Kim-Burton2002 Article TheEffectOfTaskUncertaintyAndD

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Computational & Mathematical Organization Theory, 8, 365–384, 2002


c 2003 Kluwer Academic Publishers. Manufactured in The Netherlands

The Effect of Task Uncertainty and Decentralization


on Project Team Performance∗
JISUNG KIM AND RICHARD M. BURTON
Fuqua School of Business, Duke University, 2138 Campus Drive, Durham, NC 27708-6245, USA
email: jk19@duke.edu
email: rmb2@duke.edu

Abstract
This paper investigates relationships among task uncertainty, level of centralization, and project team performance.
Team performance is measured in three dimensions: cost, time, and quality. Adopting an information processing
view and contingency theory, the authors discuss tradeoffs among the three performance dimensions of a project
team. Results from the simulation study indicate that, under high task uncertainty, a decentralized team performs
better in terms of cost and time, but a centralized team performs better in terms of quality. Under low task
uncertainty, there is no performance difference between a centralized team and a decentralized team in terms of
cost and time, but a centralized team performs better in terms of quality. The paper suggests that researchers pay
attention to the relative impact of centralization and decentralization on different dimensions of organizational
performance, and managers adopt an organizational structure that performs better in a performance dimension
that counts more to enhance overall performance.

Keywords: organizational design, decentralization, task uncertainty, quality

1. Introduction

Galbraith (1973) views an organization as an information processing system and proposes


that the greater the uncertainty of the task, the greater the amount of information that
has to be processed between organizational members. When task uncertainty is high, a
person who is responsible for completing the task will need to exchange more information
with other people to solve problems than a person with a low uncertainty task. Adopting
contingency theory, Galbraith (1973) argues that the fit between organizational structure and
task uncertainty would lead to higher performance. According to Donaldson (1996, p. 59),
a core tenet of structural contingency theory is that, for low task uncertainty, a centralized
structure will perform more effectively, while participatory, a decentralized structure is
required as task uncertainty increases.
However, studies testing contingency theory showed some mixed results and led to the
criticism of the theory (Pfeffer, 1997; Schoonhoven, 1981). In this paper, we propose that the
mixed results may be due to the lack of clarity in the theoretical statements (Schoonhoven,
1981, p. 350) and the definition of organizational performance. Using a computational
simulation method, we explore the relationships among the level of centralization in decision

∗An earlier version was presented at Computational Social Science 2000 Conference, July 2000.
366 KIM AND BURTON

making, task uncertainty, and project team performance in terms of project duration, costs,
and quality. Especially, we expect that there are tradeoffs between different dimensions of
performance. Although researchers have examined the effects of uncertainty on strategic
behavior and performance of organizations or groups, little study has examined tradeoffs
among different dimensions of organizational performance with regard to uncertainty and
the level of centralization (Carley, 1995; Murray, 1989).
The computational simulation method is a useful tool for examining the performance
of simulated organizations with various characteristics of an organization such as task,
structure, and size (Carley, 1995; Jin and Levitt, 1996). Studying organizational theory with
a simulation method has merits: (1) researchers can measure variables more precisely and
at the same time hold other parameters constant; (2) researchers can conduct multiple tests
including sensitivity analyses; and (3) researchers can test a wide range of organizational
variables in which they are interested (Lin and Carley, 1997). We add new insights to
current literature on contingency theory by investigating the relationships among the level
of centralization in decision making, task uncertainty, and different dimensions of project
team performance.
In the next section, we review theoretical backgrounds related to contingency theory and
organizational performance. Then, we explain the method used in this paper, our base model
and variables, and experiments conducted. Following are results of the experiments and
discussion on the relationships among task uncertainty, centralization of decision making,
and performance. We conduct sensitivity analyses by varying the level of task uncertainty,
and robustness tests by changing the sequence of tasks. Finally, we develop propositions
for future research and conclude with discussion.

2. Background

How does task uncertainty affect organizational performance with respect to different orga-
nizational structures? Is a decentralized structure more effective than a centralized structure
under high task uncertainty? The effect of task uncertainty and organizational structure
on performance has been a popular topic in the area of management, and most studies
have been conducted under the paradigm of contingency theory (Burton and Obel, 1998;
Donaldson, 1996; Galbraith, 1973; Lawrence and Lorsch, 1967; Thompson, 1967). Con-
tingency theory argues that the best way to organize is contingent upon factors such as task
uncertainty, environmental uncertainty, size, and technology (Donaldson, 1996; Lawrence
and Lorsch, 1967). Regarding the relationship between environmental or task uncertainty
and organizational structure, a general statement drawn from contingency theory is that,
under low uncertainty, a centralized organizational structure is more effective, while with
high uncertainty, a decentralized structure performs better.
However, empirical studies on proposed relationships between environmental or task un-
certainty and organizational structure produced mixed results and increased the necessity
of a more refined theory. For example, Keats and Hitt (1988) find that environmental uncer-
tainty affects firms’ diversification and divisionalization, and those changes also affect firm
performance. Bourgeois (1985) also finds that congruence between perceived environmental
uncertainty and volatility is positively related to firms’ financial performance. On the other
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 367

hand, Mohr (1971) and Pennings (1975) do not find supports for “consistency hypothesis,”
that predicts that task uncertainty is related to the structure of an organization. Schoonhoven
(1981) argues that contingency theory has several theoretical weaknesses. She points out
that: (1) contingency theory lacks clarity in its theoretical statements; (2) contingency theo-
rists need to consider interaction effects explicitly; (3) contingency theory needs to address
specific form of the interaction; and (4) analytic models used in contingency theory tend to
assume linear relationships among contingency factors. In this study, we address these issues
more precisely in the computational model and experiment by developing clear statements
about the relations among task uncertainty, centralization, and performance, and further
examining multiple performance measures and their tradeoffs. We argue that refining a per-
formance variable into different dimensions would help clarify some theoretical statements
of contingency theory.
The effects of organizational structure and uncertainty on organizational performance
have been studied in the context of computational simulation with various structures and
performance variables (Chang and Harrington, 1998; Lin and Carley, 1997; Phelan and Lin,
2001; Wong and Burton, 2001). Lin and Carley (1997) show that less costly/complex orga-
nizational structures perform better on decision accuracy, a performance variable; and, time
pressure was negatively associated with decision accuracy. When organizations tried to be
timely, the accuracy of decisions decreased, implying a tradeoff between the timeliness and
accuracy. They also find that, when time pressure was high, more costly/complex organiza-
tional structures performed better than less costly/complex organizational structures. This
may indicate that when there is less time for information distortion and ambiguity to occur,
the more complex organizational structure, with more information exchange, makes better
decisions. Using the rate of innovation as a performance variable, Chang and Harrington
(1998) simulated the effect of centralization and decentralization in a retail chain. They
find that, under medium level of innovative opportunities, which is defined as generation
of innovative ideas, a centralized retail chain performed better, while, with high level of
innovation opportunities, a decentralized structure performed better. It may imply that as
environment becomes less favorable, in other words, as the level of innovation opportuni-
ties decreases, a centralized structure will perform better in terms or innovation generation.
Like these studies, this paper also uses a computational simulation method in an effort
to precisely and clearly investigate the relationships among the level of task uncertainty,
centralization in decision making, and performance. We consider both quality and cost
aspects of organizational performance and address the tradeoffs among the performance
variables.
Although contingency studies can be done at any level, we choose group as the level of
analysis as the relationship between structure and performance is less influenced by other
factors such as strategy, products, and market conditions. Large firms often have multiple
businesses and overall performance is affected by the performance of individual business
as well as organizational structure. However, we believe that insights from the study can be
applied to organizational level issues. We revisit this issue in the discussion and conclusion
section.
The organizational performance at a group or team level has been studied with a focus on
group decision making and productivity (Levine and Moreland, 1996). For group decision
368 KIM AND BURTON

making, researchers find that the more information group members exchange, the better
the group decision making quality, and the more group members are informed about each
member’s expertise, the more the information is exchanged (Gigone and Hastie, 1993;
Stasser and Stewart, 1992; Stasser et al., 1995). Research on group productivity is concerned
with characteristics of group members, structural aspects of group, and task and situational
characteristics (Levine and Moreland, 1996). For example, researchers find that group
effectiveness was associated with members’ skill level and the match between skills and
tasks (Hackman, 1987), coordination and motivation (Steiner, 1972), group goals setting
(O’Leary-Kelly et al., 1994), and job characteristics and interdependence (Campion et al.,
1993).
Group productivity or effectiveness is defined, in part, as the quantity or quality of outputs
(Guzzo and Dickson, 1996). However, researchers tend to choose one of group outputs or
aggregate those outputs into an overall performance measure. We argue that organizational
performance can be measured in different dimensions, and most common dimensions of
performance in management research are time, cost, and quality (Jayaram, 1999; Ward
et al., 1998). For a project team, the duration of the project, the costs incurred during the
project, and the quality of the final outcome could be three major performance aspects (Jin
and Levitt, 1996).

3. Computational Simulation Method

We use a computational simulation method to explore relationships among task uncer-


tainty, structure, and performance. The simulation study can be viewed as a virtual ex-
periment which is useful when the problem is not tractable and a researcher wants a gen-
eral sensitivity analysis to induce predictions for organizational behavior (Carley, 1995).
Similarly, it is also useful when a researcher explicates the important tradeoffs in alter-
native design possibilities (Kumar et al., 1993). The simulation method has become an
important tool for exploring and testing the impact of contingency factors on organiza-
tional performance (Carley and Lin, 1997; Jin and Levitt, 1996; Lin and Carley, 1997;
Phelan and Lin, 2001). Using the simulation study, we can get a general prediction on how
team performance is affected by the level of centralization in decision structure and task
uncertainty.
A project team simulated in this study is a typical strategic planning consulting project
team. We chose the consulting project team because it is a popular form of a project team,
produces outputs that can be measured in terms of the three dimensions of performance:
timeliness, cost, and quality, and it fits our definition of a team as an information processing
system. We have been involved in several consulting projects previously, so the process
of the project was constructed by us. This consulting process has been used by a leading
consulting firm and its variations are widely adopted by other consulting firms (see figure 1).
In the consulting project, there is a specific set of tasks and members are assigned roles such
as a project manager or a member. Consultants process information and make decisions
to accomplish the tasks. A consulting project is evaluated by the quality of a final report
and the observance of project schedule which affects the cost of the project. The consulting
project starts with client task-force training and interviews with client managers, and ends
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 369

Figure 1. Strategy building consulting process.

with a final report. A project manager and team members are assigned tasks for which they
are responsible.
The simulation program used in this study is the SimVision R
(Vité Inc., 2002), a com-
mercial software version of the Virtual Design Team (VDT).1 The VDT incorporates an in-
formation processing view and simulates project-based organizations. It has been validated
internally by testing systematic process using small projects and theoretical predictions, and
externally by using observable real project information (Jin and Levitt, 1996). Researchers
have used the VDT to examine relative impacts of organizational characteristics on per-
formance (Wong and Burton, 2001). SimVision R
is designed to simulate project-based
organizations or teams in which members conduct tasks to finish a project. The results
one can obtain from using the SimVision R
include information such as the predicted time
to complete a project, the total costs incurred to do the project, and several measures of
process quality. Therefore, the VDT is a useful tool for testing the effects of alternative
organizational structures and task characteristics on different dimensions of organizational
performance (Jin and Levitt, 1996).
Jin and Levitt (1996) provide detailed discussions of the VDT model. VDT assumes
that actors in a simulation model are not perfect in processing information, i.e. bounded
rationality. As March and Simon (1958) proposed, all alternatives and the consequences
of them are not known. Thus, actors allocate their attention and time to work items based
on certain allocation rules. One of general behavioral assumptions of VDT is that higher
level actors are more likely to understand the need for rework than lower level actors.2 In
our model, a project manager is assumed to have more information processing capacity and
make higher quality rework decisions than project members. However, if a project manager
becomes backlogged or overloaded with tasks beyond his or her information processing
370 KIM AND BURTON

capacity, the decision making is delegated to lower level actors. So, the outcome of a project
depends on the centralization of decision making, the number of errors and exceptions that
need to be handled, and the structure of a project organization.

3.1. Modeling the Project Team in SimVision


R

The performance effects of task uncertainty and the level of centralization are investigated
using the consulting project team. There are three performance variables: project dura-
tion, project costs, and project quality. The three performance measures at each level of
task uncertainty and centralization are recorded and compared. The results show how task
uncertainty affects project team performance differentially depending on the level of the
centralization in the team decision-making structure. They will also show whether there are
tradeoffs among the three performance variables.

Project Team Performance. The performance variables include: project duration, project
costs, and project quality. The project duration is defined as the time that the entire project
is expected to take. The project costs is defined as the amount of money that the entire
project is expected to cost. The project quality is defined as the likelihood that components
produced by the project have defects based on rework and exception handling and measured
by “Functional Quality Risk.” The Functional Quality Risk contains information on the level
of risk that errors or exceptions might not be reworked in each task. The quality of a project
depends on how many of the errors or exceptions are reworked or redone. Thus, high
functional quality risk indicates low subsequent project quality. In the SimVision R
, these
variables are obtained from a project summary statistics report.

Centralization. The centralization is defined as the level of position at which a decision-


making takes place. With high centralization, more decisions will be made at the highest
level of the team—here the project manager (March and Simon, 1958). Thus, necessary
information regarding errors or exceptions will be transferred to the project manager, in-
creasing the number of information exchanges. With low centralization, the decisions will
be made by a project team member, decreasing the number of information exhanges. In
other words, with low centralization, team members can make their own decisions, without
reporting errors or exceptions to a higher level manager.

Task Uncertainty. The task uncertainty is defined as the extent to which necessary infor-
mation is available at the beginning of the task and the likelihood that a task will fail and
require rework. There are two reasons that we chose this definition. First, researchers have
adopted Perrow’s (1970) routine-nonroutine work characteristics in conceptualizing task
uncertainty, and it has been suggested that the routine-nonroutine continuum is an aggre-
gation of two task characteristics: task variety and task analyzability (Daft and Macintosh,
1981; Victor and Blackburn, 1987). Task variety means the number of exceptions or prob-
lems in the task, so high task variety increases the likelihood of errors or defects in the final
products. Task analyzability is the difficulty of searching solutions to problems, and high
task analyzability requires increased information exchange. The second reason we chose the
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 371

Table 1. Summary of variables.

Variables Definition Measure Values

IVs Centralization The level of position where Level of centralization High centralization
decision making takes place Low centralization
Task uncertainty The extent to which information Uncertainty at task level High
needed to complete a task Medium
is available at the time that Low
the task starts.
The likelihood that a task Functional error 0.1 (High)
will fail and require rework. probability at 0.05 (Medium)
task level 0.01 (Low)
DVs Project duration Time taken to finish a project Simulation project duration In weeks
Project cost Money spent to finish a project Simulation project cost In dollars
Project quality Likelihood that there are Functional quality risk In percentage
uncorrected defects based on
rework and exception handling

above definition of task uncertainty is that since our model simulates a team with not only
“choose” tasks (e.g. solving problems with correct answers) but also “execute” tasks (e.g.
executing performance tasks) (McGrath, 1984), errors and defects in the task resulting from
task exceptions as well as the lack of necessary information are included in the definition
of task uncertainty (Galbraith, 1973). Thus, in the SimVision, task uncertainty is measured
from both “Functional Error Probability” and “Task Uncertainty” at each task level. Table 1
summarizes the definition, measure, and value of each variable.

3.2. Experiments

We ran the simulation experiment with different levels of team decision structure and task
uncertainty. Each member of the team is assigned the role of project manager (PM) or sub-
team (ST) as indicated in the figure 1. The simulation teams were assigned team decision
structure, either high centralization or low centralization (i.e. decentralization), and task
uncertainty that involves either low, medium, or high task uncertainty. For low task uncer-
tainty, the functional error probability was set to 0.01 and the individual task uncertainty
was set to low. For medium task uncertainty, the functional error probability was set to 0.05,
and individual task uncertainty was set to medium. 0.1 of functional error probability and
high individual task uncertainty were assigned to high task uncertainty (normal range for
the functional error probability in the SimVision R
is 0 to 0.1) (see Table 1). There were
six sets of simulation runs, the first two sets involving high task uncertainty with either
high centralization or low centralization team structure. The next two sets of simulations
involved medium task uncertainty and either high centralization or low centralization team
structure, and the last two sets had low task uncertainty for either centralized or decentral-
ized structure. For each set of simulation, we ran 100 runs and results showed the mean and
standard deviation of the 100 simulation runs.
372 KIM AND BURTON

4. Results

4.1. Base Model Results

Table 2 summarizes the results of simulated consulting projects. Figures 1 to 3 show the
results graphically for each dimension of project team performance. Under low task un-
certainty, both centralized and decentralized teams showed same performance in project
duration and cost. In terms of project duration, it took, on average, 6.8 weeks for both the
centralized and decentralized team to finish the project. The centralized team spent $45,600
and the decentralized team spent $45,500 to finish the projects; there was no significant
difference between the two amounts. However, the functional quality risk was 0.34 for the
centralized team and 0.49 for the decentralized team. Since the functional quality risk is
the level of risk that errors or exceptions might not be reworked in each task, the lower the
risk, the higher the project quality is. Thus, the centralized project team performed better in
terms of project quality under low task uncertainty ( p < 0.01). Overall, a centralized project
team performed better as it yielded better quality without increased time or costs.
Under medium task uncertainty, the performance results are mixed. The decentralized
project team showed significantly ( p < 0.01) better performance in terms of project duration
and costs than the centralized team, but showed worse performance in project quality.
The decentralized team finished the project in 7.2 weeks (SD = 0.2 weeks), spent $48,000
(SD = $800), and showed 0.53 (SD = 0.1) of functional quality risk, which means that the
likelihood that errors or defects might not have been reworked was 53%. On the contrary,
the functional quality risk for the centralized team was 0.29 (SD = 0.082). Project duration
and costs were 7.4 (SD = 0.3) weeks and $49,100 (SD = $1,140) each for the centralized
team. The differences in time, cost, and quality were significant (n = 100, p < 0.01). Here an
appropriate choice of decentralization involves a tradeoff in the choice of which performance

Table 2. Task uncertainty, decision structure, and performance.

Low task uncertainty Medium task uncertainty High task uncertainty


Performance Centralized Decentralized Centralized Decentralized Centralized Decentralized

Duration (weeks) 6.8 6.8 7.4 7.2 8.3 8.0


Standard dev (0.1) (0.1) (0.3) (0.2) (0.4) (0.4)
Two sample test Centralized = Decentralized Centralized > Decentralized Centralized > Decentralized
(n = 100) ( p < 0.01) ( p < 0.01)
Costs ($1,000) 45.6 45.5 49.1 48 55.2 53.2
Standard dev (0.48) (0.46) (1.14) (0.8) (1.55) (1.33)
Two sample test Centralized = Decentralized Centralized > Decentralized Centralized > Decentralized
(n = 100) ( p < 0.01) ( p < 0.01)
Quality 0.34 0.49 0.29 0.53 0.29 0.53
Standard dev (0.141) (0.196) (0.082) (0.1) (0.063) (0.074)
Two sample test Centralized < Decentralized Centralized < Decentralized Centralized < Decentralized
(n = 100) ( p < 0.01) ( p < 0.01) ( p < 0.01)
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 373

dimension is the more important: project quality which suggests centralization, or project
time and costs which suggest decentralization.
Under high task uncertainty, the same tradeoffs exist and the choice of which performance
dimension remains important. The decentralized team significantly (n = 100, p < 0.01) per-
formed better in project duration and costs. It took 8.0 (SD = 0.4) weeks for the decentralized
team to finish the project, whereas it took 8.3 (SD = 0.4) weeks for the centralized team to
finish the project. Project costs incurred were $53,200 (SD = $1,330) for the decentralized
team and $55,200 (SD = $1,550) for the centralized team. The centralized team significantly
(n = 100, p < 0.01) performed better in terms of project quality. The functional quality risk
was 0.29 (SD = 0.063) for the centralized team and 0.53 (SD = 0.074) for the decentralized
team.3
In summary, the results showed that, under low task uncertainty there was no perfor-
mance difference on time and cost between the centralized and decentralized team, but the
centralized team performed better on quality. Under medium and high task uncertainty, the
decentralized team performed better in terms of time and cost, while the centralized team
did better in terms of quality (see figure 2). The maxim that under high task uncertainty, de-
centralization will yield better performance should be clarified. If the project time and costs
are the primary concern, then it is confirmed. But if project quality is more important, then
it is not confirmed. The contingency statement is general; it depends upon the performance
measure.
Based on the results of our computational simulation study, we develop a set of proposi-
tions that can be further investigated and tested in future research. Galbraith (1973), in the
discussion of contingency theory and organization design issues, viewed organization as
an information processing system. In the information processing view, one assumption is
that information processing is associated with costs. Applying this view to a project team,
the more information processing a project requires, the more costly the project becomes.
Information processing is also associated with time. The duration of a project will get longer
as more information exchanges are required by team members.
Project teams can structure themselves in different ways so that they can control the
amount of information exchange needed to accomplish the projects. With a centralized
structure, decision making occurs at the top of the team (e.g. project manager) and the
project manager requires all the information to make decisions on what happens at a lower
level. On the other hand, with a decentralized structure, decision making is delegated to
a lower level of the team (e.g. team member) so that team members can make their own
decisions without having to consult with the project manager (March and Simon, 1958).
Information necessary to make decisions does not have to be communicated with the project
manager (Jin and Levitt, 1996; Levitt et al., 1994).
Viewing an organization as an information processing system which is structured to
achieve a specific set of tasks, and comprised of information processors (Galbraith, 1973;
Jin and Levitt, 1996; Levitt et al., 1994), we propose that, under high task uncertainty, a
decentralized decision making helps a project team finish their project quicker than a team
with a centralized decision making.

Proposition 1. Under high task uncertainty, the project duration will be shorter in a
decentralized team than in a centralized team.
374 KIM AND BURTON

Figure 2. Task uncertainty, decision structure, and performance.


EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 375

We also propose that, under high task uncertainty, a decentralized team will finish the project
with lower costs than a centralized team because the former will need less information
processing than the latter when unexpected problems occur due to high task uncertainty,
and information processing is associated with costs.

Proposition 2. Under high task uncertainty, the project costs will be lower in a decen-
tralized team than in a centralized team.

For the time and cost dimensions of performance of a project team, we propose that
a decentralized team would perform better than a centralized team under high task un-
certainty. However, for the quality of a project, we propose that a centralized team,
under high uncertainty, would have higher quality than a decentralized team. A central-
ized project team would deal with the problem of neglecting errors better because de-
cisions are made by a project manager who “ensures coordinated and controlled action
(Pfeffer, 1978, p. 53).” A centralized project team would also exchange more informa-
tion than a decentralized team, and more information exchange would lead to higher
quality or accurate decision making (Gigone and Hastie, 1993; Lin and Carley, 1997;
Stasser and Stewart, 1992). Weick and Roberts (1993) argue that a group with a “col-
lective mind” show higher reliability and fewer errors, and explain how a group devel-
ops the collective mind in the study of flight operations on aircraft carriers. Although
the centralization of decision making and the “collective mind” do not exactly match,
when group members follow a centralized decision making, the group behaves as if it
has a “collective mind.” Under high task uncertainty, each project member facing high
task uncertainty may experience more exceptions and errors that have to be corrected
than under low task uncertainty. Project members will spend more time on fixing the er-
rors and redoing the tasks. Since project members are less skilled and less responsible
compared with a project manager, decisions and rework made by the project members
will end up with lower quality than decisions and rework made conducted by a project
manager.

Proposition 3. Under high task uncertainty, the project quality will be higher in a cen-
tralized team than in a decentralized team.

In summary, high task uncertainty will negatively affect all three performance variables.
However, the negative effect of high task uncertainty on project duration and costs will be
smaller for a team with decentralized decision structure than for a team with centralized
structure, and the negative effect on project quality will be smaller for a team with centralized
decision structure than for a team with decentralized structure. On the other hand, low task
uncertainty will positively affect all three performance variables. However, the positive
impact of low task uncertainty on project duration and costs will be smaller for a team with
decentralized decision structure than for a team with centralized structure, and the positive
effect on project quality will be smaller for a team with a centralized decision structure than
for a team with a decentralized structure.
376 KIM AND BURTON

4.2. Exploring the Limits: Extreme Task Uncertainty and Skill Level

In this section, we explore two dimensions that relate to the choice of decentralization.
First, does an extreme level of task uncertainty yield an unambiguous preference for decen-
tralization/centralization? And secondly, will higher levels of project member skill yield
unambiguous preferences? We replicated the test focusing on two different contingency fac-
tors: extremely high uncertainty, and project member skill level. The effect of each factor
was tested on each base model and the results are reported.
From the baseline consulting model, we changed the task uncertainty level to an ex-
tremely high level which is beyond the normal range SimVision R
recommended (0 to 0.1).
We incrementally increased Functional Error Probability to see whether the base model
results hold. As the rate increased, the duration and cost performance difference between
centralization and decentralization decreased, and when the Functional Error Probability
reached 0.5, the centralized team performed better in all three performance dimensions (see
Table 3). The task uncertainty was so high that SimVision R
generated warning signs for
several tasks that they reached activity verification limit. We call those levels of uncertainty
extreme uncertainty. Under extreme uncertainty, a project team with decentralized deci-
sion making did not perform as well as a project team with centralized decision making
in terms of duration and cost, which is the opposite of the baseline model. Quality was
consistently better in a centralized project team. This may indicate that there are limits to
the performance effect of decentralization with regard to task uncertainty. We need to be
cautious about adopting a decentralized structure, because a decentralized organization may
not function properly under extreme uncertainty.
Next, we replicated our study with different levels of project member skills. We wanted
to investigate the extent the bounded rationality of the project members drive the results.
Put differently, will less bounded project team members yield different results and thus
choices on decentralization/centralization? With high skill levels, project team members
will produce fewer errors and mistakes and they will have less difficulty in dealing with
the errors and mistakes, leading to high quality performance (Safizadeh, 1991). We ask if
the extent to which performance was enhanced would be different between centralized and
decentralized teams. We left the task uncertainty level as high, which is with Functional
Error Probability 0.1, project member skill levels as high, and compared three performance
dimensions between a centralized team and a decentralized team. Results shown in Table 3
indicate that the relative performance of each performance dimension is the same as our
baseline model results. Under high uncertainty, with highly skilled team members, a team
with decentralized decision making performed better in terms of duration and cost, but a
team with centralized decision making performed better in terms of project quality. Thus,
our replication analyses showed some external validity within the simulation context.

4.3. Robustness

To explore the validity of the results, we tested the base model with more sequential tasks.
The model used in the test is shown in figure 3. The number of project team members
and their skill levels were the same as the baseline simulation model. But we changed the
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 377

Table 3. Performance under extreme task uncertainty or high skill level.

Extreme task uncertainty from base High skill level from base model
model (Functional error prob = 0.5) (Functional error prob = 0.1)
Performance Centralized Decentralized Centralized Decentralized

Duration (weeks) 12.7 13.5 4.4 4.2


Standard dev (0.7) (0.8) (0.16) (0.04)
Two sample test (n = 100) Centralized < Decentralized ( p < 0.01) Centralized > Decentralized ( p < 0.01)
Costs ($1,000) 79.4 82.7 29.4 28.2
Standard dev (0.14) (0.34) (0.22) (0.8)
Two sample test (n = 100) Centralized < Decentralized ( p < 0.01) Centralized > Decentralized ( p < 0.01)
Quality 0.28 0.54 0.30 0.53
Standard dev (0.037) (0.034) (0.081) (0.097)
Two sample test (n = 100) Centralized < Decentralized ( p < 0.01) Centralized < Decentralized ( p < 0.01)

Figure 3. Strategy building consulting process (test of sequential process).

sequence of tasks to more sequential from a parallel sequence of the baseline model. The
sequence of tasks could change the coordination and interaction among team members
affecting the impact of task uncertainty on project team performance (Thompson, 1967).
The results showed that the relative performance of each dimension was the same as the
previous results. Under medium and high task uncertainty, a decentralized team performed
better than a centralized team in terms of project duration and costs, but performed worse in
378 KIM AND BURTON

terms of project quality. Under low task uncertainty, there was no performance difference
between the two teams in terms of duration and cost but the centralized team performed
better in terms of quality, same as the previous results.

5. Discussion and Conclusion

As Burton and Obel (1995) suggest, the current simulation study adopted a simple model,
but it addresses interesting questions that may provide insights to both organization re-
searchers and practicing managers. Although this study does not examine comprehensive
relationships among contingency factors that affect organizational performance, it tests im-
portant relationships among critical variables that have been at the central discussion of
the contingency theory. Task uncertainty affects team performance negatively by increasing
project duration, costs, and quality risk. However, our results show that the degree of the
negative effects of task uncertainty depends on performance dimensions and the level of
centralization in a team’s decision making structure. A decentralized team suffered less, as
task uncertainty increases, in terms of project duration and costs, but more in terms of project
quality, than a centralized team. Under high task uncertainty, a decentralized team finished
a project faster with less money spent on the project than a centralized team, whereas the
centralized team finished a project with higher quality. This is similar to the findings of
Lin and Carley (1997) that an organization can be either timely or accurate. The results
provide several implications for organization theory and managerial practices regarding the
relationships among organizational structure, task uncertainty, and performance.
Contingency theory states that the fit between organizational structure and task uncer-
tainty is important, and further it argues that, under high task uncertainty, adopting a de-
centralized structure leads to high performance. However, these statements lack clarity and
precision (Schoonhoven, 1981). The current study suggests that adopting a decentralized
structure may not lead to high performance because a decentralized structure may produce
poorer quality than a centralized structure. For example, despite advocated merits of de-
centralization in a retail chain, Chang and Harrington’s (1998) find that when innovation
opportunities are not very high, a centralized retail chain produced more innovations (e.g.
good quality ideas) than a decentralized retail chain. Since there is a tradeoff between costs
and quality under high task uncertainty, it is necessary that researchers specify in which
dimension of performance one form of decision structure will perform better than the other.
For example, for systems that require “collective mind (Weick and Roberts, 1993),” “high
reliability organization (Roberts, 1990),” or in the market where demand elasticity is low and
consumers want reliable products, quality performance dimension is much more important
than cost or speed performance and should get more weight when considering organizational
structure. Noting that research on effectiveness has neglected the reliability performance of
an organization, Roberts (1990) emphasizes the role of redundancy, which is consistent with
centralization of decision making, in monitoring and controlling unexpected interactions in a
high reliability organization. On the other hand, for systems in “high-velocity environments
(Eisenhardt, 1989),” “hyper competitive market (D’Aveni, 1994),” or in the market where
demand elasticity is high and consumers buy low-cost products, speed or cost performance
dimension should get more attention when matching organizational structure. Eisenhardt
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 379

(1993) discusses the differences and similarities between “high reliability organizations”
and organizations in “high velocity environment,” and suggests that their priorities are dif-
ferent: high reliability organizations with reliability and high velocity organizations with
speed. Thus, when researchers test contingency theory, they should refine organizational
performance into more specific dimensions. One way of doing this could be using speed,
cost, and quality as performance variables instead of having the overall profitability of a
firm or effectiveness of a group as a performance variable. Another way of conducting
contingency study would be finding other performance factors that mediate contingency
factors and overall effectiveness, and using it as a performance variable.
Task uncertainty negatively affects three team performance dimensions, regardless of
team structure. If the organization structure is assumed as given, then improving perfor-
mance will require decreasing task uncertainty. Our analyses indicate that once an organi-
zation can manage the uncertainty to a very low level, there is little performance difference
between centralized and decentralized structure in terms of time and cost and the organi-
zation could pursue high performance in both quality and speed. Galbraith (1973) suggests
several organization design strategies for reducing uncertainty: rules and programs, hier-
archical referral, goal setting, creation of slack resources, creation of self-contained tasks,
investment in vertical information systems, and creation of lateral relations. New technol-
ogy could increase uncertainty (Donaldson, 1996), but technology could also be used in
decreasing uncertainty. The use of Flexible Manufacturing System (FMS) or Just-In-Time
(JIT) system can be viewed as a strategy for decreasing uncertainty to achieve both high
quality and speed.
Under high task uncertainty, the performance difference in project duration and cost
between a centralized and a decentralized team was smaller than the difference in project
quality. According to the information processing approach, the poor performance of a
centralized team under high task uncertainty is due to the inability of a supervisor to handle
all the necessary information including decision-making on errors and exceptions that were
referred from the subordinates (Galbraith, 1973). In our study, although the centralized team
was suffering from the high volume of information exchange under high task uncertainty,
a project manager was able to handle information and errors that were referred. Simulation
results showed that no one had backlogged tasks, including a project manager. Since a
centralized decision making structure will burden a manager with decisions on tasks from
subordinates, it is important that the manager would not be overburdened with the tasks. As
long as a project manager is not backlogged with too many tasks to take care of, the results
in this study will hold. The smaller difference on project duration and cost than on project
quality between a centralized and a decentralized team may indicate that when a supervisor
has abilities and time to handle information and tasks, a centralized team might do as well
as a decentralized team on speed and cost dimensions of performance. In a market where
all three dimensions of performance are equally critical to win customers, an organization
may be better off to adopt a centralized structure with a highly competent manager.
We find that, under extreme uncertainty, a centralized team performed better in all three
performance dimensions. This kind of uncertainty may be reflected in the case of a crisis
situation. Crisis management researchers (Shrivastava and Mitroff, 1987) define crisis as an
economic and technical failure that threatens a company’s goal of profitability and survival.
380 KIM AND BURTON

Under a crisis situation, information needed to solve the problem will be hard to obtain
and the likelihood that errors or mistakes may occur will be extremely high. Shrivastava
and Mitroff (1987) argue that an organization needs to be prepared for a crisis situation
with a more centralized structure. Lin and Carley (2003) find similar results in a research
on the effect of, among other factors, external stress, which originates from outside of an
organization and organizational members are not familiar with. Under these conditions,
they found that, hierarchy structure consistently made better decisions than other structures
such as a team with a manager or a team with voting structure.
The study provides some practical implications for managers. At an organization level, the
overall performance of the organization can be measured in terms of financial performance
(e.g. net income, return on assets, return on investment, and return on equity). The financial
performance, in general, is the overall performance of an organization reflecting costs, speed,
and quality performance. Then what factors should managers consider in finding the better
structure that will lead to higher financial performance under high task uncertainty? It may
depend on which dimension of performance is more valuable to customers. If customers
value low costs more than high product or service quality, a decentralized organization
will have higher financial performance than a centralized organization. However, when the
quality of a product or service is more important to customers, a centralized organization will
perform better. For example, Gittel (2000) find that small supervisory span of control and an
emphasis on supervisory role on front-line employees let Southwest airline achieve highly
coordinated and on-time flight departure, which is one of the most important performance
criteria for airline companies, and one of the most difficult tasks given high uncertainty and
interdependence. On the other hand, under low task uncertainty, there is little performance
difference between the two structures. The tradeoff becomes more salient as task uncertainty
increases. This may indicate that managers can pursue both cost performance and quality
performance at the same time when the task uncertainty is low. In industries under low task
uncertainty such as consumer electronics or automobile industry, manufacturers may find it
possible to reduce costs while maintaining high quality of their products. Further, managers
under high uncertainty environment can pursue quality, speed, and cost performance at the
same time by adopting uncertainty reduction strategies as suggested by Galbraith (1973).
Contingency theory addresses specific relationships among factors such as structure,
task, size, technology, and performance (Donaldson, 1996). Our analyses show that these
relationships can be tested with robustness if researchers measure contingency factors more
precisely and clearly at the same time minimizing the effect of factors that are not in
the main focus. Simulation methodology can increase the precision of measurement since
each variable in the study can be quantified and even an extreme case can be obtained
without much trouble. In this paper, we were able to manipulate task uncertainty levels
from very low to extremely high, which is hard to observe in other research contexts.
Carley (1995) notes that computational and mathematical research, of which a simulation
method is a part, is increasingly adopted in developing and testing organizational theory,
especially in four areas: organizational design, organizational learning, organizations and
information technology, and organizational evolution and change. Another way of achieving
precision and clarity in developing and testing contingency theory may be using a group
as the level of analysis. The performance at an organization level, especially large, highly
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 381

diversified firms, can be affected by many factors, and the effects of contingency factors
researchers are interested in could be confounded with the effects of other factors, resulting
in mixed empirical findings. At a group or team level, however, researchers can observe the
effectiveness, performance, and dimensions of performance more directly. Controlling for
other factors is also easier at a group level than at an organization level.
We note that there are some limits related with the current simulation model. Addressing
these limits will enable readers to interpret the results of the current analysis with caution and
will provide some future research ideas. We tried to make the model as simple as possible
to minimize the effect of factors that are not in major interest. Thus, although the current
model is useful in exploring the relationships among task uncertainty, level of centralization,
and different performance dimensions with high precision and clarity, it leaves out other
task and contingency factors that may play a role in the proposed relationship. We did
not consider, for example, what McGrath (1984, p. 61) calls “generate” (e.g. generating
ideas and plans) and “negotiate” (e.g. resolving conflicts of interest) tasks. The choice of
performance variables and their relationships would depend on which task characteristics are
included in the definition of task. Other contingency factors in organizational design such as
technology, size, types of skills and training, types of rewards, and horizontal information
exchange would also affect organizational performance (Burton and Obel, 1998; Nadler
and Tushman, 1988). Investigating the relationships among these factors and performance
dimensions would lead to a more comprehensive model of organizational effectiveness.
Quantifying variables with a computational model enables researchers to conduct study
otherwise impossible or difficult, but at the same time it can distance the findings of the
study from actual processes or meanings. Therefore, testing the propositions we suggested
from this study in other contexts such as experiments or field study would further enhance
our understanding of uncertainty and performance and advance the contingency theory.

Acknowledgment

We thank Professor Ray Levitt and his Stanford Associates for the use of SimVision
R
in
the computational experiments.

Notes

1. SimVision R
, previously named Vité Project, is available at Vite, Inc.
2. For more discussion of the backgrounds and assumptions of VDT, see Jin and Leavitt (1996).
3. When the number of simulation trials gets larger, small difference between the means of the two project teams
becomes significant. To make the simulation more conservative, we have done the same tests with 30 trials
instead of 100, and obtained the same results.

References

Bourgeois, L.J. III (1985), “Strategic Goals, Perceived Uncertainty, and Economic Performance in Volatile Envi-
ronments,” Academy of Management Journal, 28(3), 548–573.
Burton, R.M. and B. Obel (1998), Strategic Organizational Diagnosis and Design: Developing Theory for Appli-
cation. Kluwer Academic Publishers, Boston, MA.
382 KIM AND BURTON

Burton, R.M. and B. Obel (1995), “The Validity of Computational Models in Organization Science: From Model
Realism to Purpose of the Model,” Computational and Mathematical Organization Theory, 1(1), 57–71.
Campion, M.A., G.J. Medsker and A.C. Higgs (1993), “Relations Between Work Group Characteristics
and Effectiveness: Implications for Designing Effective Work Groups,” Personnel Psychology, 46(4), 823–
850.
Carley, K.M. (1995), “Computational and Mathematical Organization Theory: Perspective and Directions,” Com-
putational and Mathematical Organization Theory, 1(1), 39–65.
Carley, K.M. and Z. Lin (1997), “A Theoretical Study of Organizational Performance under Information Distor-
tion,” Management Science, 43(7), 976–997.
Carley, K.M. and D.M. Svoboda (1997), “Modeling Organizational Adaptation as a Simulated Annealing Process,”
Sociological Methods and Research, 25(1), 138–168.
Chang, M. and J.H. Harrington, Jr. (1998), “Organizational Structure and Firm Innovation in a Retail Chain,”
Computational & Mathematical Organization Theory, 3(4), 267–288.
Daft, R.L. and N.B. Macintosh (1981), “A Tentative Exploration into the Amount and Equivocality of Information
Processing in Organizational Work Units,” Administrative Science Quarterly, 26, 207–224.
D’Aveni, R.A. (1994), Hypercompetition: Managing the Dynamics of Strategic Maneuvering. Free Press, New
York.
Donaldson, L. (1996), “Normal Science of Structural Contingency Theory,” in S.R. Clegg, C. Hardy and W.R.
Nord (Eds.) Handbook of Organizational Studies. London: Sage.
Eisenhardt, K.M. (1989), “Making Fast Strategic Decisions in High-Velocity Environments,” Academy of Man-
agement Journal, 32(3), 543–575.
Eisenhardt, K.M. (1993), “High Reliability Organizations Meet High Velocity Environments: Common Dilemmas
in Nuclear Power Plants, Aircraft Carriers, and Microcomputer Firms,” in K.H. Roberts (Ed.) New Challenges
to Understanding Organizations. New York: Macmillan Publishing Company.
Galbraith, J. (1973), Designing Complex Organizations. Addison-Wesley Publishing Co., Reading, MA.
Gigone, D. and R. Hastie (1993), “The Common Knowledge Effect: Information Sharing and Group Judgment,”
Journal of Personality and Social Psychology, 65, 959–974.
Gittell, J.H. (2000), “Paradox of Coordination and Control,” California Management Review, 42(3), 101–117.
Guzzo, R.A. and M.W. Dickson (1996), “Teams in Organizations: Recent Research on Performance and Effec-
tiveness,” Annual Review of Psychology, 47, 304–338.
Hackman, J.R. (1987), “The Design of Work Teams,” in J. Lorsch (Ed.) Handbook of Organizational Behavior.
Englewood Cliffs, NJ: Prentice-Hall.
Jayaram, J., C. Droge and S.K. Vickery (1999), “The Impact of Human Resource Management Practices on
Manufacturing Performance,” Journal of Operations Management, 18, 1–20.
Jin, Y. and R.E. Levitt (1996), “The Virtual Team Design: A Computational Model of Project Organizations,”
Computational and Mathematical Organization Theory, 2(3), 171–195.
Keats, B.W. and M.A. Hitt (1988), “A Causal Model of Linkages Among Environmental Dimensions, Macro
Organizational Characteristics, and Performance,” Academy of Management Journal, 31(3), 570–598.
Kumar, A., P.S. Ow and M.J. Prietula (1993), “Organizational Simulation and Information Systems Design,”
Management Science, 39(2), 218–240.
Lawrence, P.R. and J.W. Lorsch (1967), Organization and Environment: Managing Differentiation and Integration.
Harvard Business School Press, Boston, MA.
Levine, J.M. and R.L. Moreland (1998), “Small Groups,” in D.T. Gilbert, S.T. Fiske and G. Lindzey (Eds.) The
Handbook of Social Psychology. Boston: McGraw-Hill.
Levitt, R.E., G.P. Cohen, J.C. Kunz, C.I. Nass, T. Christiansen and Y. Jin (1994), “The ‘Virtual Design Team’:
Simulating How Organization Structure and Information Processing Tools Affect Team Performance,” in K.M.
Carley and M.J. Prietula (Eds.) Computational Organization Theory. Lawrence Earlbaum Associates.
Lin, Z. and K.M. Carley (1997), “Organizational Response: The Cost Performance Tradeoff,” Management Science,
43(2), 217–234.
Lin, Z. and K.M. Carley (2003), Designing Stress Resistant Organizations: Computational Theorizing and Crisis
Applications. Boston, MA: Kluwer Academic Publishers.
March, J.G. and H.A. Simon (1958), Organizations. Wiley, New York.
McGrath, J.E. (1984), Groups: Interaction and Performance. Prentice-Hall, Englewood Cliffs, NJ.
EFFECT OF TASK UNCERTAINTY AND DECENTRALIZATION 383

Mohr, L.B. (1971), “Organizational Technology and Organizational Structure,” Administrative Science Quarterly,
16(4), 444–459.
Murray, A.I. (1989), “Top Management Group Heterogeneity and Firm Performance,” Strategic Management
Journal, 10, 125–141.
Nadler, D. and M. Tuhman (1988), Strategic Organization Design: Concepts, Tools, & Processes. Scott, Foresman
and Company, Glenview, IL.
O’Leary-Kelly, A.M., J.J. Martocchio and D.D. Frink (1994), “A Review of the Influence of Group Goals on
Group Performance,” Academy of Management Journal, 37, 1285–1301.
Pennings, J.M. (1975), “The Relevance of the Structural-Contingency Model for Organizational Effectiveness,”
Administrative Science Quarterly, 20(3), 393–410.
Perrow, C. (1967), “A Framework for the Comparative Analysis of Organizations,” American Sociological Review,
32, 194–208.
Phelan, S.E. and Z. Lin (2001), “Promotion Systems and Organizational Performance: A Contingency Model,”
Computational Mathematical Organization Theory, 7, 207–232.
Pfeffer, J. (1978), Organizational Design. Harlan Davidson, Inc., Arlington Heights, IL.
Pfeffer, J. (1997), New Directions for Organization Theory. Oxford University Press, New York.
Roberts, K.H. (1990a), “Some Characteristics of One Type of High Reliability Organization,” Organization Sci-
ence, 1(2), 160–176.
Safizadeh, M.H. (1991), “The Case of Workgroups in Manufacturing Operations,” California Management Review,
33(4), 61–82.
Schoonhoven, C. (1981), “Problems with Contingency Theory: Testing Assumptions Hidden Within the Language
of Contingency ‘Theory,’ ” Administrative Science Quarterly, 26(3), 349–377.
Shrivastava, P. and I.I. Mitroff (1987), “Strategic Management of Corporate Crisis,” Columbia Journal of World
Business, 22(1), 5–11.
Stasser, G. and D. Stewart (1992), “Discovery of Hidden Profiles by Decision-Making Groups: Solving a Problem
Versus Making a Judgment,” Journal of Personality and Social Psychology, 63, 426–434.
Stasser, G., D. Stewart and G.M. Wittenbaum (1995), “Expert Roles and Information Exchange During Discus-
sion: The Importance of Knowing Who Knows What,” Journal of Experimental Social Psychology, 31, 244–
265.
Steiner, I.D. (1972), Group Process and Group Productivity. Academic Press, New York.
Thompson, J.D. (1967), Organizations in Action. McGraw-Hill, New York.
Victor, B. and R.S. Blackburn (1987), “Determinants and Consequences of Task Uncertainty: A Laboratory and
Field Investigation,” Journal of Management Studies, 24(4), 387–404.
Vité Inc. (2002), SimVision Version 3.8. Mountain View, CA.
Ward, P.T., J.K. McCreery, L.P. Ritzman and D. Sharma (1998), “Competitive Priorities in Operations Manage-
ment,” Decision Sciences, 29(4), 1035–1046.
Weick, K.E. and K.H. Roberts (1993), “Collective Mind in Organizations: Heedful Interrelating on Flight Decks,”
Administrative Science Quarterly, 38(3), 357–381.
Wong, S.S. and R.M. Burton (2001), “Virtual Teams: What are Their Characteristics, and Impact on Team Perfor-
mance?” Computational and Mathematical Organization Theory, 6, 339–360.

Jisung Kim is a doctoral student in Management at the Fuqua School of Business, Duke University. He holds an
MBA from Bowling Green State University, and a BS in business administration from Seoul National University.
His research interests are in the effect of uncertainty on organizational design and performance, organizational
change and adaptation, and evolutionary processes in multiple levels of organizations.

Richard M. Burton is Professor of Business Administration, The Fuqua School of Business, Duke University
where he teaches and researches organizational design issues. Professor Burton’s research and teaching is in
organizational design and the application of computational approaches in organization theory. He co-authored a
book with Professor Obel, Odense University, STRATEGIC ORGANIZATIONAL DIAGNOSIS AND DESIGN:
DEVELOPING THEORY FOR APPLICATION, Kluwer Academic Publishers. They are the co-developers of the
384 KIM AND BURTON

ORGANIZATIONAL CONSULTANT software—a knowledge based expert system for decision support in the
design of organizations.
Previously, he has published six books and some fifty articles and book chapters on organizational deisgn,
strategy and management science. He serves on editorial board of various journals and was the Department Editor
for MANAGEMENT SCIENCE in Organizational Analysis, Performance and Design. He consults on strategy
and corporate organizational design.

You might also like