). Prentice Hall.
F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618 617
Ichniowski, C., & Shaw, K. (2003). Beyond incentive pay:
Insiders’ estimates of the value of complementary human
resource management practices. Journal of Economic Per-
spectives, 17(1), 155–180.
Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The e?ects of
human resource management practices on productivity: A
study of steel ?nishing lines. American Economic Review,
83(3), 291–313.
Indjejikian, R. (1999). Performance evaluation and compensation
research: An agency perspective. Accounting Horizons, 13,
147–157.
Kandel, E., & Lazear, E. (1992). Peer pressure and partnership.
Journal of Political Economy, 100(4), 801–817.
Knez, M., & Simester, D. (2001). Firm-wide incentives and
mutual monitoring at continental Airlines. Journal of Labor
Economics, 19(4), 743–772.
Lazear, E. (1986). Salaries and piece-rates. Journal of Business,
59(3), 405–431.
Lazear, E. (1991). Labor economics and the psychology of the
organizations. Journal of Economic Perspectives, 5(2), 89–110.
Lazear, E. (2000). Performance pay and productivity. The
American Economic Review, 90(5), 1346–1361.
Macho-Stadler, I., & Perez-Castrillo, J. (1993). Moral hazard
with several agents: The gains from cooperation. International
Journal of Industrial Organizations, 11, 73–100.
Milgrom, P., & Roberts, J. (1992). Economics, organizations, and
management (1st ed.). Upper Saddle River, NJ: Prentice Hall.
Milgrom, P., & Roberts, J. (1995). Complementarities and ?t:
Strategy, structure, and organizational change in manufac-
turing. Journal of Accounting and Economics, 19, 179–208.
Nalbantian, H., & Schotter, A. (1997). Productivity under group
incentives: An experimental study. The American Economic
Review, 87, 314–341.
Paarsch, H., & Shearer, B. (2000). Piece rates, ?xed wages, and
incentive e?ects: Statistical evidence from payroll records.
International Economic Review, 41(1), 59–92.
Petersen, T. (1992). Individual, collective, and systems rationality
in work groups: Dilemmas and market-type solutions. Amer-
ican Journal of Sociology, 98, 469–510.
Prendergast, C. (1999). The provision of incentives in ?rms.
Journal of Economic Literature, XXXVII, 7–63.
Rankin, F. (2004). Coordinating e?orts under team-based and
individual incentives: An experimental analysis. Contempo-
rary Accounting Research, 21, 191–221.
Schuster, M. (1983). Forty years on scalon plan research: A
review of the descriptive and empirical literature. In Interna-
tional yearbook of organizational democracy. Chichester: John
Wiley.
Schuster, M. (1984). The scanlon plan. A longitudinal analysis.
The Journal of Applied Behavioral Science, 20, 23–38.
Sewell, G. (1998). The discipline of teams: The control of
team-based industrial work through electronic and peer
surveillance. Administrative Science Quarterly, 43,
397–428.
Shearer, B. (2004). Piece rates, ?xed wages and incentives:
Evidence from a ?eld experiment. Review of Economic Studies,
71(2), 513–534.
Sprinkle, G. (2000). The e?ects of incentive contracts on learning
and performance. The Accounting Review, 75, 299–326.
Sprinkle, G. (2003). Perspectives on experimental research in
managerial accounting. Accounting, Organizations, and Soci-
ety, 28, 287–318.
Stiglitz, J. (1975). Incentives, risk, and information: Notes
towards a theory of hierarchy. The Bell Journal of Economics,
6(2), 552–579.
Waller, W., & Chow, C. (1985). The self-selection and e?ort
e?ects of standard-based employment contracts: A framework
and some empirical evidence. The Accounting Review, LX(3),
458–476.
Weitzman, M., & Kruse, D. (1999). Pro?t sharing and produc-
tivity. In A. Blinder (Ed.), Paying for productivity (2nd ed.).
Washington, DC: The Brookings Institution.
Welbourne, T., Balkin, D., & Gomez-Mejia, L. (1995). Gain-
sharing and mutual monitoring: A combined agency-organi-
zational justice interpretation. The Academy of Management
Journal, 38(3), 881–899.
Young, M. (1999). Field research methods in management
accounting. Accounting Horizons, 13(1), 76–84.
Young, M., Fisher, J., & Lindquist, T. (1993). The e?ects of
intergroup competition and intragroup cooperation on slack
and output in a manufacturing setting. The Accounting
Review, 68(3), 466–481.
618 F.J. Roma´ n / Accounting, Organizations and Society 34 (2009) 589–618
doc_658502700.pdf