Dear participants,

Our session together will have two major themes: (1) publishing field-based research in the areas of

organizational behavior/organizational theory/international management; and (2) journal choices, and

other emerging themes, insights, and questions that cut across research areas and research methods.

I. In the first part of the session, we will discuss the joys and challenges of publishing research at the

intersection of organizational behavior, organizational theory, and international management.

Conceptually, we will focus on meso-level topics such as teamwork, communication, and collaboration

within and across organizations, especially global organizations. Methodologically, our focus will be on

intra-organizational field methods, including the use of original survey, archival, and interview data.

To focus the conversation, we will examine the development of a paper that I published in the

Academy of Management Journal in 2010. We will work through the review process that the paper

went through en route to publication, as well as discussing how it fits into an ongoing research stream.

I am providing the following materials for you to read in advance of the session:

(1) the first submission of the paper

(2) the first round of reviewers’ comments

(3) my first response to the reviewers

(4) my second submission of the paper

(5) the final published paper (note that there were two more revisions prior to this).

You may read items (1), (2) and (3) fairly quickly, to understand the early stages of the paper; then

please read item (4) carefully and prepare for our conversation by thinking about these questions:

(a) After reading item (4), my second submission of the paper:

• What are the major strengths and weaknesses that you see in the paper at this stage?

• How would you address the weaknesses while preserving the strengths?

(b) Overall, what is your sense of the greatest joys and the greatest challenges of publishing this

type of research (in terms of both areas and methods) in a leading management journal?

II. In the second part of the session, we will devote some time to discussing the differences between

the top journals, and the question of how to go about deciding where to send a given paper. We will

also aim to pull together some of the most striking themes that have arisen through your previous

sessions, and discuss insights that emerge when we look across the various journals, areas, and

methods to which you have been exposed during the program.

In preparation for this part of the session, please spend some time on the prior evening thinking

about (a) key themes that have emerged for you through the program, (b) specific insights that you

have gained, (c) particular challenges that you might like to see addressed more fully, and (d) any

questions that have arisen for you that we might usefully discuss together.

Looking forward to meeting all of you and to our session together!

Martine

2

(1) FIRST SUBMISSION

Embedded Autonomy:

Project Teams and Knowledge Work in Multinational Organizations

ABSTRACT

Combining team effectiveness theories, organizational design principles, and the knowledge-

based view of the multinational, this paper proposes that teams that deliver higher quality

projects in multinational organizations are characterized by “embedded autonomy” – extensive

use of knowledge from sources outside the team combined with control over critical task

decisions. I examine the embedded autonomy hypotheses using quantitative survey and project

performance data from a representative sample of 120 teams at a large multinational

organization. The findings shed light on the conditions that enable project teams to make

decisions that are both informed and independent, and underscore the value of a team-centered

perspective on knowledge-intensive work in multinational organizations.

Keywords:

multinational management, knowledge sharing, decision-making autonomy, project teams

3

A fundamental tenet of multinational management theory is the need to combine local

differentiation between national subsidiaries with global integration across those subsidiaries

(Bartlett & Ghoshal, 1989; Birkinshaw, Ghoshal, Markides, Stopford & Yip, 2003; Prahalad &

Doz, 1987). Organizational design research emphasizes that national subsidiaries require

substantial decision-making autonomy if they are to develop and deliver locally differentiated

products or services (e.g., Birkinshaw, Hood, & Jonsson, 1998; Garnier, 1982; Nohria &

Ghoshal, 1997; Sundaram & Black, 1992). The risk of delegating authority to national

subsidiaries, however, is that they become isolated from the rest of the multinational

organization. To mitigate this risk, cross-subsidiary integration mechanisms are necessary to

facilitate global coordination and promote worldwide learning (e.g., Martinez & Jarillo, 1989). In

particular, the knowledge-based view of the multinational argues that cross-subsidiary flows of

knowledge serve as a critical integration mechanism that can improve performance and generate

competitive advantage (e.g., Foss & Pedersen, 2004; Grant, 1996; Gupta & Govindarajan, 2000;

Kogut & Zander, 1993; Martin & Salomon, 2003; Tsai, 2001).

The tension between differentiation and integration is widely recognized at the level of

national subsidiaries, but typically these are not the units that actually carry out the critical tasks

of the multinational organization. Instead, there are multiple task units located within or across

the national subsidiaries that carry out these tasks. In particular, multinational organizations often

assign teams to carry out complex collaborative projects for which knowledge is the primary

production input, rather than labor or capital (e.g., Earley & Gibson, 2002; Lipnak & Stamps,

1997; Snell, Snow, Davison & Hambrick, 1998; Zellmer-Bruhn & Gibson, 2006). These project

teams face the macro-level challenge of combining differentiation and integration at the micro-

level of their everyday work, as they strive to deliver products or services that are both locally

4

adapted and globally informed. This raises the question of the implications of organizational

design principles and the knowledge-based view of the multinational for the teams that actually

carry out much of the daily work of multinational organizations.

Drawing on theories of team effectiveness as well as multinational management, I

propose that project teams in multinational organizations perform most effectively under

conditions of “embedded autonomy”, where they are both highly autonomous and highly

embedded in their environments (cf., Evans, 1995). While the concept of embeddedness has been

used to refer to a wide range of types of connections between individuals, groups, or

organizations and their environments,1 in this study “team embeddedness” refers to the extent of

the relations between teams and their environments in the specific form of knowledge flows from

the environment to the team, where knowledge flows are defined as inputs of task-related

information, know-how, or feedback obtained from people or documents both internal and

external to the organization (Huber, 1991). According to this definition, teams that are more

embedded in their environments obtain and use knowledge more than teams that are less

embedded in their environments. As in the work team literature, “team autonomy” refers to

collective control over critical decisions related to the team task, including decisions over the

team’s objectives, resources, design, and processes (Hackman, 1987). Teams that are more

autonomous have greater control over these critical decisions. While team effectiveness can be

evaluated along multiple dimensions, the focus of this study is on project quality, which is a

critical performance measure in many knowledge-intensive work settings (Starbuck, 1992) and

refers to the extent to which the output of a project (e.g., a product, proposal, or decision) meets

or exceeds the expectations of those who receive or use it (Hackman, 1987).

5

I argue that while decision-making autonomy can help project teams to perform more

effectively by enabling them to make independent decisions, it also creates risks of excessive

isolation. Knowledge inflows can help project teams to make more informed decisions,

meanwhile, but create their own risks of excessive external influence. The combination of high

embeddedness and high autonomy, however, can enable teams to avoid the risks of excessive

isolation or excessive external influence, and made decisions that are both informed and

independent. Consistent with this perspective, prior research has shown that knowledge

gathering is more positively associated with project performance when teams have more

decision-making autonomy (Haas, 2006). The present study builds on and extends this finding by

developing the theoretical foundations for an embedded autonomy perspective on project teams

in multinational organizations, and specifying a set of boundary conditions under which the

benefits of embedded autonomy will be greatest. After developing the theory and hypotheses, I

draw on qualitative interviews and observations to describe the research setting, and then test the

embedded autonomy framework using quantitative survey data and independently-assessed

evaluations of the quality of projects delivered by a representative sample of teams in a large

multinational organization.

PROJECT TEAMS AND KNOWLEDGE WORK

The Double-Edged Sword of Autonomy: Independence and Isolation

The multinational management literature and theories of organizational design theories

more broadly have long viewed decision-making autonomy or its obverse, centralization, as a key

attribute of organizational structure, and debated the question of how much autonomy to delegate

to subsidiaries (e.g., Barnard, 1938; Pugh, Hickson, Hinings & Turner, 1969; Simon, 1962;

Vaughan, 1990; Tsai, 2002). Autonomy conceptualized in this way is usually measured by an

6

aggregate index that measures authority over a series of key organizational decisions (e.g., Garnier,

1982; McMillan, Hickson, Hinings & Schneck, 1973).

The team effectiveness literature brings the concept of decision-making autonomy down

from the level of organizations and their subsidiaries to the level of the task units that carry out

the everyday work of the organization. In research on work group processes, empowerment, and

self-management, team autonomy refers to the level of discretion of the group in deciding what

tasks to perform and how to carry them out (Cohen & Ledford, 1994; Wellins, Wilson, Katz &

Laughlin, 1990). Team autonomy neither precludes nor necessarily implies individual autonomy

within the team (Langfred, 2000); instead, it emphasizes the extent to which a team itself can

make the critical decisions related to its task without having its authority superseded by external

parties, such as other work units, senior managers, or clients.

Hackman (1987) identified four sets of critical decisions that contribute to team

autonomy. Decisions over objectives concern the overall aims and direction for the team.

Resource decisions concern the acquisition and allocation of resources such as financial support.

Design decisions concern the design of the team and its task, such as how to structure the project

or who will join or leave the group. Process decisions concern the execution of the task and the

management of the team’s work, such as when and how to run meetings. The greater the team’s

discretion over these four sets of decisions, the greater its autonomy.

The team effectiveness literature suggests that greater collective control over critical task

decisions has motivational benefits for team members that can improve their task performance

(e.g., Kirkman & Rosen, 1999; Langfred, 2000; Latham, Winters, & Locke, 1994; Pearce &

Ravlin, 1987). These motivational benefits arise because individuals become more emotionally

committed to groups that strengthen their sense of control (Lawler, 1992), which can be

7

particularly helpful for building the sense of team identity that is often at risk in transnational

teams that rely on electronic communication (Shapiro, Furst, Spreitzer & Von Glinow, 2000).

Autonomy also provides an apparent signal of endorsement that increases group status as well as

performance expectations, further increasing commitment and promising greater rewards in the

form of bonuses, promotions, or prize assignments (Langfred, 2000).

Additionally, though less widely recognized in the team effectiveness literature,

autonomy can help to buffer teams from excessive external influence, by giving them greater

freedom to make decisions that contradict outsiders’ wishes and protecting them from possible

negative repercussions of such resistance. For this reason, new product development initiatives

or change implementation units often are given unusual latitude in multinational organizations

(Kanter, 1988; Tushman & O’Reilly, 1996). Autonomy thus both encourages and enables team

members to make decisions that are in the best interests of the team rather than outsiders.

The benefits of team autonomy can backfire, however, if the team becomes excessively

isolated from its environment. The multinational management literature has long recognized that

delegating autonomy to national subsidiaries is risky: an autonomous unit might explicitly

choose to act against the best interests of the global organization, or simply be unaware of

preferable courses of action due to isolation from the rest of the global organization (Bartlett &

Ghoshal, 1989; Prahalad & Doz, 1987). In studies of small groups, the risks of excessive

isolation have been highlighted by Janis’s (1982) analysis of “groupthink”, which occurs when

the strivings of group members for unanimity override their motivation to realistically appraise

alternative courses of action. This phenomenon is particularly prevalent where an autonomous

group is isolated from its environment: prior to the Bay of Pigs debacle of 1961, for example,

President Kennedy’s close circle of advisers failed to solicit inputs from outsiders who might

8

have broken through its blind spots and taboos on divergent opinions. The problem of excessive

isolation associated with high autonomy has since recurred in events ranging from the NASA

shuttle tragedies (Vaughan, 1990) to the failure of the US intelligence agencies to anticipate the

terrorist strikes of 9/11 (Kean & Hamilton, 2004). As these examples indicate, the benefits of

high autonomy can become substantial problems if a team is excessively isolated from its

environment.

The Double-Edged Sword of Knowledge Flows: Information and Influence

While autonomous teams run the risk of isolation, more extensive knowledge inflows can

mitigate this risk by providing inputs that help them to make more informed decisions. Research

on intra-organizational knowledge sharing has emphasized the benefits of using knowledge from

diverse sources for the performance of task units within organizations (e.g., Ancona & Caldwell,

1992; Hansen, 1999; Reagans & Zuckerman, 2001; Tsai, 2001). As suggested by the knowledge-

based view of the firm, colleagues with relevant expertise or documents from proprietary

databases can provide information, insights, and ideas that improve team performance (cf. Grant,

1996; Kogut & Zander, 1996). For example, transferring best practices can enable teams to

benchmark and improve their work processes (Szulanski, 1996), brainstorming with colleagues

who agree to devote a few hours to thinking about a specific product design problem can help

generate innovative ideas (Hargadon & Sutton, 1997), and downloading documents or soliciting

advice from experts can increase the chances of winning a competitive bid for a new client

project (Haas & Hansen, 2005). In the multinational context, knowledge acquisition from foreign

parents has been shown to improve the performance of international joint ventures (Lyles &

Salk, 1996), while knowledge management norms and procedures increase team learning

(Zellmer-Bruhn & Gibson, 2006). The implication is that knowledge inflows can increase the

9

benefits of autonomy and reduce its drawbacks by enabling teams in multinational organizations

to make decisions that are well-informed as well as independent.

Extensive knowledge inflows in the absence of high autonomy create their own

problems, however, because knowledge is a source of influence as well as information in many

multinational organizations. Organizations are often contested terrain in which actors with

divergent agendas and interests hold conflicting views about what qualifies as accurate and

relevant knowledge and how that knowledge should be used (Bacharach & Lawler, 1980; Cyert

& March, 1963; Pettigrew, 1973). Motivated by turf wars (e.g., Brown et al., 2005), competition

over scarce resources (e.g., Gresov & Stephens, 1993), or commitment to particular ideological

views (e.g., Carlile, 2002), knowledge providers often attempt to influence teams that seek inputs

from them (cf. Gresov & Stephens, 1993; Pfeffer, 1981; Spekman, 1979). They may demand

support for their agendas in return for their knowledge, for example; more subtly select the

inputs they provide in order to persuade the team members of their views; or promote their own

positions by emphasizing favoured issues and solutions without even necessarily realizing it

(Feldman, 1988; Feldman & March, 1981; Pettigrew, 1973).

While influence attempts do not always accompany knowledge inflows, and those that do

occur can be helpful in sensitizing the team to important concerns, problems of excessive

external influence frequently arise. These problems include role conflict for individual team

members who feel torn between the team and outsiders (cf. Katz & Kahn, 1966); damaging

conflicts among team members who advocate competing views based on their loyalties and

obligations to outsiders rather than the team (cf. Jehn, 1995) - or alternatively, inhibited

discussion and learning because team members avoid confrontation (Edmondson, 2002);

unproductive expenditures of time and energy on managing internal and external politics

10

(Ancona & Caldwell, 1992; Eisenhardt & Bourgeois, 1988); and even potentially cooptation of

the project if the team is diverted from a preferable course to one that more closely reflects the

agendas and interests of outsiders (cf. Selznick, 1949). These potential problems suggest that

more knowledge inflows from the team’s environment do not necessarily result in improved

project quality in the absence of team autonomy. Instead, project teams must secure the

information benefits of the knowledge inflows while avoiding the risks of excessive external

influence if they are to perform effectively.

Embedded Autonomy

In an influential political sociology theory of nation states, Evans (1995) coined the

phrase “embedded autonomy” to argue that well-functioning nation states are those that

continuously solicit and consider the opinions and concerns of their constituents but also are able

to buffer themselves from excessive pressures from those constituents. In the present context of

project teams in multinational organizations, embeddedness and autonomy provide a similarly

potent combination because they offer complementary advantages that offset each others’

disadvantages, as summarized in Figure 1.

----- insert Figure 1 here -----

As the figure shows, teams with high autonomy have independence benefits, but if they

lack information due to low embeddedness, they run the risk of excessive isolation (lower-right

quadrant). In contrast, teams with high embeddedness have information benefits, but if they lack

independence due to low autonomy, they run the risk of excessive external influence (upper-left

quadrant). The combination of high embeddedness with high autonomy, however, enables teams

to avoid the risks of excessive isolation or influence and make decisions that are both informed

and independent (upper-right quadrant). Hence:

11

Hypothesis 1: Teams that are highly autonomous and highly embedded will deliver higher

quality projects than teams that are (a) high in autonomy but low in embeddedness, (b) high

in embeddedness but low in autonomy, (c) low in autonomy and embeddedness.

Beyond this baseline hypothesis, the proposed benefits and risks of team autonomy and

team embeddedness suggest that the combination of embedded autonomy can be expected to be

most valuable under conditions where the information benefits provided by embeddedness and

the independence benefits provided by autonomy are particularly high; in contrast, embedded

autonomy will be less valuable where these benefits are low. Four such boundary conditions for

the value of embedded autonomy occur when the knowledge content is relatively scarce, the

knowledge source is outside rather than inside the organization, and the project is highly

demanding or highly prominent.

Inflows of relatively scarce knowledge content offer greater informational benefits than

inflows of more common knowledge content because the team members are less likely to possess

similar knowledge themselves, offering them greater benefits in the form of more diverse, non-

redundant inputs (Burt, 1992; Haunschild & Beckman, 1998; Reagans & Zuckerman, 2001). For

example, in a multinational organization that staffs teams with members who have high

functional expertise rather than high local market expertise, obtaining knowledge about the

country-specific conditions relevant to the project can make more difference to quality of the

team’s project than obtaining knowledge about the technical aspects of the work (cf. Lord &

Ranft, 2000). Because control over scarce resources is a source of power, however, the

independence benefits of autonomy also increase with the scarcity of the knowledge content.

Inflows of scarce knowledge expose teams more to the risks of excessive external influence more

than inflows of more common knowledge because they cannot as easily switch to different

12

providers to avoid pressures from outsiders who threaten to withhold their knowledge, or check

their trustworthiness by consulting additional sources (cf. Haunschild & Beckman, 1998;

Szulanski, Capetta, & Jensen, 2004). Further, they cannot readily judge the accuracy or

inclusiveness of this knowledge as opportunities for comparison are limited (cf. Hansen & Haas,

2001). While scarcer knowledge content increases the information benefits of embeddedness,

therefore, it simultaneously increases the independence benefits of autonomy.

Hypothesis 2: The value of embedded autonomy for teams will be greater if the

knowledge content is scarce rather than common.

Knowledge inflows from sources outside the organization offer greater informational

benefits than knowledge inflows from sources inside the organization, meanwhile, since the team

members’ external knowledge networks will typically be more differentiated than their internal

networks, resulting in more diverse and non-redundant inputs for the team (cf. Burt, 1992;

Hargadon & Sutton, 1997; Reagans & Zuckerman, 2001). The independence benefits of

autonomy are also greater for knowledge from outside than inside sources because the agendas

and interests of outsiders are less likely to be aligned with the goals of the team than those of

insiders, exposing teams to greater risks of excessive external influence (Barnard, 1938). While

there is often considerable divergence in agendas and interests within organizations (Cyert &

March, 1963), the super-ordinate identity, shared interests, and social norms that are created by

organizational membership generally increases cooperation among members relative to non-

members (cf. Schein, 1992; Sherif, 1958). Additionally, organization members tend to avoid

seeking knowledge from insiders while actively seeking knowledge from outsiders (Menon &

Pfeffer, 2003), suggesting that they are more open to influence from outsiders despite the

13

possibility that they are less committed to the team’s goals. Since both the informational benefits

of embeddedness and the independence benefits of autonomy are greater:

Hypothesis 3: The value of embedded autonomy for teams will be greater if the

knowledge source is outside rather than inside the organization.

Beyond the content and source of the knowledge inflows, the characteristics of the

project itself can also increase the benefits of embedded autonomy by heightening the risks of

excessive isolation or excessive influence. The risk of excessive isolation is greater when

projects are highly demanding (i.e., very novel or complex rather than routine or simple),

increasing the informational benefits of embeddedness. When a project is highly novel for the

team members involved, an exploitation strategy that relies exclusively on the team members’

accumulated past experiences can endanger project quality, whereas an exploration strategy that

involves seeking out knowledge from diverse sources beyond the team offers greater benefits (cf.

March, 1991). In contrast, when a project is routine, exploitation creates only limited risks, since

the potential value of new knowledge from outside the team is lower (cf. Haas & Hansen, 2005).

Similarly, excessive isolation poses greater risks for highly complex projects than for simple

ones, since the latter require more sophisticated and varied types of knowledge that are less

likely to exist within the team (cf. Tushman, 1977). Teams that work on projects that highly

novel or complex thus stand to benefit more from extensive knowledge inflows, which in turn

increases the benefits of autonomy for ensuring that teams can act independently on the

information obtained from outsiders.

Hypothesis 4: The value of embedded autonomy for teams will be greater if the project is

highly demanding rather than less demanding.

14

Highly prominent projects – those that attract more attention from stakeholders such as

managers, clients, or customers – create greater risks of excessive external influence, meanwhile.

Outsiders who provide knowledge to the team usually care more that their own agendas and

interests are reflected in the outcomes of more prominent projects, giving them more motivation

to attempt to influence the team. The team members themselves also view knowledge inputs

from outsiders as more critical for highly prominent projects, since they expect the outcomes of

such projects to be noticed, rewarded, and have more impact (cf. Meznar & Nigh, 1995). This

increases their dependence on outsiders, and thus the power of those outsiders to influence the

team (cf. Pfeffer & Salancik, 1978). Given their increased vulnerability as well as exposure to

external influence attempts, the benefits of independence conveyed by high autonomy are greater

for teams that work on highly prominent projects, simultaneously increasing the value of

embeddedness for ensuring that these independent decisions are well-informed.

Hypothesis 5: The value of embedded autonomy for teams will be greater if the project is

highly prominent rather than less prominent.

DATA AND METHODS

Research Setting

The hypotheses were tested using quantitative data collected during a multi-method field

study of project teams at a major international financial institution that makes large-scale loans

and provides high-level advice to national and regional governments in developing countries. To

understand the teams, their projects, and the work setting, the field research began with an

extensive qualitative data collection phase consisting of semi-structured interviews lasting

between one and three hours each. I started with 20 interviews with managers and staff,

including members of the units responsible for strategy and change management, knowledge

15

management, project quality monitoring, human resources, and the staff association, using these

interviews to gain an overview of the organization’s functions and operations. Next, I conducted

18 interviews with leaders and members of project teams based at headquarters, and 7 further

interviews in Russia, where I visited the organization’s Moscow office. I usually asked these

interviewees to describe a project on which they were currently working, probing for specific

details about the structure of the team, how the members carried out their work, and any

problems they encountered. Finally, I conducted a further 25 interviews as part of detailed case

studies of seven teams, interviewing the leader of each team, the members who were engaged in

the team’s work at the time, and specialists who were involved with the team on a sporadic basis.

I also observed team meetings and read materials that were generated as these teams worked.

The organization operated worldwide and supported offices in 100 countries. The official

hiring policy drew employees from every country in the world, and the organization had a matrix

structure, with operations spanning four divisions and six world geographic regions. The

interviews and observations indicated that despite these multinational characteristics, however,

the traditional multinational management paradigm of subsidiary differentiation and cross-

subsidiary integration did not apply well to this organization because the national subsidiaries

were not the primary work units responsible for its operational tasks. Instead, these tasks were

assigned to project teams that were mostly based at the U.S. headquarters where 80% of the

organization’s 10,000 employees were located. The implication of these task arrangements was

that critical task-related decisions were made by teams rather than subsidiaries, and knowledge

flows to teams were more critical for performance than knowledge flows to subsidiaries.

The team members were highly qualified experts, usually with postgraduate training in

their specialist areas. A typical team was composed of economists and technical specialists with

16

expertise in domains such as public finance, infrastructure, or engineering who were brought

together for the purposes of the specific project, joined and left the team at different points, and

flew in and out of the client country several times during the course of a project. The teams were

usually assembled by a designated team leader who sought out available individuals with the

required skills, rather than by careful assignment of particularly strong teams to selected projects,

and teams rarely worked on more than one project together.

The qualitative research indicated that decision-making autonomy helped teams to act

independently but the level of collective control over task decisions varied across teams. The

determinants of team autonomy were multiple and often idiosyncratic, including bureaucratic

requirements, the matrix structure of the organization, the particular distribution of informal as

well as informal authority in the units, the status of the team members, the style of the senior

managers, the limitations of the budgets, and the extent of client participation in the projects.

The interviews and observations also highlighted the importance of knowledge inflows

for providing valuable information to the project team. Each team member was responsible for

researching and designing a specific component of the project and integrating it with the other

components, or for providing input to several project components, and these tasks required high

levels of expertise, analysis, and input from multiple sources. To facilitate the flow of knowledge

to project teams, senior management had introduced a high-profile knowledge management

initiative, making substantial investments in communities of practice, expert directories, help

desks, and electronic document repositories. However, the qualitative research further indicated

that knowledge flows to a team were frequently accompanied by influence attempts. Loans and

advice to developing countries are intensely controversial issues where the best approaches are

highly contested and multiple parties have stakes in the outcomes of each project (Haas, 1990).

17

Teams typically solicited inputs for their projects from managers in the organization, experts

around the world, client governments, multinational corporations, local industry, and non-

governmental organizations. Given the contested nature of knowledge in this setting and the

diverse agendas and interests of these parties, attempts to influence the teams were common.

Quantitative Variables

The quantitative data came from three main sources: the organization’s quality-

monitoring unit, which provided independent ratings of project quality, a survey administered to

members of the teams that participated in the quality evaluations, and archival project records.

To evaluate the quality of the projects delivered to its clients, a dedicated quality-

monitoring unit of 20 full-time staff annually drew a random sample of financial loan projects

(stratified by region and division) and technical advice projects (stratified by cost) from the full

population of projects completed across the organization in the previous year. To assess each of

these projects, the unit then assembled customized panels that included at least two respected

experts in the area of the project. The panelists were drawn from both within and outside the

organization and had no prior connections with the project. Along with the widespread

recognition that their professional reputations were on the line, these precautions minimized the

chances that panelists might have a personal interest in making a project look good or bad. They

were responsible for reviewing the project documentation, interviewing the team leader,

completing a detailed evaluation protocol, and jointly arriving at an overall evaluation of the

quality of the project. As well as observing this evaluation process and interviewing some

participants, I obtained the quality ratings for 120 projects that were undergoing quality

evaluation in the year of the study (60 financial teams and 60 technical teams).

18

To collect data from the members of these teams, I sent them a customized survey that

directed them to focus on the project selected for quality evaluation. Pre-tests of the full survey

with 12 individuals and selected sections with another 42 individuals served to refine the survey

questions and to ensure the validity and shared understanding of the items in the organizational

context. Responses were received from 550 of the 1021 team members who were surveyed

(response rate = 54%), and 96 of the 120 teams qualified for the study by returning at least 50%

of their core team members’ surveys (50 financial and 46 technical teams; qualifying rate =

80%). Tests for selection bias showed no significant differences between the quality ratings,

project type, region, or division with the 24 teams that did not qualify for the study, though

qualifying teams had worked on more costly and lengthier projects than disqualified teams.

Project quality. The dependent variable in this study was the overall project quality rating

for a team determined by the panel of experts on an ordinal scale on which 3 was “highly

satisfactory”, 2 was “satisfactory”, and 1 was “marginal or unsatisfactory” (project quality).

Using a series of over 100 detailed questions, the panels rated the projects on multiple quality

dimensions that loaded onto a single factor for both financial and technical projects (α = 0.82 and

α = 0.75 respectively). The final ordinal project quality rating was based on the panel’s overall

assessment of the quality of the project, taking into account not only the raw scores but also their

full understanding of the project and its distinctive challenges. Although each project was

evaluated by a different independent panel to ensure a match between the focus of the project

and the expertise of the panelists, the quality-monitoring unit took considerable care to ensure

that the results of these quality evaluations were robust across panels. In addition to providing

clearly specified evaluation criteria, detailed supporting questions, and hands-on guidance to the

panels during the evaluation process, the unit had conducted extensive tests of the inter-panel

19

reliability of the quality ratings and established that different panels were highly likely to rate the

same project similarly. Inter-rater reliability within a panel of experts was not a concern because

the quality of the project was determined by all the panelists working collectively rather than by

individual panelists working separately. Of the 96 projects in the final dataset, 16% received a

quality rating of 3, 70% received a quality rating of 2, and 14% received a quality rating of 1.2

Team autonomy. In the survey, the team members were asked to report on the team’s

level of discretion over a list of 20 critical task-related decisions generated through the field

interviews. Following Hackman (1987), I identified five decisions in each of the four main

categories of objectives, resources, design, and processes. Decisions about objectives concerned

project initiation, overall priority, boundaries and scope, specific components, and level of

innovation. Decisions about resources concerned budget size, additional funding, level of

information input, team training or coaching, and team rewards or recognition. Decisions about

task and team design concerned project pacing, feedback solicitation, quality standards, staffing

requirements, and selection of team members. Decisions about team processes concerned setting

up and managing missions, level of interaction with clients and management, and handling

conflict. For each decision, the team members were asked, “How was influence over this

decision distributed between the team itself (including the team leader) and others outside the

team (including managers, the client country, and the development community)?”, using 5-point

scales with anchors of “the team had very little influence over the decision; others had almost all

the influence” and “the team had almost all the influence over the decision; others had very little

influence”. The responses of the team members were averaged across all 20 decisions to create

an overall measure of team autonomy with a Cronbach’s alpha of 0.90 (team autonomy).

20

Team embeddedness. The interviews revealed that the team members typically classified

the sources outside the team from which they obtained knowledge into four categories that

corresponded to geographic and organizational boundaries. Using these categories, the survey

respondents were asked: “During the course of the project, how much relevant technical

knowledge did you gather from (a) the country office? (b) the rest of the organization? (c) the

client country? (d) the global community?”, where technical knowledge was defined as

“knowledge about the technical aspects of the work – the professional skills, competencies, and

expertise relevant to the project.” The client country was defined to include the client

government, intended beneficiaries of the project, and local stakeholders such as NGOs and

businesses, while the global community was defined to include those who worked on

development issues around the world, including international NGOs, foundations, think tanks,

academics, and members of professional networks. They were then asked the same set of

questions about country knowledge, which was defined as “knowledge about the local

environment – the country-specific conditions relevant to the project.” The responses to these

eight items (on 5-point scales with anchors of “very little knowledge” and “a lot of knowledge”)

were averaged across the respondents to create an overall measure of the knowledge inflows to

the team (team embeddedness).3 To test hypothesis 1, I multiplied this overall embeddedness

score by the team’s overall autonomy score.

Knowledge content and source. My interviews indicated that project teams in this

organization typically found that country knowledge content was relatively scarce compared

with technical knowledge content. While the organization had national hiring quotas, project

teams were staffed according to functional expertise rather than familiarity with the client

country; indeed, the organization purposefully avoided staffing its teams with nationals of the

21

client country and rotated its employees onto projects in different countries frequently to

preserve a professional distance from the client. Additionally, country knowledge in the form of

reliable information about economic and social conditions in developing countries is often very

poor or non-existent, whereas technical knowledge usually builds on formal education or prior

experiences in other countries and so is more abundant. Finally, the project teams could not

always identify and access country knowledge that did exist as easily as technical knowledge

because the team members mostly were based in the U.S. rather than the client countries, and

therefore tended to be more deeply entrenched in professional rather than national knowledge-

sharing networks. I therefore tested hypothesis 2 by splitting the aggregate knowledge variable

into two sub-constructs that captured technical and country knowledge inflows separately

(Cronbach’s alphas of 0.70 for technical knowledge and 0.72 for country knowledge). A

multitrait-multimethod matrix analysis indicated that the average within-scale correlations for

these two knowledge content sub-constructs exceeded the average between-scale correlations,

indicating convergent and discriminant validity (Campbell & Fiske, 1959). According to

hypothesis 2, the interaction effect with team autonomy should be stronger for the relatively

scarce country knowledge than for the relatively common technical knowledge variable.

To test hypothesis 3, I created two further sets of knowledge sub-constructs by separately

averaging the four items for knowledge obtained from sources inside the organization (the

country office or rest of the organization) and the four items for knowledge obtained from

sources outside the organization (the client country or global community) (Cronbach’s alphas of

0.70 for inside knowledge and 0.80 for outside knowledge). The average within-scale

correlations again exceeded the average between-scale correlations, establishing convergent and

discriminant validity for the two knowledge source sub-constructs. Hypothesis 3 predicts that the

22

interaction effect with team autonomy should be stronger for the inside knowledge variable than

for the outside knowledge variable.

Demanding and prominent projects. My interviews with teams and managers indicated

that projects were viewed as more demanding in the organization if they were novel rather than

routine, and complex rather than straightforward. To capture project novelty, the team members

were asked: “To what extent did the work depart from the usual work of a routine

financial/technical assistance project?” (5-point scale with anchors “the work was very routine”

and “the work was very non-routine”) (project novelty). To capture project complexity, the team

members were asked: “To what extent did the task require complex approaches and solutions?”

(5-scale with anchors “the work was not particularly complex” to “the work was very complex

indeed”) (project complexity).

The interviews further revealed that the projects that were more prominent in this

organization were those that had larger budgets committed to their preparation because they were

larger in scale or considered a higher strategic priority, and/or those that were being prepared for

richer countries, which tended to be powerful clients and regional role models with larger

populations. I therefore measured prominence by project costs and project country wealth. To

determine the cost of preparing each project, I used archival budget data from the project files

and logged the U.S. dollar amounts (project cost). To capture the relative wealth of the client

countries, which typically was considered in relation to other countries in the same region, I

constructed a measure of the country GNP (gross national product) as a percentage of regional

GNP (project country wealth).

Using median splits, I divided the sample into high- and low-demand projects using each

of the two criteria that captured how demanding they were (high versus low project novelty and

23

high versus low project complexity), and into high- and low-prominence projects using each of

the two prominence criteria (high versus low project costs and high versus low country wealth).

To test hypotheses 4 and 5, I compared the models across the paired sets of high- versus low-

demand projects and high- versus low-prominence projects, to establish whether the effects were

stronger for more demanding and prominent projects. An alternative approach would be to use

three-way interactions between embeddedness, autonomy, and the project variables instead of

median split comparisons, but the number of observations in the dataset made this infeasible.

Team, respondent, and project control variables. Because more prior knowledge could

reduce the need for knowledge inflows, I asked how much relevant country and technical

knowledge the survey respondents personally possessed prior to starting work on the project

(using 5-point scales with anchors of “very little” and “a lot”) and averaged their responses (team

knowledge). The models also controlled for the number of team members (team size), for the

location of the team using a binary variable coded 1 if the team was based at headquarters or 0 if

it was based in a country office (team location), and for the extent to which the team was a

clearly identifiable interdependent group using two five-point scales items (Hackman &

Wageman, 2004) with a Cronbach’s alpha of 0.66 (real team). I also constructed measures of the

proportion of survey respondents in each team who were core team members (core respondents),

the proportion who returned their surveys more than seven days after their project’s evaluation

interview (late respondents), and the average length of organizational tenure in years at the start

of the project (tenure). Because preliminary analyses showed consistent project quality

differences in one region and one division, binary variables were coded 1 for this region or

division respectively, and 0 otherwise (region, division). Using archival records, I also calculated

the logged duration of a project from initiation to completion (project duration), and the logged

24

costs of the project in dollars (project cost). Finally, the models included a binary variable coded

1 for financial projects and 0 for technical projects to capture mean differences between the two

types of projects conducted by the teams in the study (project type).

RESULTS

----- insert Tables 1, 2, and 3 about here -----

Table 1 reports descriptive statistics and correlations. The main explanatory variables were

standardized (by subtracting by the mean and dividing by the standard deviation) to avoid high

levels of multicollinearity with the interaction terms (Neeter, Wasserman & Kutner, 1985). One

noteworthy observation is that the correlation between team autonomy and team embeddedness is

low and positive (r=0.15, not significant), indicating that these constructs are orthogonal:

embeddedness does not reduce autonomy, not does autonomy preclude embeddedness.

As shown in Tables 2 and 3, ordinal logit analysis was used to test the hypotheses because

the dependent variable of project quality was categorical and ordered (Long, 1997).4 Model 1 in

Table 2 reports the baseline model including only the control variables. This model shows that the

teams that delivered higher quality projects had more members and longer average member tenure,

they were more clearly identifiable groups, more often came from one particular division, and had

more late respondents. Model 2 shows that the main effect of team autonomy is positive and

significant: teams with more control over their task-related decisions delivered higher quality

projects. Model 3 shows that the main effect of team embeddedness is not significant, a finding

that is consistent with the argument that knowledge inflows have drawbacks as well as benefits.

Model 4 reports the results for hypothesis 1, which proposed that project teams perform

most effectively under conditions of high embeddedness and high autonomy. As previously

established in Haas (2006), this model shows that knowledge gathering is more positively

25

associated with project quality when teams have more decision-making autonomy. Put

differently, the interaction between team embeddedness and team autonomy is positive and

significant, providing support for hypothesis 1. This interaction effect is plotted in Figure 2 to

illustrate the implications for teams with varying levels of embeddedness and autonomy. High

and low levels of embeddedness and autonomy are set at one standard deviation above and

below their mean levels, respectively (Aiken & West, 1991). The project quality scale ranges

from 1 to 3, giving a maximum of 2-points difference between high and low quality projects. The

plot shows that teams with high embeddedness and high autonomy delivered higher quality

projects on average than teams with high embeddedness but low autonomy (a 0.50-point

difference), teams with high autonomy but low embeddedness (a 0.32-point difference), or teams

with low embeddedness and low autonomy (a 0.22-point difference). These findings support

hypotheses 1a, 1b, and 1c.

The plot also reveals two further findings: teams with low autonomy delivered lower

quality projects on average if they had high knowledge inflows than if they had low knowledge

inflows (0.28 points lower), and teams with low inflows delivered lower quality projects on

average if they had high autonomy than if they had low autonomy (0.10 points lower). The

implication is that the combination of low embeddedness and low autonomy endangers project

quality less than the combination of low embeddedness and high autonomy, which in turn

endangers project quality less than the combination of high embeddedness and low autonomy.

----- insert Figure 2 about here -----

Model 5 and Model 6 present the results for hypotheses 2 and 3, which proposed that the

value of embedded autonomy varies with the content and source of the knowledge inflows to the

team. Model 5 shows that the interaction between autonomy and country knowledge is

26

substantial and strongly significant whereas the interaction between autonomy and technical

knowledge is small and not significant. These two interactions are statistically significantly

different (chi2=3.99, p<0.05). Model 6 shows that the interaction between autonomy and

knowledge from sources inside the organization is not significant whereas the interaction with

knowledge from sources outside the organization is significant. This model also reveals that the

main effect of knowledge flows from sources inside the organization is negative whereas the

main effect of knowledge flows from sources outside the firm is positive. Accounting for the

combined main and interaction effects, the difference between embeddedness in internal and

external knowledge sources is statistically significant (chi2=3.49, p<0.10). These results indicate

that autonomy was more valuable in combination with relatively scarce knowledge content or

knowledge sourced from outside the organization than for relatively common knowledge content

or knowledge sourced from inside the organization, as the hypotheses predicted.

The four sets of paired models presented in Table 3 report the results for hypotheses 4

and 5, which proposed that embedded autonomy is especially advantageous for teams that work

on projects that are more demanding or more prominent. These results show a consistent pattern:

the interactions between team embeddedness and team autonomy are larger and significant for

projects above the median levels of project novelty and project complexity (models 7b and 8b),

but smaller and non-significant for less demanding projects (models 7a and 8a). Likewise, the

interactions are larger and significant for projects above the median levels of project cost and

country wealth (models 9b and 10b), but smaller and non-significant below the median levels of

these indicators of project prominence (models 9a and 10a). Using a seemingly unrelated

estimation procedure to test for the statistical significance of these differences indicates that the

coefficients in the paired novelty, cost, and wealth models are significantly different from each

27

other, but the coefficients in the paired complexity models are not (chi2=36.12, p<0.01 for

novelty; chi2=21.81, p<0.15 for complexity; chi2=23.15, p<0.05 for cost; chi2=28.01, p<0.05 for

wealth). These results thus provided mixed evidence for the hypothesis that the combination of

embeddedness and autonomy is more valuable for more demanding projects, while supporting

the hypothesis that embedded autonomy is more valuable for more prominent projects.

Alternative Explanations

While the models provide converging evidence to support the importance of embedded

autonomy for project team effectiveness, there might be alternative explanations for some of

these results. One possibility is that external factors are driving the observed relationships

between knowledge inflows, decision-making autonomy, and project quality. For example,

perhaps some teams received more knowledge inflows, were given more autonomy, and also

performed better because their members were more expert or experienced. This argument would

suggest that team member quality should be strongly correlated with both team embeddedness

and team autonomy. However, two measures of team member quality utilized in this study, their

prior knowledge and their organizational tenure, reveal only low and non-significant correlations

with levels of knowledge inflows and decision-making autonomy. A related argument is that

more demanding projects that required more knowledge inflows were staffed with better teams

that were given more autonomy. But correlations between project novelty or complexity and

team autonomy again were low and non-significant. The qualitative data also provided little

support for this argument, since team leaders often complained that they could not strategically

staff more demanding projects with better team members because they had to take whomever

was available and also because it was hard to predict the difficulty of a project in advance.

28

A different alternative explanation is that the findings of the study might be due to post-

evaluation attribution bias. Some projects had undergone the full quality evaluation process

before the surveys were distributed, raising the possibility that the members of these teams knew

the outcome of their evaluations and might have made self-serving attributions in their responses

to the survey (Miller & Ross, 1975). Specifically, team members who believed their team had

performed poorly might have been likely to report both that they had insufficient autonomy and

that they had solicited extensive knowledge inflows, in an effort to make themselves look as

though they had done everything they could. The research design allowed for testing of such

biases, however, by comparison of 19 teams whose members all returned their surveys before

their quality evaluation outcome was announced with 37 teams whose members all returned their

surveys at least seven days after their evaluations were completed.5 The tests revealed no

significant differences on the main variables in this study, particularly knowledge inflows (t=-

0.01, n.s.) and decision-making autonomy (t=0.98, n.s.), or on the correlations between project

quality and knowledge inflows (z=0.42, n.s.) or project quality and decision-making autonomy

(z=1.20, n.s.), indicating little evidence of attribution bias in the data. There were also no

significant differences in the quality ratings across these two sets of teams (t=-0.74, n.s.).

It remains possible, however, that these tests are insufficient because the teams did not

need to await the results of their quality evaluations to have a fairly accurate sense for how they

performed. The available evidence, though, does not support this explanation. First, it was far

from easy for a team to assess the quality of its project prior to the quality evaluation process, as

frequent complaints from teams that received poor quality ratings attested. Second, the four

boundary condition hypotheses are difficult to explain using the post-hoc attribution argument,

which would assume a very high level of careful calculation in the survey respondents’ answers.

29

Third, although a multi-item survey measure of respondents’ satisfaction with their project’s

quality (based on Wageman & Hackman, 2004) was positively correlated with team autonomy,

as might be expected (r=0.31, p<0.01), it was also positively correlated with the knowledge

inflows reported (r=0.16, p<0.01), indicating that the team members were not likely to report

more knowledge inflows if they were less satisfied with their performance. Attribution bias thus

does not provide a convincing explanation for the interaction effects found in this study.

DISCUSSION

This study shows that embedded autonomy enables project teams to deliver higher

quality projects, where embeddedness takes the form of knowledge inflows from the

environment and autonomy takes the form of control over critical task-related decisions. In the

multinational organization studied here, teams with high embeddedness but low autonomy and

teams with high autonomy but low embeddedness delivered lower quality projects than teams

that were both highly embedded and highly autonomous. Unexpectedly, project quality suffered

less in teams with low embeddedness and low autonomy, suggesting that the influence risks

outweighed the information benefits for teams with high embeddedness but low autonomy, while

the isolation risks outweighed the independence benefits for teams with high autonomy but low

embeddedness. The combination of embedded autonomy was beneficial particularly when the

knowledge content was relatively scarce, the knowledge source was outside the organization, and

the projects were more demanding or more prominent. The study thus extends previous research

that has shown that decision-making autonomy can increase the benefits of knowledge gathering

for project performance (Haas, 2006) by developing an embedded autonomy perspective on team

effectiveness and identifying boundary conditions under which the combination of

embeddedness and autonomy is most valuable. The results provide support for a team-centered

30

theory of knowledge work in multinational organizations that views decision-making autonomy

as providing independence at the risk of excessive isolation, and knowledge inflows as providing

information at the risk of excessive influence.

Though potentially relevant to settings other than multinational organizations too, the

embedded autonomy framework is particularly relevant for understanding the conditions that

improve performance in multinational settings because it provides insight into the fundamental

challenge of combining differentiation with integration at the level of the project teams that carry

out the everyday work of many multinationals (cf., Earley & Gibson, 2002; Snell et al, 1998).

Autonomy is a central concept in theories of multinational organizational design (e.g. Bartlett &

Ghoshal, 1989), while knowledge flows are the foundations for the knowledge-based view of the

multinational (e.g., Grant, 1996; Kogut & Zander, 1993). In focusing on headquarter-subsidiary

relations, however, the extant theory and research offers little guidance to explain variations in

the performance of critical tasks within a multinational subsidiary that has a given level of

autonomy and knowledge inflows. Moreover, the traditional models of headquarter-subsidiary

relations that have long dominated theories of multinational management seem increasingly

inflexible and static in the face of trends such as the emergence of regional powerhouses like

India and China outside the headquarters of U.S. and European-based multinationals, the

proliferation of centers of excellence, the growth of global account management, and worker

mobility (cf., Birkinshaw, Toulan & Arnold, 2001; Frost, Birkinshaw & Ensign, 2002; Song,

Almeida & Wu, 2003). Rather than focusing on subsidiary autonomy, the embedded autonomy

framework developed in this paper identifies team autonomy as a critical locus of differentiation

that enables appropriate decisions to be made. Rather than focusing on cross-subsidiary

knowledge flows, the framework emphasizes team embeddedness in the form of knowledge

31

inflows to the team as a critical integrating mechanism. The framework thus views the macro-

level challenges of multinational organizations from the micro-level perspective of their project

teams, and shifts attention toward the micro-level conditions that enhance team effectiveness.

By highlighting the influence risks as well as the informational benefits of knowledge

flows, the study also extends and enriches theories of organizational knowledge. Research on the

sharing of dispersed knowledge within firms usually focuses on technical, social, or cognitive

factors that impede knowledge flows, such as search and transfer problems (Hansen, 1999), lack

of psychological safety (Edmondson, 1999), or limited absorptive capacity (Szulanski, 1996).

However, organization scholars have long recognized that knowledge is a source of power in

organizations, and recent ethnographic studies of knowledge sharing processes and practices

echo this view (e.g., Bechky, 2003; Carlile, 2002). The implication is that knowledge flows can

be used to influence as well as inform, making their value more complex to evaluate and realize

than is often recognized. For managers, this suggests that investing in knowledge management

innovations like document databases and communities of practice may be only a partially

effective strategy for improving project performance. Even successful search and transfer efforts

do not necessarily help teams to perform more effectively and might even hurt, not just because

the knowledge may be poor quality or outdated but because it may be biased, misleading, or

intended to persuade rather than assist those who use it.

This study also contributes to theories of team effectiveness by explicating the benefits of

autonomy for helping teams to resist excessive external influence. Prior research has paid much

more attention to the intra-team implications of autonomy, perhaps due partly to the internal

focus of most team studies (Ancona, 1993). As the effects of organizational contexts on team

processes and outcomes attract increasing attention from group scholars (e.g., Mannix, Neale, &

32

Wageman, 1999), however, the question arises of how teams can manage their external relations

effectively. Scholars have noted the need for ambassadorial activities as a means to manage

organizational politics (Ancona & Caldwell, 1992), but this study emphasizes that the risks of

exposure to excessive external influence requires buffering abilities as well as outreach activities.

The implication is that autonomy can be usefully viewed as a group resource that enables teams

to avoid excessive external influence.

Another theoretical implication of this study is that buffering and bridging are not

mutually exclusive boundary management techniques for teams, where buffering refers to

disengagement while bridging refers to engagement with the team environment (Scott, 2003).

Indeed, successful boundary management for teams seems to require both buffering and bridging

(as for organizations, cf. Fennell & Alexander, 1987; Meznar & Nigh, 1995). Without autonomy,

teams engaged in bridging activities such as knowledge gathering risk being unable to protect

themselves from influence attempts by outsiders. Without these bridging activities, however,

teams risk overlooking important information in their environments, especially if they are further

insulated by high decision-making autonomy (cf. Janis, 1982). These risks help explain why

teams with high embeddedness but low autonomy or low embeddedness but high autonomy may

perform worse than those that run lower risks of either excessive influence or excessive isolation.

Project work within multinational organizations is an under-explored area for

organizational studies, yet one that is becoming increasingly critical as globalization creates new

challenges across economic, social, and military domains. Some aspects of the empirical setting

studied here are probably unique, and knowledge-intensive work certainly lends itself more to

the risks of isolation and influence, but these issues are salient in many fields. The growing role

of information resources and specialist expertise in advanced economies increasingly places

33

knowledge flows at the center of influence dynamics in organizations, while simultaneously

making isolation an unviable strategy (Child & McGrath, 2001). Still, the extent to which the

theory and results of the study hold in other settings is worthy of further exploration. For

example, the risk that knowledge is used to influence rather than inform might be less acute in

settings where the knowledge needed for tasks is more standardized or fact-based, cause-effect

relationships are more predictable, novices need rely less on experts, there is more consensus

around organizational goals and means, and knowledge is shared more through codification and

document databases than through interpersonal exchanges.

Other questions raised by this study emerge from the limitations of the data, which did

not allow, for example, for direct examination of the various types of influence exerted by

knowledge providers or of internal team processes such as their levels of conflict, which would

further illuminate the findings. Future research could also usefully distinguish between different

knowledge providers, perhaps using network methods, to see whether autonomy has different

implications under different configurations of embeddedness. Likewise, autonomy relative to

different external stakeholders could be unpacked to establish whether the benefits of

independence depend on who else is involved in decision-making.

In conclusion, this paper has focused attention on the reality that the fundamental

multinational management mandate of combining local differentiation and global integration is

not simply a structural challenge at the level of national subsidiaries. Instead, it is a daily

challenge for the project teams that carry out the critical everyday tasks of the multinational

organization. As they strive to address this challenge, embedded autonomy can help them to

avoid the risks of excessive isolation or excessive influence and make decisions that are both

informed and independent.

34

REFERENCES

Aiken, L. S., & West, S. G. 1991. Multiple regression: Testing and interpreting interactions.

Newbury Park, CA: Sage.

Ancona, D. G. 1993. The classics and the contemporary: A new blend of small group theory. In

J. K. Murninghan (Ed.), Social psychology in organizations: Advances in theory and

research: 225-243. Englewood Cliffs, NJ: Prentice Hall.

Ancona, D. G., & Caldwell, D. F. 1992. Bridging the boundary: External activity and

performance in organizational teams. Administrative Science Quarterly, 37: 634-665.

Bacharach, S. B., & Lawler, E. J. 1980. Power and politics in organizations. San Francisco, CA:

Jossey-Bass.

Barnard, C. I. 1938. The Functions of the Executive. Boston: Harvard University Press.

Bartlett, C.A., & Ghoshal, S. 1989. Managing across borders: The transnational solution. Boston:

Harvard Business School Press.

Bechky, B. 2003. Share meaning across occupational communities: The transformation of

understanding on a production floor. Organization Science, 13: 312-329.

Birkinshaw, J., Ghoshal, S., Markides, C. C., Stopford, J. & Yip, G. (Eds.). 2003. The future of

the multinational company. New York: John Wiley.

Birkinshaw, J., Hood, N., & Jonsson, S. 1998. Building firm-specific advantages in multinational

corporations: The role of subsidiary initiative. Strategic Management Journal, 19: 221-242.

Birkinshaw, J., Toulan,O., & Arnold, D. 2001. Global account management in multinational

corporations: Theory and evidence. Journal of International Business Studies, 32: 231-248.

Brown, G., Lawrence, T. B., & Robinson, S. L. 2005. Territoriality in organizations. Academy of

Management Review, 30: 577-594.

Burt, R. S. 1992. Structural holes: The social structure of competition. Cambridge, MA:

Harvard University Press.

Campbell, D. T. & Fiske, D. W. 1959. Convergent and discriminant validation by the multitrait-

multimethod matrix. In D. N. Jackson & S. Messnick (eds.), Problems in human

assessment. New York: McGraw-Hill.

Carlile, P. 2002. A pragmatic view of knowledge and boundaries: Boundary objects in new

product development. Organization Science, 13(4): 442-455.

Child, J., & McGrath, R. G. 2001. Organizations unfettered: Organizational form in an

information-intensive economy. Academy of Management Journal, 44(6): 1135-1148.

35

Cohen, S. G. & Ledford, G. E. 1994. The effectiveness of self-managing teams: A quasi-

experiment. Human Relations, 47(1): 13-34.

Cyert, R. M., & March, J. G. 1963. A behavioral theory of the firm. Cambridge, MA: Blackwell.

Earley, P. C. & Gibson, C. B. 2002. Multinational work teams: A new perspective. Mahwah,

NJ: Lawrence Erlbaum.

Edmondson, A. C. 1999. Psychological safety and learning behavior in work teams.

Adminstrative Science Quarterly, 44: 350-383.

Edmondson, A. C. 2002. The local and variegated nature of learning in organizations: A group-

level perspective. Organization Science, 13(2): 128-146.

Eisenhardt, K. M., & Bourgeois, L. J. 1988. Politics of strategic decision making in high-velocity

environments: Toward a midrange theory. Academy of Management Journal, 31(4): 737-770.

Evans, P. 1995. Embedded autonomy: States and industrial transformation. Princeton, NJ:

Princeton University Press.

Feldman, M. S., &March, J. G. 1981. Information in organizations as signal and symbol.

Administrative Science Quarterly, 26: 171-186.

Feldman, S. P. 1988. Secrecy, information, and politics: An essay in organizational decision

making. Human Relations, 41: 73-90.

Fennell, M. L., & Alexander, J. A. 1987. Organizational boundary spanning in institutionalized

environments. Academy of Management Journal, 30: 356-475.

Foss, N. J. & Pedersen, T. 2004. Organizing knowledge processes in the multinational

corporation: An introduction. Journal of International Business Studies, 35(5): 3490-349.

Frost, T. S., Birkinshaw, J. M., & Ensign, P. C. 2002. Centers of excellence in multinational

corporations. Strategic Management Journal, 23(11): 997-1018.

Garnier, G. H. 1982. Context and decision-making autonomy in the foreign affiliates of US

multinational corporations. Academy of Management Journal, 25(4): 893-908.

Granovetter, M. 1985. Economic action and social structure: The problem of embeddedness.

American Journal of Sociology, 91: 481-510.

Grant, R. M. 1996. Prospering in dynamically competitive environments: Organizational

capability as knowledge integration. Organization Science, 7: 375-387.

Gresov, C., & Stephens, C. 1993. The context of interunit influence attempts. Administrative

Science Quarterly, 38: 252-276.

Gupta, A.K., & Govindarajan, V. 2000. Knowledge flows within multinational corporations.

Strategic Management Journal, 21: 473-496.

36

Haas, E. B. 1990. When knowledge is power: Three models of change in international

organizations. Berkeley, CA: University of California Press.

Hackman, J. R. 1987. The design of work teams. In J. Lorsch (ed.), Handbook of organizational

behavior. Englewood Cliffs, NJ: Prentice-Hall.

Haas, M. R. 2006. Knowledge gathering, team capabilities, and project performance in

challenging work environments. Management Science, 52(8): 1170-1184.

Haas, M. R., & Hansen, M. T. 2005. When using knowledge can hurt performance: The value of

organizational capabilities in a management consulting company. Strategic Management

Journal, 26: 1-24.

Hansen, M. T. 1999. The search-transfer problem: The role of weak ties in sharing knowledge

across organization subunits. Administrative Science Quarterly, 44: 82-111.

Hansen, M.T., & M. R. Haas, M. R. 2001. Competing for attention in knowledge markets:

Electronic document dissemination in a management consulting company. Administrative

Science Quarterly, 46: 1-28.

Hargadon, A., & Sutton, R. I. 1997. Technology brokering and innovation in a product

development firm. Administrative Science Quarterly, 42: 716-749.

Haunschild, P. R., & Beckman, C. M. 1998. When do interlocks matter?: Alternative sources of

information and interlock influence. Administrative Science Quarterly, 43: 815-845.

Huber, G. 1991. Organizational learning: The contributing processes and literatures.

Organization Science, 2(1): 88-115.

Janis, I. L. 1982. Groupthink (2nd

ed). Boston, MA: Houghton Mifflin.

Jehn, K. 1995. A multimethod examination of the benefits and detriments of intragroup conflict.

Administrative Science Quarterly, 40(2): 245-282.

Kanter, R. M. 1988. When a thousand flowers bloom: Structural, collective, and social

conditions for innovation in an organization. In B. M. Staw & L. L. Cummings (eds.),

Research in organizational behavior, 10: 169-211. Greenwich, CG: JAI Press.

Katz, D., & Kahn, R. L. 1966. The social psychology of organizations. New York: Wiley.

Kean, T. H. & Hamilton, L. 2004. The 9/11 commission report: Final report of the national

commission on terrorist attacks upon the United States. St Martin’s Press.

Kirkman, B. & Rosen, B. 1999. Beyond self-management: Antecedents and consequences of

team empowerment. Academy of Management Journal, 42(1): 58-74.

Kogut, B., & Zander, U. 1993. Knowledge of the firm and the evolutionary theory of the

multinational corporation. Journal of International Business Studies, 24: 625-645.

37

Kogut, B., & Zander, U. 1996. What firms do? Coordination, identity, and learning.

Organization Science, 7(5): 502-518.

Kostova, T. 1999. Transnational transfer of strategic organizational practices: A contextual

perspective. Academy of Management Journal, 24(2): 308-324.

Langfred, C. W. 2000. The paradox of self-management: Individual and group autonomy in

work groups. Journal of Organizational Behavior, 21: 563-585.

Latham, G. P., Winter, D. C., & Locke, E. A. 1994. Cognitive and motivation effects of

participation: A mediator study. Journal of Organizational Behavior, 15: 49-63.

Lawler, E. J. 1992. Affective attachments to nested groups: A choice-process theory. American

Sociological Review, 57: 327-339.

Lipnak, J. & Stamps, J. 1997. Virtual teams: Reaching across space, time, and organizations

with technology. New York: John Wiley & Sons, Inc.

Long, J. S. 1997. Regression models for categorical and limited dependent variables. Thousand

Oaks, CA: Sage.

Lord, M. D. & Ranft, A. L. 2000. Organizational learning about new international markets:

Exploring the internal transfer of local market knowledge. Journal of International

Business Studies, 31(4): 573-589.

Lyles, M. A. & Salk, J. 1996. Knowledge acquisition from foreign parents in international joint

ventures: An empirical examination. Journal of International Business Studies, 27(5): 877-903.

Mannix, E. A., Neale, M. A., & Wageman, R. 1999. Research on managing groups and teams

(volume II): Groups in context. Stamford, CT: JAI Press.

March, J. S. 1991. Exploration and exploitation in organizational learning. Organization

Science, 2(1): 71-87.

Martin, X. & Salomon, R. 2003. Knowledge transfer capacity and its implications for the theory

of the multinational corporation. Journal of International Business Studies, 34(4): 356-373.

Martinez, J. I. & Jarillo, J. C. 1989. The evolution of research on coordination mechanisms in

multinational corporations. Journal of International Business Studies, 20(3): 489-514.

McMillan, C. J., Hickson, D. J., Hinings, C. R., & Schneck, R. E. 1973. The structure of work

organizations across societies. Academy of Management Journal, 16(4): 555-569.

Menon, T., & Pfeffer, J. 2003. Valuing internal vs. external knowledge: Explaining the

preference for outsiders. Management Science, 49(4): 497-513.

Meznar, M. B., & Nigh, D. 1995. Buffer or bridge? Environmental and organizational

determinants of public affairs activities in American firms. Academy of Management

38

Journal, 4: 975-996.

Miller, S. T., & Ross, M. 1975. Self-serving biases in the attribution of causality: Fact or fiction?

Psychological Bulletin, 82: 93-118.

Neeter, J. W., Wasserman, S. & Kutner, M. H. 1985. Applied linear statistical models (2nd ed.).

Homewood, IL: Richard D. Irwin Inc.

Nohria, N., & Ghoshal, S. 1997. The differentiated network: Organizing multinational

corporations for value creation. San Francisco: Jossey-Bass.

Pearce, J. A., & Ravlin, E. C. 1987. The design and activation of self-regulating work groups.

Human Relations, 40: 751-782.

Pettigrew, A. M. 1973. The politics of organizational decision-making. London, UK: Tavistock.

Pfeffer, J. 1981. Power in organizations. Marshfield, MA: Pitman Publishing.

Pfeffer, J., & Salancik, G. 1978. The external control of organizations: A resource dependency

perspective. New York: HarperCollins.

Prahalad, C.K., & Doz, Y. L. 1987. The multinational mission. New York: Free Press.

Pugh, D. S., Hickson, D. J., Hinings, C. R., & Turner, C. 1969. The context of organization

structures. Administrative Science Quarterly, 14(1): 91-114.

Reagans, R. & Zuckerman, E. Z. 2001. Networks, diversity, and productivity: The social capital

of corporate R&D teams. Organization Science, 12: 502-517.

Rugman, A. M., & Verbeke, A. 2001. Subsidiary-specific advantages in multinational

enterprises. Strategic Management Journal, 22(3): 237-250.

Schein, E. 1992. Organizational culture and leadership. 2nd

edition. San Francisco: Jossey-Bass.

Scott, W. R. 2003. Organizations: Rational, natural, and open systems. 5th

edition. Upper

Saddle River, NJ: Prentice-Hall.

Selznick, P. 1949. TVA and the grass roots: A study in the sociology of formal organization.

Berkeley, CA: University of California Press.

Shapiro, D. L., Furst, S. A., Spreitzer, G. M., & Von Glinow, M. A. 2002. Transnational teams in

the electronic age: Are team identity and high performance at risk? Journal of

Organizational Behavior, 23: 455-467.

Sherif, M. Superordinate goals in the reduction of intergroup conflict. American Journal of

Sociology, 4: 349-356.

Simon, H. 1962. The architecture of complexity. Proceedings of the American Philosophical

Society, 106(6): 467-482.

39

Song, J., Almeida, P. & Wu, G. 2003. Learning-by-hiring: When is mobility more likely to

facilitate interfirm knowledge transfer? Organization Science, 49(4): 351-365.

Snell, S. A., C. C. Snow, S. C. Davison, D. C. Hambrick. 1998. Designing and supporting

transnational teams: The human resource agenda. Human Resource Management, 37: 147-158.

Spekman, R. E. 1979. Influence and information: An exploration of the boundary role person’s basis

of power. Academy of Management Journal, 22: 104-117.

Starbuck, W. H. 1992. Learning by knowledge-intensive firms. Journal of Management

Studies, 29(6): 713-740.

Szulanski, G. 1996. Exploring internal stickiness: Impediments to the transfer of best practice within

the firm. Strategic Management Journal, Winter Special Issue, 17: 27-43.

Szulanski, G., Cappetta, R. & Jensen, R. J. When and how trustworthiness matters: Knowledge

transfer and the moderating effect of causal ambiguity. Organization Science, 15(5): 600-613.

Tsai, W. 2001. Knowledge transfer in intraorganizational networks: Effects of network position

and absorptive capacity on business unit innovation and performance. Academy of

Management Journal, 44(5): 996-1004.

Tsai, W. 2002. Social structure of “coopetition” within a multiunit organization: Coordination,

competition, and intraorganizational knowledge sharing. Organization Science, 13(2): 179-190.

Tushman, M. L. 1977. Communication across organizational boundaries: Special boundary roles

in the innovation process. Administrative Science Quarterly, 22: 581-606.

Tushman, M. L. & O’Reilly, C. A. 1996. Ambidextrous organizations: managing evolutionary

and revolutionary change. California Management Review, 38(4): 8-30.

Uzzi, B. 1996. The sources and consequences of embeddedness for the economic performance of

organizations: The network effect. American Sociological Review, 61(4): 674-698.

Vaughan, D. 1990. Autonomy, interdependence, and social control: NASA and the Space Shuttle

Challenger. Administrative Science Quarterly, 35(2): 225-257.

Wageman, R., & Hackman, J. R. 2004. Development of the team diagnostic survey. Unpublished

manuscript. Tuck School of Business, Dartmouth University.

Wellins, R. S., Wilson, R., Katz, A. J., & Laughlin, P. 1990. Self-directed teams: A study of

current practice. Pittsburgh, PA: Development Dimensions International.

Zellmer-Bruhn, M. E., & Gibson, C. B. 2006. Multinational organization context: Implications

for team learning and performance. Academy of Management Journal, 49(3): 501-518.

Zukin, S. & DiMaggio, P. 1990. Structures of capital: The social organization of the economy.

New York: Cambridge University Press.

40

FIGURE 1. Information and Independence Benefits Associated with Combinations of

Team Embeddedness and Team Autonomy

FIGURE 2. Interaction Plot of Team Embeddedness with Team Autonomy

Note: To illustrate the direction and magnitude of effects, low values are set at 1 standard deviation below the

mean, high values are set at 1 standard deviation above the mean, and the plots are constructed using OLS

regression.

41

TABLE 1. Descriptive Statistics and Bivariate Correlations (n = 96)

Variable Mean S.D. Min Max 1 2 3 4 5 6 7 8 9

1. Project quality (dep. variable) 2.02 0.54 1 3

2. Project type 0.52 0.50 0 1 .08

3. Project cost [log] 5.34 0.85 3.00 6.84 .03 .13

4. Project duration [log] 5.61 0.75 3.76 7.27 -.10 -.16 .21

5. Region 0.21 0.41 0 1 -.07 .03 .06 .08

6. Division 0.20 0.40 0 1 .13 .11 -.09 .17 .00

7. Tenure 8.46 4.40 1.00 25.5 .09 .12 -.15 -.14 -.02 -.12

8. Late respondents 0.56 0.41 0.00 1.00 .01 .40 -.01 .11 -.07 -.06 .09

9. Core respondents 0.76 0.22 0.17 1.00 .16 .14 -.05 -.10 .09 .05 .14 .03

10. Real team 3.82 0.56 2.38 5.00 .14 .19 -.14 -.10 -.06 .24 -.07 .05 .36

11. Team location 0.80 0.40 0 1 -.03 .31 .06 -.02 .06 .05 .02 .02 -.05

12. Team size 8.48 4.15 2 23 .14 .29 .41 .00 .20 -.14 -.13 -.07 -.10

13. Team knowledge 3.62 0.54 2.00 4.60 .07 -.10 .09 .03 -.02 .05 .00 -.14 .02

14. Team autonomy1 3.62 0.31 2.75 4.32 .15 .04 .03 .03 -.05 .17 .05 -.15 .03

15. Team embeddedness1 3.00 0.43 1.91 3.98 .10 .00 -.03 -.02 .18 .19 -.02 -.04 .05

16. Team embeddedness (country knowledge)1 3.08 0.47 1.88 4.17 .12 .09 -.01 .01 .12 .25 -.04 -.02 .01

17. Team embeddedness (technical knowledge)1 2.91 0.44 1.75 3.92 .05 -.06 -.09 -.05 .23 .14 .00 -.08 .08

18. Team embeddedness (inside knowledge)1 0.51 0.03 0.44 0.60 .01 .06 -.00 -.00 .20 .10 .04 .03 .07

19. Team embeddedness (outside knowledge)1 3.18 0.48 1.84 4.89 .14 -.05 -.06 -.03 .16 .22 -.06 -.11 .02

20. Project country wealth 0.14 0.21 0.01 0.76 .19 -.01 .08 -.00 -.16 .06 -.13 -.12 .03

21. Project novelty 3.73 0.57 2.00 5.00 .13 .02 .16 .03 .03 .15 .27 -.21 .26

22. Project complexity 3.77 0.58 2.00 4.88 .06 .20 .33 .18 .01 .07 .23 -.01 .18

Variable 10 11 12 13 14 15 16 17 18 19 20 21 22

11. Team location .19

12. Team size -.23 .21

13. Team knowledge .08 -.14 -.04

14. Team autonomy1 .11 -.04 .02 .11

15. Team embeddedness1 .15 -.08 .06 .19 .15

16. Team embeddedness (country knowledge)1 .18 .00 .08 .19 .15 .94

17. Team embeddedness (technical knowledge)1 .12 .14 .04 .19 .12 .95 .80

18. Team embeddedness (inside knowledge)1 .14 -.07 .09 .21 .03 .84 .77 .82

19. Team embeddedness (outside knowledge)1 .11 -.07 .03 .12 .23 .90 .86 .83 .52

20. Project country wealth .17 -.02 -.05 .05 -.05 -.09 -.11 -.06 -.12 -.04

21. Project novelty -.02 .13 .15 .25 .12 .23 .23 .21 .18 .22 -.02

22. Project complexity .03 .20 .25 .21 .10 .22 .27 .16 .21 .17 -.04 .63

1 Variable is standardizing by subtracting the mean and dividing by the standard deviation in the analyses.

42

TABLE 2. Results of Ordinal Logit Analyses of Project Quality (n=96) †

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Project type -1.07

(0.65)

-1.27*

(0.69)

-1.06

(0.65)

-1.25*

(0.74)

-1.40*

(0.74)

-1.13

(0.76)

Project cost -0.11

(0.32)

-0.12

(0.32)

-0.10

(0.32)

-0.22

(0.34)

-0.18

(0.35)

-0.19

(0.35)

Project duration -0.30

(0.36)

-0.33

(0.37)

-0.28

(0.37)

-0.47

(0.40)

-0.61

(0.41)

-0.44

(0.41)

Region -0.36

(0.63)

-0.27

(0.64)

-0.39

(0.64)

-0.54

(0.68)

-0.46

(0.70)

-0.36

(0.72)

Division 1.92***

(0.74)

1.81**

(0.76)

1.88**

(0.75)

1.97**

(0.80)

1.91**

(0.82)

1.88**

(0.81)

Tenure 0.12*

(0.06)

0.11*

(0.06)

0.12*

(0.06)

0.09

(0.07)

0.14*

(0.07)

0.10

(0.07)

Late respondents 1.47*

(0.75)

1.79**

(0.79)

1.44*

(0.76)

2.08**

(0.87)

2.13**

(0.89)

2.36***

(0.91)

Core respondents

1.49

(1.24)

1.55

(1.24)

1.51

(1.24)

1.32

(1.31)

2.10

(1.40)

1.49

(1.36)

Real team 0.90*

(0.53)

0.91*

(0.54)

0.87

(0.53)

1.31**

(0.61)

1.28**

(0.64)

1.54**

(0.65)

Team location -1.04

(0.68)

-0.99

(0.69)

-1.01

(0.69)

-1.35*

(0.75)

-1.51*

(0.80)

-1.52**

(0.77)

Team size 0.34***

(0.09)

0.35***

(0.09)

0.33***

(0.09)

0.39***

(0.10)

0.41***

(0.10)

0.44***

(0.11)

Team knowledge 0.42

(0.44)

0.42

(0.45)

0.39

(0.45)

0.29

(0.48)

0.25

(0.49)

0.54

(0.51)

Team autonomy1

0.56**

(0.27)

0.59*

(0.30)

0.67**

(0.31)

0.54*

(0.31)

Team embeddedness1

0.08

(0.26)

0.09

(0.29)

Team embeddedness (country knowledge)1

0.24

(0.47)

Team embeddedness (technical knowledge)1

-0.11

(0.47)

Team embeddedness (inside knowledge)1

-0.68*

(0.35)

Team embeddedness (outside knowledge)1

0.61*

(0.33)

Team embeddedness * Team autonomy

0.90***

(0.34)

Team embeddedness (country knowledge) * Team autonomy 1.55***

(0.58)

Team embeddedness (technical knowledge) * Team autonomy -0.50

(0.51)

Team embeddedness (inside knowledge) * Team autonomy

0.41

(0.34)

Team embeddedness (outside knowledge) * Team autonomy

0.63**

(0.32)

Cut 1 4.70

(3.60)

4.63

(3.61)

4.01

(3.97) 4.19

(4.17) 6.60

(4.28)

Cut 2 9.44

(3.82)

9.36

(3.83)

9.34

(4.19) 9.77

(4.44) 12.31

(4.61)

Degrees of freedom 12 13 15 17 17

Log likelihood -59.52 -59.47 -53.12 -51.24 -50.78

LL χ2 ratio test2 0.10 12.80*** 16.56*** 17.48***

Pseudo R-squared 0.18 0.18 0.26 0.29 0.30

† Standard errors are in parentheses 1 Variable is standardized by subtracting the mean and dividing by the standard deviation; 2 Compared to model 1

•p < .10, ••p < .05, •••p < .01; two-tailed test for variable coefficients.

43

TABLE 3a. Results of Ordinal Logit Analyses of Project Quality

for High v. Low Demand Projects (n=96) †

Model 7a Model 7b Model 8a Model 8b

Median split criteria Project novelty Project complexity

Low High Low High

Team autonomy 1 -0.16

(0.68)

0.79

(0.56)

0.84

(0.55)

0.60

(0.52)

Team embeddedness 1 -0.18

(0.59)

0.29

(0.47)

-0.25

(0.52)

0.47

(0.47)

Team embeddedness * Team autonomy 0.26

(0.90)

1.18**

(0.56)

0.67

(0.69)

0.94*

(0.58)

Cut 1 14.88

(10.71)

-5.29

(7.34)

8.08

(9.73)

-0.97

(6.22)

Cut 2 23.02

(11.89)

1.18

(7.26)

16.38

(10.56)

4.00

(6.30)

Number of observations 48 47 50 45

Degrees of freedom 2 16 16 16 16

Log likelihood -19.56 -23.03 -20.82 -26.70

LL χ2 ratio test 3 0.38 10.28** 3.22 6.40*

Pseudo R-squared 0.47 0.39 0.44 0.29

TABLE 3b. Results of Ordinal Logit Analyses of Project Quality

for High v. Low Prominence Projects (n=96) †

Model 9a Model 9b Model 10a Model 10b

Median split criteria Project cost Project country wealth

Low High Low High

Team autonomy 1 0.84

(0.53)

1.11

(0.74)

-0.96

(1.06)

0.98

(0.69)

Team embeddedness 1 0.30

(0.54)

-0.52

(0.51)

1.39

(1.44)

-0.58

(0.53)

Team embeddedness * Team autonomy 0.62

(0.58)

2.17**

(0.86)

0.59

(0.94)

1.72**

(0.79)

Cut 1 7.85

(7.46)

6.68

(8.32)

3.51

(11.66)

2.10

(8.00)

Cut 2 15.45

(8.06)

14.45

(9.33)

14.80

(12.95)

8.87

(8.38)

Number of observations 45 46 40 40

Degrees of freedom 2 14 14 16 16

Log likelihood -20.18 -20.51 -15.14 -17.70

LL χ2 ratio test 3 5.20 11.02** 3.66 10.64**

Pseudo R-squared 0.46 0.41 0.48 0.39

† Standard errors are in parentheses

1 Variable is standardized by subtracting the mean and dividing by the standard deviation. 2 All models include control variables in Table 4 and an additional dummy variable for missing cost data, except 9a and 9b which exclude cost controls. 3 Compared to the same models including control variables only

•p < .10, ••p < .05, •••p < .01; two-tailed test for variable coefficients.

44

1 The concept of team embeddedness has parallels in the sociological literature on organizational

embeddedness, which examines the relations between firms and their implications for economic logics of

exchange (Granovetter, 1985; Uzzi, 1996). The concepts of team embeddedness and organizational

embeddedness differ in their levels of analysis but they share an emphasis on the nested nature of

organizing systems within broader webs of relations. Although the specific team-environment relations of

interest in the present study are knowledge flows, the concept of team embeddedness, like organizational

embeddedness, could be extended to address other types of relations, including social and cultural ties (cf.

Zukin & DiMaggio, 1990). In the multinational literature, similarly, embeddedness has been used to refer

to the involvement of foreign subsidiaries in their host countries’ knowledge development systems

(Rugman & Verbeke, 2001), as well as more broadly to the social, organizational, and relational

requirements for successful transfers of strategic organizational practices (Kostova, 1999).

2 An alternative measure of project quality constructed by summing each project’s raw quality dimensions

scores correlated 0.86 with the ordinal measure and generated the same substantive results.

3 Calculating Cronbach’s alphas for the knowledge flow measures in this study is not appropriate because

these are composite indices rather than constructs in which the underlying items should be correlated with

each other or show high agreement within teams.

4 Because calculation and interpretation of interaction effects in nonlinear models can be problematic, I also

generated the marginal effects for the interaction terms and ran the models using an ordinary least squares

specification instead. These two alternative approaches both generated the same pattern of results.

5 The quality-monitoring unit reported that seven days was a reasonable estimate of the time it usually took to

finalize the results of a panel’s review and notify the team of its project quality rating. Shorter and longer cut-off

periods were used to construct alternative measures, but these did not change the results.

Feb. 1, 2008

Dear Professor Haas:

Thank you for submitting your manuscript to AMJ. All three reviewers have

completed their evaluations of your paper, "Embedded Autonomy: Project

Teams and Knowledge Work in Multinational Organizations"

(Manuscript #2007-0880). Although the reviewers have some critical

concerns, they all feel your manuscript has good potential. I agree. As a

result, I would like to offer you the opportunity to revise and resubmit your

paper for further evaluation at the AMJ. Please note, however, that a

substantial revision is expected as the reviewers have some major

reservations.

I was fortunate to obtain excellent reviews from highly qualified experts in

this area. The reviewers' comments are of very high quality, and they offer

detailed, thorough, and constructive suggestions for revising the paper. Please

note that this invitation carries no guarantee for the ultimate success of your

revision, even if you believe every issue is successfully resolved. Please also

note that even through you are not required to agree with the reviewers or

myself, you are encouraged to be as responsive as possible.

The reviewers are very receptive to your study, and have several positive

comments about your paper. Reviewer 1 believes that your paper develops “a

very interesting construct” and draws on “a wonderful dataset of a relatively

large sample of project teams where real project quality is assessed (first

paragraph). Reviewer 2 notes, “your study is thoughtfully designed, carried

out with notable and consistent care, and potentially makes a genuine

contribution” (first paragraph). In addition, Reviewer 3 says, “The study of

project teams and knowledge flows in MNEs is relevant and important” (first

paragraph). I share with the reviewers’ overall enthusiasm for your study, and

concur with their assessment regarding the need for additional scholarship to

be completed within the general research domain of your study.

EDITOR R. Duane Ireland Texas A&M University direland@mays.tamu.edu ASSOCIATE EDITORS Peter A. Bamberger Technion - Israel Institute of Technology peterb@tx.technion.ac.il Jason A. Colquitt University of Florida jason.colquitt@cba.ufl.edu K. Michele Kacmar University of Alabama amjmk@ua.edu

David J. Ketchen, Jr.

Auburn University ketchda@auburn.edu

Elizabeth W. Morrison

New York University

emorriso@stern.nyu.edu

Michael G. Pratt

University of Illinois at Urbana-Champaign

amjpratt@business.uiuc.edu

Wm. Gerard Sanders Brigham Young University

gsanders@byu.edu

Wenpin Tsai

The Pennsylvania State University

wtsai@psu.edu

MANAGING EDITOR Michael P. Malgrande Academy of Management mmalgrande@pace.edu 914-923-2607 Fax: 914-923-2615 PRODUCTION EDITOR COPY EDITOR Persephone Doliner Ithaca, NY pdoliner@twcny.rr.com 607-277-5283

2

Despite these strengths, the reviewers also identify some serious weaknesses in your paper. Clearly,

the reviewers want to make this the best paper possible, and they are very thorough in reading your

paper, in raising issues, and in making constructive suggestions. Their comments are fairly

consistent and clearly stated, so I will not repeat all of them in this letter. Instead, I will highlight the

more salient issues that you will need to address in your revision.

1. Contribution. You need to articulate your contribution and clarify how your study is

theoretically linked to and disconnected from prior work on knowledge and project teams,

particularly the research reported in Haas 2006. As Reviewer 1 points out, “While the prior

research doesn’t develop the specific construct of embedded autonomy, it does look examine the

interaction between knowledge gathering and autonomy or what you call embedded autonomy.

In this way, your H1 has already been tested and supported.” (point 5). You might want to drop

your current H1 and focus on new theoretical insights in your other hypotheses. It is important

to convince your readers that your paper represents a major theoretical breakthrough rather than

just an incremental extension of prior work. I encourage you to identify a theoretical advance

and develop it more clearly and convincingly.

You also need to provide solid empirical evidence to support your proposed contribution. Given

that the different aspects of team embeddedness in your study are very highly correlated with one

another (as reported in your table 1), it is not clear what we actually gain by separating these

aspects and testing their relative difference in interacting with autonomy in different hypotheses.

If you want to identify the boundary conditions that split team embeddedness into different

aspects, you need to make sure that you can clearly differentiate these aspects empirically. In

your revision, you must present a stronger case that your study provides a more compelling body

of learning for both management theory and knowledge.

2. Theory. Reviewer 2 questions your motivation to choose the four boundary conditions for

studying the value of embedded autonomy. This reviewer asks “Why these four and not others?”

(point 7). Indeed, Reviewer 1 suggests a different boundary condition for you to consider (point

6). You need to convince your readers that you have a strong, coherent theory that guides you to

select your variables. I suggest you put more efforts in developing a good theory for the

boundary conditions for embedded autonomy, and use the theory to justify why particular

boundary conditions are included in your study.

3. Embeddedness. All three reviewers question your definition of embeddedness. Reviewer 3 is

concerned about the way you define embeddedness in terms of both internal and external factors,

and urges you to provide clearer explanations (points 1 and 2). Reviewer 2 has a similar

concern, and asks you to clarify in what level of environment a team is embedded. In addition,

Reviewer 1 questions whether you really study embeddedness (point 4) given your emphasis on

knowledge flow in operationalizing embeddedness. As this reviewer points out, “Knowledge

flows don’t tell us the extent to which the teams actually pay attention to and use the knowledge

that flows to them” (point 4). Like these reviewers, I am not convinced that embeddedness is the

right term for your current study unless you change your definition and improve your measures.

Alternatively, you may reconceptualize your idea of knowledge flows and use a different term to

replace embeddedness in your study.

3

4. Conceptual Development. The reviewers also raise concerns about the logic and premises

underlying your predictions. Some of the arguments leading to your hypotheses are not very

convincing. You need to clarify your logic and strengthen your arguments. In particular,

Reviewer 3 suggests you rework your arguments for H3 (point 6). Reviewer 2 is concerned

about team usage of knowledge in your arguments, and asks you to pay attention to the

quality/relevance of available knowledge (point 4). Reviewer 1 believes that some of your

hypotheses are intertwined (points 6 and 9) and asks you to clarify your logic. This reviewer

also suggests you consider within team dynamics and power in your theory development (point

18). Considered collectively, the theory section of your paper is not developed clearly or

sufficiently enough to provide the basis for your hypothesized relationships. In your revision,

the theoretical and/or conceptual basis for your hypotheses must be substantially strengthened.

5. Qualitative Data. You mention your interviews and qualitative research many times in your

paper. But, you provide little content about your qualitative data in the paper. All three

reviewers have questions about your qualitative data (Reviewer 1, point 10; Reviewer 2, points

12-14 and 17; Reviewer 3, point 8). You need to provide more detailed information about your

qualitative data in terms of how you collect, analyze, and incorporate such data in your study.

Consider using qualitative data (such as quotes or examples) to clarify some of your statements.

6. Methodological Issues. The reviewers also identify some methodological issues in your paper.

In particular, they have questions about your sample (Reviewer 2, point 18; Reviewer 3, point 7),

survey (Reviewer 3, point 9), variation in dependent variable (Reviewer 2, point 33), validity of

your measures (Reviewer 1, points 13-15), inter-rater reliability (Reviewer 1, point 16), and

alternative explanations for your study (Reviewer 1, points 11, 12, 17). You need to clearly

address all these concerns in your revision.

The reviewers have provided you many other comments that I would encourage you to consider

carefully. These comments are insightful and reflect efforts to help improve your paper. I urge you

not to ignore these other ideas, or take them less seriously because I did not emphasize them.

I sincerely hope that you will accept this invitation to revise and resubmit the manuscript. Please

consider our remarks carefully and then let me know if you do not intend to revise and resubmit the

manuscript. If you do accept the revision invitation, please follow the following procedures:

Procedures for Revising Your Manuscript

While revising your manuscript, please consider each reviewer comment carefully given that even

relatively minor comments can sometimes trigger significant improvements. Once the revision is

completed, please focus your responses to reviewers’ document mainly (but not necessarily

exclusively) on the issues highlighted in my letter rather than on the specific points raised by each

reviewer. Our intent is to have you spend most of your time improving your paper rather than

documenting in great detail all specific changes you made while revising it. Accordingly, please

indicate (in approximately 10-12 pages) how you have responded to the issues raised in the decision

letter. The decision letter synthesizes the reviewers and my concerns; as such, dealing with these

points will generally ensure that you have addressed the major issues. However, if you decide not to

make a substantive change suggested by a reviewer, I advise you to explain your rationale, even if

4

that particular point was not mentioned in my letter. At your discretion, you may also decide to

present an original reviewer or action editor comment in your responses’ document before

describing the action(s) you took in response to individual comments or requests. Overall, our intent

is to maximize your freedom to choose a presentation style when preparing the responses’ document

that allows you to best deal with the core issues included in the decision letter and in the reviewers’

commentaries. Beginning on a new page, the responses’ document should appear at the end of the

revised manuscript.

This procedure for preparing responses to the reviewers and the action editor is new for AMJ.

Because of this, we will inform the reviewers of this change in procedures so they will adjust their

expectations accordingly.

Instructions for Resubmission on Manuscript Central

Once your revised manuscript is completed, log into http://mc.manuscriptcentral.com/amj and enter

your Author Center, where you will find your manuscript titled listed under “Manuscripts with

Decisions.” Under “Actions,” click on “Create a Revision.” Your manuscript number will have been

modified to denote a revision. IMPORTANT: Your original files are available to you when you

upload your revised paper. Please delete any redundant files before completing the submission.

Our desire is to facilitate timely publication of papers accepted for inclusion in AMJ. As such, please

upload your revision on the Manuscript Central system within four months. Please let me know if

you need more than four months to complete the revision process. We will work with you regarding

the time needed to complete a revision while maintaining AMJ’s commitment to timely editorial

processes and decisions.

Also, as a final note, ensure that your revised manuscript meets every submission requirement in

terms of line spacing, font size and so on. I urge you to ensure that your revised manuscript is in

compliance. You can find the submission specifications in the August Issue (“Style Guide for

Authors”) of AMJ. Note that failure to meet these requirements will result in the manuscript being

returned to you and ultimately delay our review and decision.

It is our intention at the Academy of Management Journal that the review process be constructive

and developmental for both our authors and their research. The reviewers and I think your

manuscript has promise, but much work remains to clarify and validate the contribution your

manuscript offers. Regardless, I hope that you will find the feedback we have provided helpful, and

in keeping with the spirit of your submission. Whether you choose to revise your submission or not,

we hope that you will continue to think of AMJ as a primary publication outlet for your very best

work, both on this topic and others.

Sincerely yours,

Wenpin Tsai

Wenpin Tsai

Academy of Management Journal

5

Feedback for the Author(s)

Reviewer #1 Manuscript #2007-0880

This paper develops the idea of embedded autonomy in project teams in multinational organizations.

Embedded autonomy is a very interesting construct having to do with being relatively independent in

decision making but embedded in their context through knowledge flows to ensure good decision

making. The paper draws on a wonderful dataset of a relatively large sample of project teams where

real project quality is assessed. Overall, I think this research has real promise. That said, I have

some suggestions that I will outline in my points below that will help improve the contribution of

your work and address some of the current limitations. Best of luck to you as you continue to

develop this line of research.

1. I am glad to see you embed your work in team embeddedness theories. However, I think

your two main sub-elements of team autonomy and information flows are very consistent

with models of high involvement coming out of the research of the Center for Effective

Organizations at the University of Southern California, particularly the work of Ed Lawler

(he as a number of books including Ultimate Advantage which may give you an overview).

He describes effective high involvement as being a combination of knowledge, power,

autonomy and rewards. As you can see your embedded autonomy is a combination of two

elements of his model. I think it’s important to situate your work within the research on high

involvement/team design research.

2. I didn’t find your framing of the front end around multinationals to be that compelling. It

made me expect that you were going to be examining subsidiaries of multinational

corporations in your paper. While the analogy may be appropriate in your discussion, I don’t

think it’s the strongest introduction to your ideas. You are dealing with project teams. Why

not just go immediately to the issue at hand which you begin to introduce at the bottom of p.

3.

3. I also found that during the times you referenced the multinational management literature

that it seemed to mix up the level of analysis. For example, on p. 7, you cite concerns about

national subsidiary autonomy from the parent organization. However, your theory is about

project teams. It gets a bit confusing when you so easily cross levels of analysis in your

conceptual development. Might not there some different issues at hand depending on what

level you are referring to?

4. I also wonder whether your definition of embeddedness (i.e., knowledge flows from the

environment to the team) is really embeddedness. Knowledge flows sound more like

embedding forces rather than embeddedness. Knowledge flows don’t tell us the extent to

which the teams actually pay attention to and use the knowledge that flows to them. At the

very least, it seems like in order for knowledge flows to create embeddedness is when the

team is requesting the information but not necessarily when the information comes from the

outside in a way that is forced or unwelcomed. I think the argument is stronger to the extent

that you can show that teams obtain and use the knowledge that is provided to them.

6

5. You cite Haas 2006 as having examined the same question – whether knowledge gathering

combined with decision-making autonomy is related to project performance. This makes me

wonder about the extent to which your work goes beyond this cited research. When I went

back to look at this piece, I had even more concerns about this issue. To what extent does

your research go beyond the research reported in Haas 2006? While the prior research

doesn’t develop the specific construct of embedded autonomy, it does look examine the

interaction between knowledge gathering and autonomy or what you call embedded

autonomy. In this way, your H1 has already been tested and supported. Thus, your key

contribution is a better understanding of the conditions under which embedded autonomy

will make a difference to project quality. It seems like this is an important issue to highlight

in your framing because it highlights the potential for a more limited contribution.

6. One issue you raise that could be an important boundary condition is the presence of

influence attempts. It seems to have come out strongly in your interviews as well. Is this

another relevant boundary condition for embedded autonomy?

7. It strikes me that H2 and H3 are intertwined. Isn’t knowledge from outside the organization

usually more scarce then information within the organization? Also it is important that the

knowledge come from outside the organization? What about from inside the organizations

but outside of the team? Would the logic be similar for this boundary condition?

8. I wonder about your point that projects are more prone to isolation when they are demanding

or complex. Wouldn’t that be particular the time when the organization would want to put

more knowledge in the team’s hands with more outside advisors and such?

9. Similar to my point #6 above, it strikes me that H4 and H5 are also intertwined. Won’t more

complex projects by nature be more highly prominent in the sense that they probably involve

more stakeholders?

10. I like the idea that you have a multi-method study consisting of interview data as well as the

survey data! Bravo. However, I was disappointed to see that you provide almost no

information from your interview data. While I can understand that a full-blown study 1

based on your qualitative data would make your paper very long, in the present form, the

information provided on your interviews provides little content. You tell us about all the

interviews you did but little in the way of what you heard/found. I would at least hope to see

some quotes in your discussion that elaborate or support your quantitative findings. It seems

like that would bring to life your findings. It also seems like your interview data could help

provide information about the kinds of measures you chose and why they were appropriate.

11. Are there any differences you might expect across the two types of project teams? Would the

nature of these projects difference in terms of the need for embedded autonomy? Do

stakeholders play a bigger role in the financial projects, for example?

12. Given there are differences in response rate for teams based on their cost and length, is the

confounding your hypothesis about complexity?

13. Given that you have developed your measures specifically for this study and sample, I think

it would be important to do some additional validation work. Did you compute a

7

confirmatory factor analysis to show the convergence and discrimination of your construct

items across embeddedness and autonomy? Are the two constructs distinct? Also what is the

reliability of your embeddedness scale? For the multitrait, multimethod analysis for your

knowledge scales, why not do this with confirmatory factor analysis? If your sample is too

small, it’s possible to do these measurement analyses at the individual level of analysis.

14. I do worry a bit about the reliability of your single item measures of project novelty and

project complexity. Have these items been used anywhere else? I wonder if the length of the

project (duration) could be a proxy for complexity?

15. I do like how you use archival measures for project quality and for project cost and country

wealth!

16. Did you see how much agreement that there was for each of your measures across the

multiple respondents? What are the rwg scores?

17. I also like how you address a number of alternative explanations that I also had been

considering.

18. I really like in your discussion that you think of this paper as input to building a team-

centered theory of knowledge work in multinational organizations. I would like to see you

further flesh out this theory in your discussion. Clearly an understand of boundary

conditions is important to the building of that theory. I also wonder how things like within

team dynamics and power issues would play out in this theory. Clearly issues for future

conceptual and empirical work.

Best of luck as you continue to develop these ideas.

8

Reviewer #2 Manuscript #2007-0880

Overall, your study is thoughtfully designed, carried out with notable and consistent care, and

potentially makes a genuine contribution. Thus, I congratulate you on work well done in pursuit of

understanding an important phenomenon. To help realize the potential of this work, I believe you

can strengthen as well as enrich your argument in multiple ways, which I outline below. Most of my

concerns focus on apparent gaps or disconnects in your argument – calling for needs to clarify

theoretical connections, logic, descriptions of your sample, procedures, and analyses, and your

interpretations of results. Doing so should help to illuminate your contribution and enable other

researchers to replicate your study if so desired, particularly since you sample only one organization.

Unlike most of my reviews where I can point to one, two or three critical issues, I instead point to

the logical flow of your argument and the underlying premises that frame its identity.

1. None of your appendices seem necessary as appendices. As it, they detract attention from the flow

of your argument. Personally, I would include in your text what you believe is essential and delete

the appendices.

2. When you speak of team embeddedness you refer “to the extent of the relations between teams

and their environments.” This statement warrants clearer specification of the level(s) of

“environment” that you have in mind. A team’s environment may be the division in which it works.

Since the concept of “team embeddedness” is central to your study, this should be addressed more

clearly at the outset. In what level of their environment must they be embedded? Can they be

embedded less or more deeply across different levels?

3. When you talk about the two “double-edged” swords, three of the “edges” describe impacts on a

team: independence, isolation, and information. The fourth, influence, seems to have an opposite

effect. Is not “loss of influence” more consistent with your framing of the other three?

4. One of my most salient concerns pertains to team usage of knowledge. By the time I get fully

immersed in your argument, it seems as if there is too little attention paid to the quality/relevance of

available knowledge, as if all knowledge is equal. Yet, an embedded team may be able to discern

more accurately the potential quality of knowledge.

5. I do not think Figure 1 adds sufficient value to the paragraph that precedes it – it seems too

obvious to warrant additional space.

6. On page 11, you claim that “embedded autonomy can be expected to be most valuable under

conditions where the information benefits provided by embeddedness and the independence benefits

provided by autonomy are particularly high.” As opposed to saying that these “benefits are high,” I

wonder if there is more meaning inherent in the thought that such “benefits are more relevant to or

important for a decision situation at hand.”

7. It may strengthen your argument to explain your reason(s) for choosing the four “boundary

conditions” that you describe on page 11. Why these four and not others?

8. This concern exceeds the focus of your study, but I’ll share it for later if not for now. In your

specification of hypotheses 2-5, it may refine your argument to begin with a specification of “kinds

9

of situations or challenges” at hand -- a situational variable that helps to guide or inform different

arrays of subsequent variables. For example, the extent of value of embedded autonomy if a

knowledge source is outside or inside the organization likely depends upon the knowledge “needed”

for a challenge at hand.

9. On page 13 you say that “the risk of excessive isolation is greater when projects are highly

demanding (i.e., very novel or complex).” Why don’t you then say this more directly: “...greater

when projects are highly novel or complex”?

10. On page 13, in the fifth line above hypothesis 4, I think you mean “former” where you say

“latter.”

11. Where possible, minimize the use of passive voice, such as in sentence one of “research setting”

(page 14), where you say “the hypotheses were tested.”

12. At the beginning of your “data and methods” section, you say “I started with 20 interviews.”

How did you make these selections?

13. On page 15, you mention conducting 18 additional “interviews with leaders and members of

project teams.” When did these occur? It may help to include a timeline that situates and connects

your procedures and data gathering steps.

14. You also say on page 15 that you “conducted a further 25 interviews as part of detailed case

studies of seven teams.” How did you divide up 25 interviews among seven teams?

15. When you say that you “read materials that were generated as these teams worked,” to what

materials are you referring? Examples may help.

16. On page 15 you talk as if the organization you studied no longer exists: “the organization

operated worldwide and supported offices in 100 countries.” Did the organization cease to exist or

are you referring to the time period when you were there.

17. It is unclear why you conducted seven interviews in Russia. What were you expecting to learn

and how did you integrate these into your study?

18. I would like to know more about the U.S. headquarters where most of your teams operate(d). For

example, what is this work environment like – are members of teams interacting with each other or

are they separated? Are these teams mostly ad hoc? Are individual member demographics, such as

age and education, relevant to team differences? I ask this final question partly because you note on

page 16 that “each team member was responsible for researching and designing a specific

component of the project.”

19. On page 16 – middle of the page – you write “informal as well as informal.” I suspect that you

mean “formal as well as informal.”

10

20. I would like to see clearer descriptions of your qualitative data gathering procedures – so that

others could replicate your study if they wanted to do so.

21. On the top of page 18 you say “to collect data from the members of these teams, I sent them a

customized survey.” Does this mean that you sent a customized survey to each “member” or to each

“team”? And what do you mean by a “customized” survey?

22. On page 18 you say “96 of the 120 teams qualified for the study by returning at least 50% of

their core team members’ surveys.” How did you decide upon this cutoff point? What do you mean

by “core” team members?

23. To what degree are “financial” and “technical” projects comparable? At times you seem to refer

to them collectively and at other times separately.

24. On page 18 you say that “each project was evaluated by a different independent panel.” Describe

in more detail what you mean by this. For example, did every single one of the 96 projects have a

completely different evaluative panel?

25. Explain your rationale for multiplying “overall embeddedness scores by the team’s overall

autonomy score” to test hypothesis one.

26. On page 23 you talk about assessing “the extent to which the team was a clearly identifiable

interdependent group.” Explain what you mean by this.

27. Why did you differentiate “late respondents” from other respondents and how did you measure

“lateness”?

28. On page 23 you talk about “the average length of organizational tenure in years at the start of the

project.” Were you measuring tenure of teams, core members, or both?

29. I do not understand why “teams that delivered higher quality projects ... had more late

respondents.” You must have anticipated a potential connection here but I don’t see it in your paper.

30. At the end of your “results” section you include a subsection entitled “alternative explanations.”

I think this belongs in your discussion.

31. I think it would enrich your argument to describe “teams” in greater detail.

32. It seems possible that certain teams had some of the same members as other teams, i.e., that there

were some team members who participated on more than one team. If so, to what extent did this

occur – were some teams substantively similar?

33. There is not much variation in your dependent variable. Perhaps you could address this more

carefully in your discussion.

11

34. Because your study focuses so heavily on ONE organization, I think it would enrich your

argument to describe this one organization more intimately and completely.

12

Reviewer #3 Manuscript #2007-0880

Overall, I was impressed with the paper. The study of project teams and knowledge flows in MNEs

is relevant and important, and the author is familiar with the relevant literature. Furthermore, the

author has a relatively large sample of multinational teams as well as both qualitative and

quantitative data to draw upon. Thus, there is much in the paper to like. There were a few areas that

could use improvement, which I discuss below, broadly in terms of theory, methods, and discussion.

Main Points:

Theory

1. The definition of team embeddedness and autonomy thankfully appear early in the paper (p. 4).

However, considering the importance of, in particular, team embeddedness, a little more

clarification seems warranted. You say that that embeddedness refers to inputs inside and outside

the organization. Does this also include team members themselves? In other words, teams in

which group members communicate more with one another is considered part of team

embeddedness? I thought of embeddedness in terms of being embedded in the environment

outside the team, but this isn’t necessarily true given your definition.

2. Related to my previous point on definitions, I was a bit confused early-on with the notion of

external influence being a problem. It makes sense to me that outside forces can be problematic

in terms of their influence on a team. However, if you are defining embeddedness and autonomy

in terms of both internal and external factors, this relationship is less clear. Internal factors (i.e.

those within the firm), it would seem, would help the team gain internal information and stay on

track in terms of organizational goals. If you assume (as I do, and as you mention in the last

paragraph of p. 7) that the team is acting on behalf of the organization, it would seem that

autonomy from outside forces is more potentially problematic than inside forces. This

differentiation between internal and external becomes more confusing when you refer to

“external influence” (e.g. p. 10), as I am not sure whether by “external” you mean outside the

team, outside the organization, or both.

3. On p. 7 you mention the issue of groupthink, which makes sense to mention given your study of

teams, but in my view groupthink is more about interactions and decision-making within the

team, as well as having outside input, than a team’s level of autonomy. It seems that the level of

embeddedness (more so than autonomy) is critical, as having information outside the team would

reduce the propensity to have groupthink. Also, internal team-level norms, roles, and culture

would make a difference regardless of level of autonomy and embeddedness. For example, the

team could have a culture of questioning its members.

4. Figure 1 should be better explained and justified (p. 10). It makes sense, but even if you are

adapting from Evan’s (1995) work, you need to clarify your reasoning more, particularly since

this is the basis for Hypothesis 1.

5. What do you mean by “most valuable” (p. 11)? Do you mean they will have the best

performance, that they have the potential for the best performance, or something else? The word

“value” is particularly problematic in your hypotheses (pp. 12-14), as it was not defined, seems

13

vague, and does not sound testable. Also, the level of analysis needs to be clarified. Is it of most

value for the team, the organization, or something else?

6. You need to rework your arguments for Hypothesis 3 (pp. 12-13). At first, it sounds as though

you are advocating firm outsiders as improving inputs and autonomy. However, then it sounds as

though you are advocating information from insiders since they have similar identities, interests,

etc., and that outsiders may have external influence that could hurt the team’s goals. Both

arguments are important to mention, but justification for your viewpoint needs to be

strengthened.

Methods

7. I would like a little more information about your sample.

a. In your initial 20 interviews, who were these managers and staff? Were they part of the

same team? Different teams? Were they in teams at all? (p. 14)

b. Where was the headquarters (p. 15)?

c. Were the other 25 interviews made with the same teams or different teams (p. 15)?

d. Does the organization no longer exist (“operated”) (p. 15)?

e. More information on the teams themselves would be helpful. What exactly did these

teams do? Did they give loans to countries, to private businesses, to individuals? What

are some characteristics of the team members (e.g. team members at the same

hierarchical level, level of demographic diversity in the team)? (pp. 15-16)

f. Where were the inflows of information coming from? Inside the organization? Outside of

it? (p. 16)

8. How did you analyze the qualitative research? Did you conduct a content analysis? Did someone

else independently verify or confirm your conclusions in a separate analysis? I realize the paper

is already quite long and this is not directly linked to the testing of your hypotheses and hence

not critical to do, but I would enjoy reading a few quotes or examples to clarify and provide

more face validity to some of your statements (e.g. top of p. 17 regarding influence attempts) if

you think this might add to the richness of this paper.

9. More information and clarification on the quantitative side would also be helpful.

a. By “quality evaluations” (p. 17) do you mean teams that rated for project quality by the

project-monitoring unit? This is unclear.

b. What do you mean by “assembled customized panels” (p. 17)? That they had a panel of

experts to assess the team?

c. I assume you mean that you received the quality ratings from the panelists (p. 17)?

d. Did you send your survey to each team to complete as a whole, or to all individuals on a

team, or to certain members of the team (p. 18)?

e. Who were the people used to refine the survey (p. 18)? Team members? Organizational

members? Others?

f. A separate header about “measurement” would be helpful. Also, for measures that were

changed based on refinement (see point 9e or p. 18), be sure to mention how they were

refined (i.e. questions omitted, reworded, etc.).

g. Were the 100 project quality questions from you or from the company (p. 18)?

14

h. Did you develop your measure of team autonomy based on Hackman (1987) or are you

using some measure developed by him or someone else?

i. How did you determine which countries were in which region (p. 22)?

Discussion

10. Your “alternative explanations” (pp. 27-29) would in my opinion fit better in the “discussion”

section where you would delve into the meaning of your findings and limitations. This section

also seems inordinately long, particularly the explanations of how the alternative explanations

were not supported. In my opinion, this section could be omitted or significantly reduced.

11. Practical implications of your study, as well as a more in-depth discussion of limitations and

areas for future research would be good to mention.

Minor Points:

12. On p. 5 sentence three you may want to switch the order of words and say “The combination of

high embeddedness and high autonomy…avoid the risks of excessive external influence or

excessive isolation” as it would seem embeddedness leads to the problem of external influence

and autonomy leads to the problem of isolation.

13. You may want to consider omitting the second paragraph on p. 6 about Hackman’s four sets of

critical decisions, as it seems to suddenly appear in the paper and it was unclear to me how this

relates to your paper. It may make sense to just discuss it later in the methods section (p. 19), as

you did.

14. On p. 12 the first sentence after Hypothesis 2 is a run-on.

15. On p. 13 you say “Similarly excessive isolation poses greater risks for highly complex projects

than for simple ones, since the latter require more sophisticated…” Do you mean the former? It

makes more sense to me that isolation is more problematic for complex projects.

16. Have clearer referents (e.g. “their” and “they” p. 17 as panelists, project teams, or someone else).

17. On p. 24 last sentence of first full paragraph “nor” rather than “not does autonomy…”

1

(3) FIRST REVIEWER RESPONSE

AMJ2007-0880: Responses to Associate Editor and Reviewers

SUMMARY

I really want to thank all four of you for providing an immensely helpful set of comments. They

helped me to make substantial changes to the paper that I hope strengthen it considerably. I have

rewritten the original manuscript from start to finish and conducted new empirical analysis as

detailed in my responses below. To briefly summarize the main changes, I have:

• reframed the paper to focus on the contribution of the embedded autonomy perspective to

theories of team self-management as well as the knowledge-based view of the firm;

• focused on “knowledge embeddedness” rather than embeddedness more generally;

• introduced two new dependent variables, strategic effectiveness and operational efficiency,

in place of the previous dependent variable, project quality;

• reduced the previous five disconnected hypotheses to three hypotheses that share and

advance a common theoretical framework;

• included illustrative examples from my qualitative data;

• provided new methodological details, new supporting analyses, and a new discussion.

As requested in the Associate Editor’s letter, below I respond in detail to the Associate Editor’s

comments (p1-9 of this response) and then much more briefly to the Reviewers’ comments (p10-

23 of this response). For your convenience, I have included your comments in italics prior to my

responses.

RESPONSE TO ASSOCIATE EDITOR

1. Contribution. You need to articulate your contribution and clarify how your study is

theoretically linked to and disconnected from prior work on knowledge and project teams,

particularly the research reported in Haas 2006. As Reviewer 1 points out, “While the prior

research doesn’t develop the specific construct of embedded autonomy, it does look examine the

interaction between knowledge gathering and autonomy or what you call embedded autonomy.

In this way, your H1 has already been tested and supported.” (point 5). You might want to drop

your current H1 and focus on new theoretical insights in your other hypotheses. It is important

to convince your readers that your paper represents a major theoretical breakthrough rather

than just an incremental extension of prior work. I encourage you to identify a theoretical

advance and develop it more clearly and convincingly.

I have rewritten the paper completely from start to finish in order to develop a clear and

convincing theoretical contribution, clarify how my study builds on and extends prior

research, and articulate how it advances our understanding of project teams and knowledge

work. I have also undertaken a completely new set of empirical analyses that uses two new

dependent variables not utilized in previous studies to further distinguish this study from the

research reported in Haas (2006). After much thought, I have retained the embedded

2

autonomy concept as central to the contribution of the paper, but I have reframed H1 and

invested considerable effort in developing new theoretical insights and empirical evidence to

support it. I have also used my new theoretical framing to drive the boundary conditions that

I explore in the new H2 and H3 that follow, creating greater coherence for the paper. I

believe that the comprehensive new theory of embedded autonomy together with the two

new dependent variables provide a major advance beyond previous work, both theoretically

and empirically.

i) The primary theoretical contribution that I now offer is a team-centered theory of

knowledge work that advances theories of team self-management by proposing that

embedded autonomy, rather than autonomy alone, enables project teams to perform more

effectively. Theories of team self-management have proposed many benefits of team

autonomy, yet the empirical evidence for these benefits is weak, especially for project

teams. To explain this puzzle, I take an external perspective on project teams that

suggests that autonomy can lead teams to become isolated from their environments,

resulting in missed opportunities for learning that can reduce their effectiveness.

However, autonomous teams can counteract this risk through knowledge embeddedness –

obtaining knowledge from sources outside the team.

Thus, my aim is to contribute to theories of team self-management by using an external

perspective to help explain when and why team autonomy is not as beneficial as expected

for project teams. I also contribute to the knowledge-based view of the firm by focusing

on the project teams that carry out much of the knowledge work in organizations today,

in contrast to prior research that focuses on knowledge flows to business units or to

national subsidiaries. My focus on project teams helps to illuminate why knowledge

flows do not always deliver the strategic advantages proposed by the knowledge-based

view, and increases the capacity of the knowledge-based view to explain variance in

performance outcomes within and across firms. Finally, I contribute to research on

multinational management by viewing the macro-level challenge of combining

differentiation (autonomy) and integration (knowledge flows) from the micro-level

perspective of the project teams that face this challenge in their everyday work, and

identifying the micro-level conditions that help them to meet this challenge.

ii) In the revised paper, I have replaced the previous empirical analyses with completely new

analyses that use two different dependent variables. In the previous version of this

manuscript, my dependent variable was project quality, the same performance metric as

in Haas (2006). In this revised version, I examine the effects of embedded autonomy on

(1) strategic effectiveness (2) operational efficiency. These two new dependent variables

are significantly but not highly correlated with each other (r=0.39, p<0.01), enabling me

to test the embedded autonomy hypotheses on two substantively different measures of

project performance. In the revised discussion section (p32), I note that Haas (2006)

found a similar interaction effect for project quality, a performance metric that is different

from strategic effectiveness and operational efficiency. While that study did not utilize

the concept of embedded autonomy or identify its boundary conditions, this previous

finding taken together with the findings of the present study indicates that the embedded

autonomy effect has some generalizability across multiple performance metrics.

3

You also need to provide solid empirical evidence to support your proposed contribution. Given

that the different aspects of team embeddedness in your study are very highly correlated with one

another (as reported in your table 1), it is not clear what we actually gain by separating these

aspects and testing their relative difference in interacting with autonomy in different hypotheses.

If you want to identify the boundary conditions that split team embeddedness into different

aspects, you need to make sure that you can clearly differentiate these aspects empirically. In

your revision, you must present a stronger case that your study provides a more compelling body

of learning for both management theory and knowledge.

I have addressed the concern about the different aspects of knowledge embeddedness

(formerly team embeddedness) in two ways: first, by recognizing their common theoretical

underpinnings in the theory and combining the predictions into a single hypothesis, and

second, by demonstrating their empirical distinctiveness more rigorously.

i) I have combined the two old hypotheses that separated out the different aspects of

knowledge embeddedness into a single new hypothesis, H2. This new hypothesis

provides a common theoretical rationale for comparing the effects of the different aspects

of knowledge embeddedness – that teams should benefit more from knowledge that is

non-redundant than from knowledge that is more redundant with what they already know.

I provide two operationalizations of non-redundant versus more redundant knowledge:

country versus technical knowledge (see p22) and extrafirm versus intrafirm knowledge

(see p23). For more on this hypothesis, please see my responses in point 2 ii) and 4 iii)

below.

ii) To clearly differentiate the different aspects of knowledge embeddedness empirically, I

conducted confirmatory analysis as well as multi-trait multi-method analysis on the two

pairs of embeddedness sub-constructs (Venkatraman & Grant, 1986). I also now report

Cronbach’s alphas, intra-class correlations, and rwg for these sub-constructs. While the

correlations between the paired aspects of knowledge embeddedness are fairly high for

country/technical knowledge (r=0.80) and moderate for intrafirm/extrafirm knowledge

(r=0.52), the results of these analyses indicate that the pairs of sub-constructs have both

convergent and discriminant validity, indicating that they can be meaningfully

differentiated empirically. I report these results in the methods section (p23, p24).

2. Theory. Reviewer 2 questions your motivation to choose the four boundary conditions for

studying the value of embedded autonomy. This reviewer asks “Why these four and not others?”

(point 7). Indeed, Reviewer 1 suggests a different boundary condition for you to consider (point

6). You need to convince your readers that you have a strong, coherent theory that guides you to

select your variables. I suggest you put more efforts in developing a good theory for the

boundary conditions for embedded autonomy, and use the theory to justify why particular

boundary conditions are included in your study.

As described above, in the revised paper I have simplified the theory considerably to make it

stronger and more coherent. I have also identified two boundary conditions that emerge from

it clearly and directly, in place of the previous four boundary conditions.

4

i) As outlined above (see my response to AE comment 1), the revised theory argues that

autonomous teams have independence but are in danger of isolation that can impede their

performance, but knowledge embeddedness - obtaining knowledge from sources outside

the team – reduces this danger. This simplified argument leads directly to the following

implication: the benefits of knowledge embeddedness for autonomous teams will depend

on the team’s need for outside knowledge. I identify two conditions under which the

team’s need for such knowledge is greater: (i) when the knowledge is different from their

own rather than similar to their own, and (ii) when their project is more novel or

complex. From this, I derive two boundary conditions. First, because teams’ need for

outside knowledge is greater if that knowledge is different from their own, autonomous

teams will benefit more from embeddedness when the knowledge obtained is non-

redundant rather than redundant with their own (H2). Second, because teams’ need for

outside knowledge is greater if the projects is more novel or complex, autonomous teams

will benefit more from embeddedness when the knowledge is obtained for highly novel

or complex projects (H3). Thus, the benefits of knowledge embeddedness for

autonomous teams will be greater when characteristics of the knowledge itself or

characteristics of the project make obtaining that knowledge more valuable. The revised

paper therefore offers two boundary conditions that are clearly grounded in the theory

rather than the previous four ungrounded ones.

ii) Please note that in the previous version of the paper, I argued that autonomy offers

independence but also isolation, while knowledge offers information but also influence.

However, I have dropped all the theoretical arguments about influence in the revised

version. While I find the notion that knowledge is a source of influence very compelling,

I decided to drop it for three reasons: first, because I believe that the theory is more

powerful and parsimonious without it; second, because I am concerned about a possible

disconnect between theory and data; and third, because none of the reviewers seemed to

view it is a major contribution of the paper. One implication of this decision is that the

boundary condition raised by R1, the presence of influence attempts, is no longer so

relevant.

3. Embeddedness. All three reviewers question your definition of embeddedness. Reviewer 3

is concerned about the way you define embeddedness in terms of both internal and external

factors, and urges you to provide clearer explanations (points 1 and 2). Reviewer 2 has a

similar concern, and asks you to clarify in what level of environment a team is embedded. In

addition, Reviewer 1 questions whether you really study embeddedness (point 4) given your

emphasis on knowledge flow in operationalizing embeddedness. As this reviewer points out,

“Knowledge flows don’t tell us the extent to which the teams actually pay attention to and

use the knowledge that flows to them” (point 4). Like these reviewers, I am not convinced

that embeddedness is the right term for your current study unless you change your definition

and improve your measures. Alternatively, you may reconceptualize your idea of knowledge

flows and use a different term to replace embeddedness in your study.

I thought long and hard about whether to replace the term “embeddedness” with a different

term. Ultimately, I believe that the concept of “embedded autonomy” really captures the idea

that I wish to convey in a succinct way that makes the contribution of the paper memorable.

5

For this reason, I decided to retain “embedded autonomy” as the core concept of the paper.

However, I have focused my theoretical arguments about embeddedness on the more specific

idea of “knowledge embeddedness.” While this decision is risky, at this point I still believe

that the payoff is worth it.

That said, I am very open to feedback here. The best alternative to “embedded autonomy”

that I could come up with is “connected autonomy”, which I think captures the right idea

without perhaps invoking other literatures. If you and the reviewers prefer this wording, I

will be happy to make changes accordingly.

Given that I am retaining “embedded autonomy” in this revision, here is how I have tried to

address some concerns that have arisen or might arise for the reviewers:

i) The term “embeddedness” has been used in many different ways in the organizational

literature. In the revised paper, I have aimed to recognize this variation and explain

where my use of the term fits without undertaking an extensive review of the

embeddedness literature, which would be beyond the scope of the paper. I have noted

that the term generally refers to the strength of the relations between organizational

actors and their environments and that those with stronger relations are viewed as

more embedded. For this study, in which the organizational actors are project teams

engaged in knowledge work, I focus specifically on “knowledge embeddedness,”

which refers to the strength of the knowledge-sharing relations between teams and

their environments (p4 and p10).

ii) The term “embeddedness” needs to be clearly defined for the purposes of my study.

• In the revised introduction (p5), I have conceptually defined teams as “more

embedded in their knowledge environments if their members obtain more

task-relevant knowledge from sources outside the team, including sources

within as well as beyond the organization.”

• In the section on the role of knowledge embeddedness (p11), I have provided

a fuller operational definition, as follows:

“Operationally, I define knowledge embeddedness as the total amount of

knowledge that a team obtains from sources outside the team. These

sources can be situated both within and beyond the organization. Sources

within the organization include directories of in-house experts who can

provide advice on specialized issues as well as company databases or

libraries. Sources beyond the organization include personal contacts in

other organizations who can offer valuable advice and information,

professional associations and conferences, and government websites that

provide access to public data. A team’s level of knowledge embeddedness

is a group-level aggregation of individual behaviors: in a typical project

team, some members obtain more outside knowledge than others, and

members obtain knowledge from different sources.”

• I have used more consistent and careful terminology throughout to addresses

R3’s questions (points 1 and 2) about what I mean by inputs from within and

6

beyond the organization (p11), as well as R2’s request (point 2) for further

clarification of the environments in which teams are embedded (p11).

iii) The term “embeddedness” may create confusion about whether the teams actually

pay attention to and use the knowledge that flows to them. Reviewer 2 (point 4) noted

that “it seems like in order for knowledge flows to create embeddedness is when the

team is requesting the information but not necessarily when the information comes

from the outside in a way that is forced or unwelcomed.” I agree with this point. I

think that the concern here arose because my previous arguments about knowledge as

a source of influence suggested that outsiders might provide knowledge to the team in

order to influence them. In that situation, knowledge flows to the team would not

necessarily be attended to and used. In the revised paper, I focus exclusively on

teams’ own efforts to obtain knowledge from outsiders, and make no arguments about

knowledge as a source of influence. This focus is consistent with the emphasis of the

survey questions which asked the team members “how much relevant knowledge did

you get from…” While I agree that the survey responses do not allow me to establish

with absolute certainty whether teams actually pay attention to and use the knowledge

they obtain, I believe that the wording of the questions made them much more likely

to recall knowledge that they had found useful (relevant) than knowledge that they

did not use. I have noted this in the description of the measure (p22).

4. Conceptual Development. The reviewers also raise concerns about the logic and premises

underlying your predictions. Some of the arguments leading to your hypotheses are not very

convincing. You need to clarify your logic and strengthen your arguments. In particular,

Reviewer 3 suggests you rework your arguments for H3 (point 6). Reviewer 2 is concerned

about team usage of knowledge in your arguments, and asks you to pay attention to the

quality/relevance of available knowledge (point 4). Reviewer 1 believes that some of your

hypotheses are intertwined (points 6 and 9) and asks you to clarify your logic. This reviewer

also suggests you consider within team dynamics and power in your theory development (point

18). Considered collectively, the theory section of your paper is not developed clearly or

sufficiently enough to provide the basis for your hypothesized relationships. In your revision, the

theoretical and/or conceptual basis for your hypotheses must be substantially strengthened.

Beyond strengthening the theoretical and conceptual basis for my hypotheses as already

outlined,

i) I have reworked my arguments about the benefits of knowledge from beyond the firm

relative to knowledge from within the firm (old H3, now part of H2), as requested by

Reviewer 3 (point 4) (p13).

ii) I agree with Reviewer 2’s observation that an embedded team may be able to discern

more accurately the potential quality/relevance of knowledge. This is consistent with my

argument that teams that are more embedded can benefit from increased insights and

decreased duplication of past mistakes or existing solutions more than teams that are less

embedded – the former not only obtain more knowledge, but they may also be more able

to identify the potential quality/relevance of that knowledge, and utilize it accordingly. Of

7

course, the opposite may also occur – some teams may choose to obtain less knowledge

in total because they want only knowledge of high quality/relevance. However, my

survey explicitly asked the team members about how much relevant knowledge they

obtained from the various sources, so the requirement that the knowledge is of high

quality/relevance it is integral to my empirical as well as my theoretical definition of

knowledge embeddedness. Because this is an important point to recognize, I have now

made it explicit in the paper (p22).

iii) I agree with Reviewer 1 (points 6 and 9) that some of my original hypotheses were

intertwined. I found this observation very helpful, as it highlighted the common

theoretical roots of the intertwined hypotheses. I have now incorporated these insights

into the paper by combining these pairs of intertwined hypotheses into the new H2

(combining old H2/H3) and new H3 (combining old H4/H5), and clarifying the

underlying logic for them as relating to characteristics of the knowledge (H2) and

characteristics of the projects (H3) that increase the benefits of knowledge embeddedness

for autonomous teams.

iv) While this paper takes an external perspective on project teams, I now discuss the

research on within-team dynamics in the theoretical framing (p6-8) as well as addressing

some possible implications of within-team dynamics in the discussion section, as

Reviewer 1 (point 18) suggested (p38).

5. Qualitative data. You mention your interviews and qualitative research many times in your

paper. But, incorporate such data in your study. Consider using qualitative data (such as quotes

or examples) to clarify some of your statements. you provide little content about your qualitative

data in the paper. All three reviewers have questions about your qualitative data (Reviewer 1,

point 10; Reviewer 2, points 12-14 and 17; Reviewer 3, point 8). You need to provide more

detailed information about your qualitative data in terms of how you collect, analyze, and

incorporate such data in your study. Consider using qualitative data (such as quotes or

examples) to clarify some of your statements.

i) I have followed the suggestions that I enrich the paper by including more of the insights

from the qualitative data (AE, point 5; R1, point 10; R3, point 8). I now use examples

from the teams I studied to illustrate three points (see p16-19): (1) variation in team

autonomy – the West African slum upgrading team versus the East-Central European

education team, (2) the problems of isolation in self-managing teams – the example of the

Latin American urban infrastructure team, (3) the benefits of knowledge embeddedness –

the West African team again and the Romanian social services team. I do not include

quotes mostly because I believe that the examples illustrate my points effectively but also

because these already add three pages to the length of the paper.

ii) I have also more carefully explained that my interviews and observations were intended

to help me understand the teams and projects in the organization I studied, as well as to

develop the survey instrument. My qualitative data collection efforts thus were

supplementary to the quantitative data collection, rather than a fully-fledged stand-alone

qualitative study. In response to specific points raised by yourself and the reviewers (AE,

point 5; R2, points 12-14, 17; R3, point 8), I have explained how and why I selected the

8

interviewees, when I conducted the interviews, the materials I collected, and how I

analyzed them (p15).

6. Methodological issues. The reviewers also identify some methodological issues in your paper.

In particular, they have questions about your sample (Reviewer 2, point 18; Reviewer 3, point 7),

survey (Reviewer 3, point 9), variation in dependent variable (Reviewer 2, point 33), validity of

your measures (Reviewer 1, points 13-15), inter-rater reliability (Reviewer 1, point 16), and

alternative explanations for your study (Reviewer 1, points 11, 12, 17). You need to clearly

address all these concerns in your revision.

i) To address the questions about my sample, I have provided more descriptive details about

the teams and their projects, including who the team members were, their demographic

characteristics, how they worked on multiple projects at a time but rarely on more than

one project together, how the teams were constituted, and the nature of the projects on

which they worked (p15-16). As requested, I have also clarified the location of the

headquarters and the continuing existence of the organization (p16).

ii) To address the questions about the survey, I have clarified the terminology I use to refer

to the project evaluation process and the project-evaluation unit that managed it, my

description of the panels who conducted the assessments, and how I received the project

evaluations (p19). I have also explained that I sent the survey to all the individual

members of each team, that it was pre-tested with team members who were not part of

the sample, that the project evaluation questions were developed and administered by the

organization rather than myself, and that I developed my autonomy measure by taking the

four general categories of critical team decisions identified by Hackman (1987) and then,

through my interviews, identifying five decisions within each of these categories that

were important in this organizational context (p20-21).

iii) To address the questions about my measures, I have done additional validation work to

demonstrate the inter-rater reliability and convergent and discriminant validity of the

survey items. I now report Cronbach’s alphas, intra-class correlations, and rwg for all the

multi-item constructs (p21, p22, p23, p24, p25). Additionally, I report the results of a

confirmatory factor analysis as well as a multi-trait multi-method analysis to demonstrate

the distinctiveness of the knowledge embeddedness and team autonomy constructs (p27),

as well as of the country/technical knowledge and intrafirm/extrafirm knowledge sub-

constructs (p23, p24).

iv) In response to R3’s question about the variation in my categorical dependent variables, I

conducted two additional analyses:

• First, because there were relatively few projects rated “marginal/unsatisfactory”, I

checked whether these were driving my results by creating alternative

dichotomous measures of strategic effectiveness and operational efficiency that

were coded 1 if the project was rated “highly satisfactory” or 0 if the project was

rated either “satisfactory” or “marginal/unsatisfactory”. The results using these

alternative dichotomous measures were the same as the results for the original

categorical measures for H1, and substantively similar though weaker for H2 and

H3.

9

• Second, because variation in the dependent variable was a particular concern for

the categorical measure of strategic effectiveness, I created an alternative

continuous measure of strategic effectiveness by summing the project’s scores on

the 10 questions that the expert panels used to inform their overall strategic

effectiveness ratings. This continuous measure is highly correlated with the

categorical measure of strategic effectiveness (r=0.88), but not identical to it

because the expert panels took into account their full understanding of the project

and its distinctive challenges as well as the scores on the 10 questions when

determining the overall strategic effectiveness rating for a project. However, this

continuous measure does allow for more fine-grained variation in the strategic

effectiveness dependent variable. The results using this alternative continuous

measure were same as results using the original categorical measure for H1, H2,

and H3.

• Having established that the results are robust to these alternative specifications of

the dependent variables, I retain and report the original categorical measures in

this study because the project-evaluation unit that generated the project ratings

viewed the three categories as meaningfully different from each other and the

categorical ratings as capturing the assessment of the project more accurately than

the continuous ratings. I report this in a footnote (p21, footnote 2).

v) In response to R1’s concern about the reliability of single-item measures of project

novelty and project complexity, I have now revised these measures by replacing the

original project novelty measure with a more robust 2-item measure (p25) and the

original project complexity measure with a more robust 3-item measure (p25). I have

provided more details about these new measures and their benefits in my responses to this

reviewer, including noting that the differences in response rates for teams based on their

cost and length do not confound my results using this revised complexity measure (R1

response, points 12 and 14).

vi) In response to the same reviewer’s questions about alternative explanations, again please

see the detailed responses to this reviewer (R1 response, points 11 and 12). In the revised

paper I address the reviewer’s question of possible differences across the two types of

project teams by explaining that I did not expect to see differences in the value of

embedded autonomy since the two types of teams were substantively similar for the

purposes of this study (p15), and I report the results of additional analyses showing that

the results did in fact hold across both types of teams (p32). This reviewer noted that s/he

found my discussion of possible alternative explanations for my findings helpful, so I

have retained this section but I have now moved it to the discussion section as suggested

by R2 and R3 (p31-33).

10

RESPONSE TO REVIEWER 1

This paper develops the idea of embedded autonomy in project teams in multinational

organizations. Embedded autonomy is a very interesting construct having to do with being

relatively independent in decision making but embedded in their context through knowledge

flows to ensure good decision making. The paper draws on a wonderful dataset of a

relatively large sample of project teams where real project quality is assessed. Overall, I

think this research has real promise. That said, I have some suggestions that I will outline in

my points below that will help improve the contribution of your work and address some of the

current limitations. Best of luck to you as you continue to develop this line of research.

Thank you! I found your comments incredibly helpful – many thanks for your thoughtful and

thought-provoking review.

1. I am glad to see you embed your work in team embeddedness theories. However, I think

your two main sub-elements of team autonomy and information flows are very consistent

with models of high involvement coming out of the research of the Center for Effective

Organizations at the University of Southern California, particularly the work of Ed Lawler

(he as a number of books including Ultimate Advantage which may give you an overview).

He describes effective high involvement as being a combination of knowledge, power,

autonomy and rewards. As you can see your embedded autonomy is a combination of two

elements of his model. I think it’s important to situate your work within the research on high

involvement/team design research.

Thanks for the suggestion. I have now referenced this work in the theory section as well as the

discussion section. Please see p7 and p35 in the paper.

2. I didn’t find your framing of the front end around multinationals to be that compelling. It

made me expect that you were going to be examining subsidiaries of multinational

corporations in your paper. While the analogy may be appropriate in your discussion, I

don’t think it’s the strongest introduction to your ideas. You are dealing with project teams.

Why not just go immediately to the issue at hand which you begin to introduce at the bottom

of p. 3.

I have followed this suggestion. Please see p3 in the paper.

3. I also found that during the times you referenced the multinational management literature

that it seemed to mix up the level of analysis. For example, on p. 7, you cite concerns about

national subsidiary autonomy from the parent organization. However, your theory is about

project teams. It gets a bit confusing when you so easily cross levels of analysis in your

conceptual development. Might not there some different issues at hand depending on what

level you are referring to?

Good point. In response to your previous comment above, I now spend less time on the

multinational management literature, and I have revised the text to reduce confusion about my

level of analysis. Please see p3-5 and p36 in the paper.

11

4. I also wonder whether your definition of embeddedness (i.e., knowledge flows from the

environment to the team) is really embeddedness. Knowledge flows sound more like

embedding forces rather than embeddedness. Knowledge flows don’t tell us the extent to

which the teams actually pay attention to and use the knowledge that flows to them. At the

very least, it seems like in order for knowledge flows to create embeddedness is when the

team is requesting the information but not necessarily when the information comes from the

outside in a way that is forced or unwelcomed. I think the argument is stronger to the extent

that you can show that teams obtain and use the knowledge that is provided to them.

I agree; I have revised the text to address this concern. Please see my response to the Associate

Editor, point 3 iii), and p22 in the paper.

5. You cite Haas 2006 as having examined the same question – whether knowledge gathering

combined with decision-making autonomy is related to project performance. This makes me

wonder about the extent to which your work goes beyond this cited research. When I went

back to look at this piece, I had even more concerns about this issue. To what extent does

your research go beyond the research reported in Haas 2006? While the prior research

doesn’t develop the specific construct of embedded autonomy, it does look examine the

interaction between knowledge gathering and autonomy or what you call embedded

autonomy. In this way, your H1 has already been tested and supported. Thus, your key

contribution is a better understanding of the conditions under which embedded autonomy

will make a difference to project quality. It seems like this is an important issue to highlight

in your framing because it highlights the potential for a more limited contribution.

This is clearly an important issue. Thinking it through at length has me to very substantially

revise the paper, both theoretically and empirically. Please see my response to the Associate

Editor, points 1 i) and 1 ii), and p32 in the paper.

6. One issue you raise that could be an important boundary condition is the presence of

influence attempts. It seems to have come out strongly in your interviews as well. Is this

another relevant boundary condition for embedded autonomy?

Please see my response to the Associate Editor, point 2 ii). As I note there, this boundary

condition is no longer so relevant because I no longer focus on influence attempts in my theory.

7. It strikes me that H2 and H3 are intertwined. Isn’t knowledge from outside the organization

usually more scarce then information within the organization? Also it is important that the

knowledge come from outside the organization? What about from inside the organizations

but outside of the team? Would the logic be similar for this boundary condition?

This was a very helpful observation. I have now combined these two hypotheses into a new H2.

Please see p12-13 in the paper, and my response to the Associate Editor, point 2 i) and point 4

iii).

12

8. I wonder about your point that projects are more prone to isolation when they are

demanding or complex. Wouldn’t that be particular the time when the organization would

want to put more knowledge in the team’s hands with more outside advisors and such?

This is a fair point; I have revised my arguments to suggest that teams can benefit more from

knowledge from sources outside the team when their projects are more novel or complex. Please

see p13-14 in the paper.

9. Similar to my point #6 above, it strikes me that H4 and H5 are also intertwined. Won’t more

complex projects by nature be more highly prominent in the sense that they probably involve

more stakeholders?

Again, this observation was spot on. I have now combined these two hypotheses into a new H3.

Please see p13-14 in the paper, and my response to the Associate Editor, point 2 i) and point 4

iii).

10. I like the idea that you have a multi-method study consisting of interview data as well as the

survey data! Bravo. However, I was disappointed to see that you provide almost no

information from your interview data. While I can understand that a full-blown study 1

based on your qualitative data would make your paper very long, in the present form, the

information provided on your interviews provides little content. You tell us about all the

interviews you did but little in the way of what you heard/found. I would at least hope to see

some quotes in your discussion that elaborate or support your quantitative findings. It seems

like that would bring to life your findings. It also seems like your interview data could help

provide information about the kinds of measures you chose and why they were appropriate.

I have now included more information from the qualitative data in the text, including several

examples that illustrate the roles of team autonomy and knowledge embeddedness in the

organization I studied. Please see p16-19 in the paper, and my response to the Associate Editor,

point 5 i).

11. Are there any differences you might expect across the two types of project teams? Would the

nature of these projects make a difference in terms of the need for embedded autonomy? Do

stakeholders play a bigger role in the financial projects, for example?

While the two types of teams were engaged in projects that were in some ways quite different,

both stood to benefit more from decision-making autonomy if they also obtained more

knowledge from sources outside the team, since such knowledge could offer important new

insights and help them to avoid duplicating prior mistakes and existing solutions. Based on my

theoretical model, I therefore did not expect the nature of the projects to make a difference in

their need for embedded autonomy. To examine this empirically, I conducted additional analyses

for the two types of teams separately. The findings show that embedded autonomy helped teams

to improve the strategic effectiveness of both financial and technical projects. It also helped them

to improve the operational efficiency of technical projects, but not significantly for financial

projects, although the effect is in the right direction. Please see p15-16 in the paper for more

extensive discussion of the two types of teams, p32 in the paper for the results of the additional

analysis, and my response to the Associate Editor, point vi).

13

While your specific query about whether stakeholders might play a bigger role in financial

projects is an interesting one (I’d say, based on my interviews and observations at the

organization, that the answer isn’t obvious since the organization is highly politicized internally),

now that I have moved away from the politics/influence framing it seems no longer so relevant to

address this directly in my theorizing.

12. Given there are differences in response rate for teams based on their cost and length, is this

confounding your hypotheses about complexity?

The original single-item complexity measure to which you refer was significantly correlated with

project cost (r=0.33, p<0.01) and with project length (r=0.18, p<0.10), so it is possible that there

was some confounding occurring with this measure. In the revised paper, however, I have

replaced the single-item complexity measure with a new more robust 3-item measure of project

complexity. The correlations between this new measure and project cost/project length are low

(r=0.26, p<0.05; r=0.08, n.s.). These low correlations generally suggest that the differing

response rates are unlikely to confound the hypotheses about project complexity substantially.

More specifically, the implication of team members being less likely to return their surveys if

they worked on costlier or longer projects is that some costlier or longer projects may be

excluded from the dataset. In the case of longer projects, since the correlation of project length

with project complexity is low, this bias is very unlikely to affect the project complexity

analyses. In the case of costlier projects, this bias should make it more difficult to observe

differences between high and low complexity projects using the median split approach, since the

median level of complexity in the sample of responding team members will be lower than the

median in the sample of team members surveyed, blurring the boundary between high and low

complexity projects. Nevertheless, as Table 5 (Models 7a/b) indicate, I find evidence for

different effects of embedded autonomy in high complexity versus low complexity projects,

despite any increased difficulty of detecting these differences due to response rate bias; this

suggests that these differences are robust to small variations in the median.

13. Did you compute a confirmatory factor analysis to show the convergence and discrimination

of your construct items across embeddedness and autonomy? Are the two constructs distinct?

Also, what is the reliability of your embeddedness scale? For the multitrait multimethod

analysis for your knowledge scales, why not do this with confirmatory factor analysis? If

your sample is too small, it’s possible to do these measurement analyses at the individual

level of analysis.

I have now reported the results of confirmatory factor analyses as well as MMMT analyses, as

requested. Please see p23, p24, and p27 in the paper, and my response to the Associate Editor,

point 1 iv) and point 6 iii).

14. I do worry a bit about the reliability of your single item measures of project novelty and

complexity. Have these items been used anywhere else? I wonder if the length of the project

(duration) could be a proxy for complexity?

14

I now use multi-item measures for these constructs. Please see p25 in the paper, and my response

to the Associate Editor, point 4 v).

The correlation between this new measure of project complexity and project duration is low and

not significant (r=0.08, n.s.), indicating that the length of the project is not a proxy for

complexity. In case you were wondering, the correlation between the original single-item project

complexity measure and project duration was also low (r=0.18, p<0.10); removing this particular

item from the new 3-item scale does not change the results.

15. I do like how you use archival measures for project quality, cost, country wealth

Thank you; in the revised paper, I use archival measures for the two new dependent variables.

16. Did you see how much agreement there was for each of your measures across the multiple

respondents? What are the rwg scores?

I now report rwg scores for the relevant constructs. Please see p21-26 in the paper, and my

response to the Associate Editor, point 6 iii).

17. I also like how you address a number of alternative explanations that I also had been

considering.

Thank you; I have now moved this section into the discussion, following the advice of the other

reviewers. Please see p31-33 in the paper.

18. I really like in your discussion that you think of this paper as input to building a team-

centered theory of knowledge work in multinational organizations. I would like to see you

further flesh out this theory in your discussion. Clearly an understand of boundary

conditions is important to the building of that theory. I also wonder how things like within

team dynamics and power issues would play out in this theory. Clearly issues for future

conceptual and empirical work.

Thank you; I have now made this theory more central to my contribution in the introduction as

well as the discussion section. The revised discussion section now considers the applicability of

my theory to other performance measures, other types of teams engaged in other types of work,

and other organizational settings, as well as the possible implications of within-team dynamics

that were beyond the scope of the present study. Please see p32, p37, and p38 in the paper.

Best of luck as you continue to develop these ideas.

Again, many thanks for all your help with improving the paper!

15

RESPONSE TO REVIEWER 2

Overall, your study is thoughtfully designed, carried out with notable and consistent care, and

potentially makes a genuine contribution. Thus, I congratulate you on work well done in pursuit

of understanding an important phenomenon. To help realize the potential of this work, I believe

you can strengthen as well as enrich your argument in multiple ways, which I outline below.

Most of my concerns focus on apparent gaps or disconnects in your argument – calling for needs

to clarify theoretical connections, logic, descriptions of your sample, procedures, and analyses,

and your interpretations of results. Doing so should help to illuminate your contribution and

enable other researchers to replicate your study if so desired, particularly since you sample only

one organization. Unlike most of my reviews where I can point to one, two or three critical

issues, I instead point to the logical flow of your argument and the underlying premises that

frame its identity.

Thank you very much for all your detailed and careful comments, I found them very helpful in

revising the paper!

1. None of your appendices seem necessary as appendices. As it, they detract attention from the

flow of your argument. Personally, I would include in your text what you believe is essential

and delete the appendices.

Thank you for this suggestion. I have incorporate several of the original footnotes into the text,

as you recommended, but I have continued to footnote some points that are important but

difficult to incorporate into the flow of the text without creating a lengthy digression.

2. When you speak of team embeddedness you refer “to the extent of the relations between

teams and their environments.” This statement warrants clearer specification of the level(s)

of “environment” that you have in mind. A team’s environment may be the division in which

it works. Since the concept of “team embeddedness” is central to your study, this should be

addressed more clearly at the outset. In what level of their environment must they be

embedded? Can they be embedded less or more deeply across different levels?

I have now focused more specifically on “knowledge embeddedness” and clarified these issues.

In the introduction, I explain that I view teams as more embedded in their knowledge

environments if they obtain more task-relevant knowledge from sources outside the team,

including sources within as well as beyond the organization. Thus, a team’s environment

includes the internal environment within the organization as well as the external environment

beyond the organization. In the section on the role of knowledge embeddedness, I provide more

detail including examples of these sources. I do not differentiate between “levels” of the

environment, although this would certainly be an alternative way to conceptualize

embeddedness. Please see p5 and p11 in the paper, and my response to the Associate Editor,

point 3 ii).

3. When you talk about the two “double-edged” swords, three of the “edges” describe impacts

on a team: independence, isolation, and information. The fourth, influence, seems to have an

opposite effect. Is not “loss of influence” more consistent with your framing of the other

three?

16

This is a fair point. However, since I no longer argue that knowledge is a source of influence, it

is not as relevant for the revised paper. Please see my response to the Associate Editor, point 2

iii).

4. One of my most salient concerns pertains to team usage of knowledge. By the time I get fully

immersed in your argument, it seems as if there is too little attention paid to the

quality/relevance of available knowledge, as if all knowledge is equal. Yet, an embedded

team may be able to discern more accurately the potential quality of knowledge.

This is a good point. I have addressed it by paying more attention to the relevance issue

throughout the text. Specifically, I have highlighted the importance of obtaining relevant

knowledge, not just any knowledge, in the theory and hypotheses, and I have also noted that the

empirical measures of knowledge embeddedness focuses on how much relevant knowledge the

teams obtain. Please see p22 in the paper, and my response to the Associate Editor, point 4 ii).

5. I do not think Figure 1 adds sufficient value to the paragraph that precedes it – it seems too

obvious to warrant additional space.

I have now omitted the figure.

6. On page 11, you claim that “embedded autonomy can be expected to be most valuable under

conditions where the information benefits provided by embeddedness and the independence

benefits provided by autonomy are particularly high.” As opposed to saying that these

“benefits are high,” I wonder if there is more meaning inherent in the thought that such

“benefits are more relevant to or important for a decision situation at hand.”

I have revised the text to be more precise, as suggested.

7. It may strengthen your argument to explain your reason(s) for choosing the four “boundary

conditions” that you describe on page 11. Why these four and not others?

Again, this was a very helpful comment. I have now proposed two boundary conditions that are

firmly grounded in my theoretical framework, in place of the four less well-grounded ones. For

more detail, please see p12-14 of the paper, and my response to the Associate Editor, point 2 i).

8. This concern exceeds the focus of your study, but I’ll share it for later if not for now. In your

specification of hypotheses 2-5, it may refine your argument to begin with a specification of

“kinds of situations or challenges” at hand -- a situational variable that helps to guide or

inform different arrays of subsequent variables. For example, the extent of value of

embedded autonomy if a knowledge source is outside or inside the organization likely

depends upon the knowledge “needed” for a challenge at hand.

Thank you for this insightful suggestion. I have now made it central to my hypotheses.

Specifically, the boundary conditions I identify are situations in which teams are more likely to

need the knowledge available from sources outside the team: when that knowledge is non-

17

redundant and the project is novel or complex. Please see p12-14 in the paper, and my response

to the Associate Editor, point 2 i).

9. On page 13 you say that “the risk of excessive isolation is greater when projects are highly

demanding (i.e., very novel or complex).” Why don’t you then say this more directly:

“...greater when projects are highly novel or complex”?

I have revised the text as suggested. Please see p14 in the paper.

10. On page 13, in the fifth line above hypothesis 4, I think you mean “former” where you say “latter.”

Thank you.

11. Where possible, minimize the use of passive voice, such as in sentence one of “research

setting” (page 14), where you say “the hypotheses were tested.”

I have revised the text as suggested.

12. At the beginning of your “data and methods” section, you say “I started with 20 interviews.”

How did you make these selections?

I have clarified this. Please see p15 in the paper.

13. On page 15, you mention conducting 18 additional “interviews with leaders and members of

project teams.” When did these occur? It may help to include a timeline that situates and

connects your procedures and data gathering steps.

I have clarified this. Please see p15 in the paper.

14. You also say on page 15 that you “conducted a further 25 interviews as part of detailed case

studies of seven teams.” How did you divide up 25 interviews among seven teams?

I have clarified this. Please see p15 in the paper.

15. When you say that you “read materials that were generated as these teams worked,” to what

materials are you referring? Examples may help.

I have clarified this. Please see p15 in the paper.

16. On page 15 you talk as if the organization you studied no longer exists: “the organization

operated worldwide and supported offices in 100 countries.” Did the organization cease to

exist or are you referring to the time period when you were there.

The organization still exists; I have amended the language accordingly. Please see p14 in the

paper.

17. It is unclear why you conducted seven interviews in Russia. What were you expecting to learn

and how did you integrate these into your study?

I have clarified this. Please see p15 in the paper.

18

18. I would like to know more about the U.S. headquarters where most of your teams operate(d).

For example, what is this work environment like – are members of teams interacting with

each other or are they separated? Are these teams mostly ad hoc? Are individual member

demographics, such as age and education, relevant to team differences? I ask this final

question partly because you note on page 16 that “each team member was responsible for

researching and designing a specific component of the project.”

I have provided more details as requested. Please see p16 in the paper.

19. On page 16 – middle of the page – you write “informal as well as informal.” I suspect that

you mean “formal as well as informal.”

Thank you.

20. I would like to see clearer descriptions of your qualitative data gathering procedures – so

that others could replicate your study if they wanted to do so.

I have clarified this. Please see p15 in the paper.

21. On the top of page 18 you say “to collect data from the members of these teams, I sent them

a customized survey.” Does this mean that you sent a customized survey to each “member”

or to each “team”? And what do you mean by a “customized” survey?

I have clarified this. Please see p20 in the paper.

22. On page 18 you say “96 of the 120 teams qualified for the study by returning at least 50% of

their core team members’ surveys.” How did you decide upon this cutoff point? What do you

mean by “core” team members?

I have clarified this. Please see p20 in the paper.

23. To what degree are “financial” and “technical” projects comparable? At times you seem to

refer to them collectively and at other times separately.

I have now described the two types of teams more fully. Please see p15-16 in the paper. I have

also revised the text to make my references more clear and consistent. Additionally, I have

reported the results of an additional analysis where I compared the two types of projects. Please

see p32 in the paper.

24. On page 18 you say that “each project was evaluated by a different independent panel.”

Describe in more detail what you mean by this. For example, did every single one of the 96

projects have a completely different evaluative panel?

I have clarified this. Please see p19 in the paper.

25. Explain your rationale for multiplying “overall embeddedness scores by the team’s overall

autonomy score” to test hypothesis one.

This is necessary to test the proposed interaction effect; I have revised the text to make this

clearer. Please see p22 in the paper.

19

26. On page 23 you talk about assessing “the extent to which the team was a clearly identifiable

interdependent group.” Explain what you mean by this.

I have now explained this. Please see p25 in the paper.

27. Why did you differentiate “late respondents” from other respondents and how did you

measure “lateness”?

I have now explained this; please see p26 in the paper.

28. On page 23 you talk about “the average length of organizational tenure in years at the start

of the project.” Were you measuring tenure of teams, core members, or both?

I have clarified this. Please see p26 in the paper.

29. I do not understand why “teams that delivered higher quality projects ... had more late

respondents.” You must have anticipated a potential connection here but I don’t see it in

your paper.

This issue no longer arises with the revised dependent variables..

30. At the end of your “results” section you include a subsection entitled “alternative

explanations.” I think this belongs in your discussion.

I have now moved the alternative explanations to the discussion session. Please see p31-33 in the

paper.

31. I think it would enrich your argument to describe “teams” in greater detail.

I have now described the teams more fully; please see p16 in the paper. I have also incorporated

examples from the qualitative data to illustrate the variations in team autonomy and knowledge

embeddedness and their implications for project teams in this organization. Please see p16-19 in

the paper, and my response to the Associate Editor, point 1 i).

32. It seems possible that certain teams had some of the same members as other teams, i.e., that

there were some team members who participated on more than one team. If so, to what extent

did this occur – were some teams substantively similar?

I have now addressed this. Please see p20 in the paper.

33. There is not much variation in your dependent variable. Perhaps you could address this

more carefully in your discussion.

Thank you for raising this important point. I have now addressed it in the methods section.

Please see p21, footnote 2 in the paper, and my response to the Associate Editor, point 6 iv).

34. Because your study focuses so heavily on ONE organization, I think it would enrich your

argument to describe this one organization more intimately and completely.

I have now provided more description of the organization, the teams, and their projects. Please

see p14-19 in the paper.

20

RESPONSE TO REVIEWER 3

Overall, I was impressed with the paper. The study of project teams and knowledge flows in

MNEs is relevant and important, and the author is familiar with the relevant literature.

Furthermore, the author has a relatively large sample of multinational teams as well as both

qualitative and quantitative data to draw upon. Thus, there is much in the paper to like. There

were a few areas that could use improvement, which I discuss below, broadly in terms of theory,

methods, and discussion.

Thank you very much for your careful and constructive comments; they helped me to develop

my ideas and sharpen my writing.

Main Points:

Theory

1. The definition of team embeddedness and autonomy thankfully appear early in the paper (p.

4). However, considering the importance of, in particular, team embeddedness, a little more

clarification seems warranted. You say that that embeddedness refers to inputs inside and

outside the organization. Does this also include team members themselves? In other words,

teams in which group members communicate more with one another is considered part of

team embeddedness? I thought of embeddedness in terms of being embedded in the

environment outside the team, but this isn’t necessarily true given your definition.

Knowledge embeddedness (formerly team embeddedness) refers to environments outside the

team (including the environment within the organization and the environment beyond the

organization). I have clarified this in the text. Please see p4 and p11 in the paper, and my

response to the Associate Editor, point 3 ii).

2. Related to my previous point on definitions, I was a bit confused early-on with the notion of

external influence being a problem. It makes sense to me that outside forces can be

problematic in terms of their influence on a team. However, if you are defining

embeddedness and autonomy in terms of both internal and external factors, this relationship

is less clear. Internal factors (i.e. those within the firm), it would seem, would help the team

gain internal information and stay on track in terms of organizational goals. If you assume

(as I do, and as you mention in the last paragraph of p. 7) that the team is acting on behalf of

the organization, it would seem that autonomy from outside forces is more potentially

problematic than inside forces. This differentiation between internal and external becomes

more confusing when you refer to “external influence” (e.g. p. 10), as I am not sure whether

by “external” you mean outside the team, outside the organization, or both.

I have now removed all references to external influence from my theory, so this should no longer

cause confusion. Please see my response to the Associate Editor, point 2 ii).

3. On p. 7 you mention the issue of groupthink, which makes sense to mention given your study

of teams, but in my view groupthink is more about interactions and decision-making within

the team, as well as having outside input, than a team’s level of autonomy. It seems that the

21

level of embeddedness (more so than autonomy) is critical, as having information outside the

team would reduce the propensity to have groupthink. Also, internal team-level norms, roles,

and culture would make a difference regardless of level of autonomy and embeddedness. For

example, the team could have a culture of questioning its members.

I agree. I now use the groupthink example to illustrate the problem of isolation, while

recognizing that this was not the sole problem; rather, it exacerbated the effects of dysfunctional

dynamics within the team. Please see p8 in the paper. I also note that internal team relations

more generally are important for self-managing teams. Please see p7-8 in the paper.

4. Figure 1 should be better explained and justified (p. 10). It makes sense, but even if you are

adapting from Evan’s (1995) work, you need to clarify your reasoning more, particularly

since this is the basis for Hypothesis 1.

I have now omitted this figure. I have also clarified my reasoning for Hypothesis 1. Please see

p10-12 in the paper.

5. What do you mean by “most valuable” (p. 11)? Do you mean they will have the best

performance, that they have the potential for the best performance, or something else? The

word “value” is particularly problematic in your hypotheses (pp. 12-14), as it was not

defined, seems vague, and does not sound testable. Also, the level of analysis needs to be

clarified. Is it of most value for the team, the organization, or something else?

I have reworded the hypotheses (and the text where relevant) to avoid the term “value” and

instead replace it with clearer predictions of the expected relationships between my independent

and dependent variables. Please see p12-14 in the paper.

6. You need to rework your arguments for Hypothesis 3 (pp. 12-13). At first, it sounds as though

you are advocating firm outsiders as improving inputs and autonomy. However, then it

sounds as though you are advocating information from insiders since they have similar

identities, interests, etc., and that outsiders may have external influence that could hurt the

team’s goals. Both arguments are important to mention, but justification for your viewpoint

needs to be strengthened.

I have reworked these arguments as suggested. Please see p13 of the paper.

Methods

7. I would like a little more information about your sample.

a. In your initial 20 interviews, who were these managers and staff? Were they part of

the same team? Different teams? Were they in teams at all? (p. 14)

b. Where was the headquarters (p. 15)?

c. Were the other 25 interviews made with the same teams or different teams (p. 15)?

d. Does the organization no longer exist (“operated”) (p. 15)?

e. More information on the teams themselves would be helpful. What exactly did these

teams do? Did they give loans to countries, to private businesses, to individuals?

22

What are some characteristics of the team members (e.g. team members at the same

hierarchical level, level of demographic diversity in the team)? (pp. 15-16)

f. Where were the inflows of information coming from? Inside the organization?

Outside of it? (p. 16)

I have now addressed these questions in the text. Please see p15-16 in the paper.

8. How did you analyze the qualitative research? Did you conduct a content analysis? Did

someone else independently verify or confirm your conclusions in a separate analysis? I

realize the paper is already quite long and this is not directly linked to the testing of your

hypotheses and hence not critical to do, but I would enjoy reading a few quotes or examples

to clarify and provide more face validity to some of your statements (e.g. top of p. 17

regarding influence attempts) if you think this might add to the richness of this paper.

I have now clarified my qualitative methods, and I have also incorporated examples from the

qualitative data, as suggested. This adds three pages to the paper, but hopefully it also adds

richness. Please see p16-19 in the paper.

9. More information and clarification on the quantitative side would also be helpful.

a. By “quality evaluations” (p. 17) do you mean teams that rated for project quality by

the project-monitoring unit? This is unclear.

b. What do you mean by “assembled customized panels” (p. 17)? That they had a panel

of experts to assess the team?

c. I assume you mean that you received the quality ratings from the panelists (p. 17)?

d. Did you send your survey to each team to complete as a whole, or to all individuals

on a team, or to certain members of the team (p. 18)?

e. Who were the people used to refine the survey (p. 18)? Team members?

Organizational members? Others?

f. A separate header about “measurement” would be helpful. Also, for measures that

were changed based on refinement (see point 9e or p. 18), be sure to mention how

they were refined (i.e. questions omitted, reworded, etc.).

g. Were the 100 project quality questions from you or from the company (p. 18)?

h. Did you develop your measure of team autonomy based on Hackman (1987) or are

you using some measure developed by him or someone else?

i. How did you determine which countries were in which region (p. 22)?

I have now addressed these questions in the text. Please see p19-21 in the paper.

Discussion

10. Your “alternative explanations” (pp. 27-29) would in my opinion fit better in the

“discussion” section where you would delve into the meaning of your findings and

limitations. This section also seems inordinately long, particularly the explanations of how

the alternative explanations were not supported. In my opinion, this section could be omitted

or significantly reduced.

23

I have now moved the alternative explanations to the discussion section and reduced it

somewhat, though not substantially since other reviewers found it helpful. Please see p31-33 in

the paper.

11. Practical implications of your study, as well as a more in-depth discussion of limitations and

areas for future research would be good to mention.

I have now discussed implications, limitations and directions for future research more

thoroughly. Please see p34-38 in the paper.

Minor Points:

12. On p. 5 sentence three you may want to switch the order of words and say “The combination

of high embeddedness and high autonomy…avoid the risks of excessive external influence or

excessive isolation” as it would seem embeddedness leads to the problem of external

influence and autonomy leads to the problem of isolation.

Thank you; I have now revised this sentence.

13. You may want to consider omitting the second paragraph on p. 6 about Hackman’s four sets

of critical decisions, as it seems to suddenly appear in the paper and it was unclear to me

how this relates to your paper. It may make sense to just discuss it later in the methods

section (p. 19), as you did.

I have now introduced this in the methods section, as you suggest. Please see p21 in the paper.

14. On p. 12 the first sentence after Hypothesis 2 is a run-on.

Thank you; I have now revised this sentence.

15. On p. 13 you say “Similarly excessive isolation poses greater risks for highly complex

projects than for simple ones, since the latter require more sophisticated…” Do you mean

the former? It makes more sense to me that isolation is more problematic for complex

projects.

Thank you; I have now revised this sentence.

16. Have clearer referents (e.g. “their” and “they” p. 17 as panelists, project teams, or someone

else).

Thank you; I have clarified the referents throughout.

17. On p. 24 last sentence of first full paragraph “nor” rather than “not does autonomy…”

Thank you; I have now revised this sentence.

2

(4) SECOND SUBMISSION

Embedded Autonomy:

A Team-Centered Perspective on Knowledge Work in a Multinational Organization

ABSTRACT

Integrating and extending team self-management and knowledge-based theories, I propose that

project teams perform more successfully when they are characterized by “embedded autonomy”:

control over critical task decisions combined with extensive use of task-relevant knowledge from

sources outside the team. This study examines the effects of embedded autonomy on the

strategic effectiveness and operational efficiency of 120 representative teams in a large

multinational organization, using quantitative survey data and independent performance ratings.

Findings show that self-managing project teams perform better when they are independent but

not isolated, and shed light on the conditions under which embedded autonomy is most

beneficial.

Keywords:

project teams, self-management, autonomy, knowledge transfer, multinational organizations

3

In many multinational organizations, self-managing teams are assigned to carry out

complex collaborative projects (e.g., Earley & Gibson, 2002; Lipnak & Stamps, 1997; Shapiro,

Von Glinow, & Cheng, 2005; Snow, Snell, Davison, & Hambrick, 1996). Self-managing teams

are work groups that have high levels of team autonomy: collective control over the critical

decisions related to their work, including what tasks to perform and how to carry them out

(Hackman, 1987; Manz & Sims, 1993).1 According to theories of team self-management, such

autonomy can improve team performance by enabling and encouraging team members to make

independent decisions that serve the best interests of their tasks (e.g., Cohen, Ledford, &

Spreitzer, 1996; Cordery, Mueller, & Smith, 1991; Janz, Colquitt, & Noe, 1997; Langfred, 2000;

Pearce & Ravlin, 1987). However, empirical evidence on the performance benefits of team

autonomy is mixed. Indeed, a comprehensive review of the group effectiveness literature

concluded that although autonomy is beneficial for stable work teams that are responsible for

producing goods or providing services on a continuing basis, there is no compelling evidence

that it improves the performance of project teams, which are time-limited, produce one-time

outputs, and usually work on nonrepetitive tasks involving considerable knowledge, judgment,

and expertise (Cohen & Bailey, 1997).

To explain why autonomy may not be beneficial for project teams, I take an external

perspective that focuses on these teams’ relations with their environments (Ancona, 1987). I

propose that the benefits anticipated by theories of team self-management may not be realized

because autonomous teams can easily become isolated from their environments. Such isolation

can hinder project teams’ performance by leading them to miss valuable opportunities for

learning. Autonomy thus can be a double-edged sword: while enabling independence, it

encourages isolation. Given this tension, how can project teams capture the advantages of

autonomy but avoid the dangers?

4

The knowledge-based view of the firm suggests a possible answer to this autonomy

dilemma. According to this view, firms continuously generate and acquire new knowledge

through their wide-ranging activities, creating opportunities for learning: if organizational units

obtain and use knowledge from elsewhere, improved firm performance and competitive

advantage can result (Grant, 1996; Kogut & Zander, 1996). Most empirical research taking a

knowledge-based view of the firm has focused on knowledge flows to business units or national

subsidiaries, however, not on flows to project teams (e.g., Gupta & Govindarajan, 2000; Hansen,

1999; Lord & Ranft, 2000; Monteiro, Arvidsson, & Birkinshaw, 2008; Schulz, 2003; Tsai, 2001).

In this study, I extend the knowledge-based view of the firm to project teams by arguing that

autonomous teams can perform better if they obtain and use more task-relevant knowledge from

sources outside the team—and hence, avoid the dangers of isolation.

To provide an integrated theoretical framework for this team-centered perspective on

knowledge work, I propose that project teams perform most effectively under conditions of

embedded autonomy, where they are not only highly autonomous but also highly embedded in

their knowledge environments (cf. Evans, 1995). The concept of team autonomy is well

established in the work group literature (see, e.g., Cohen & Bailey, 1997). The concept of

embeddedness, on the other hand, has many meanings in organizational research. A pervasive

element of the embeddedness concept is strong relations between organizational actors and their

environments (Dacin, Ventresca, & Beal, 1999). Organizational actors who have strong relations

with their environments, including social, cultural, political, and institutional relations, are

viewed as more embedded (e.g., Baum & Dutton, 1996; Granovetter, 1985; Zukin & DiMaggio,

1990). For this study, in which the organizational actors are project teams engaged in knowledge

work, I focus specifically on knowledge embeddedness, which refers here to the strength of the

knowledge-sharing relations between teams and their environments, as measured by the extent to

5

which teams obtain knowledge from sources outside the team. I view teams as more embedded

in their knowledge environments if their members obtain more task-relevant knowledge from

sources outside the team, including sources within as well as beyond the organization.

In this study, I investigate the effects of embedded autonomy in a multinational

organizational setting. Although the theoretical framework is relevant for domestic firms too,

examining project teams in multinational organizations is particularly important for extending

research on team self-management because the complexity of multinational settings amplifies the

tension between independence and isolation (cf. Roth & Kostova, 2003). The need for

independence is particularly high in multinational organizations because project teams must

deliver locally differentiated products and services that fit the wide variety of contexts in which

these organizations operate, but the risks of isolation are also particularly high because these

organizations’ global operations are so extensive and dispersed (Bartlett & Ghoshal, 1989;

Prahalad & Doz, 1987). Additionally, the wealth of knowledge created within multinational

organizations increases opportunities for worldwide learning (Kogut & Zander, 1993),

amplifying the potential value of knowledge embeddedness not only for national subsidiaries

(Kostova, 1999; Rugman & Verbeke) but also for the project teams that work within them.

Below, I develop theoretical foundations for the embedded autonomy perspective and

build on them to identify boundary conditions under which autonomous teams benefit most from

knowledge embeddedness. I examine the effects of embedded autonomy on two dimensions of

performance that are critical for many project teams: strategic effectiveness, which refers here to

the extent to which the project outputs delivered by the team meet and further the organization’s

strategic goals, and operational efficiency, which refers here to how economically the team

utilizes available resources in delivering the project outputs.

6

AUTONOMY AND EMBEDDEDNESS IN PROJECT TEAMS

The Double-Edged Sword of Team Autonomy: Independence and Isolation

According to theories of team self-management, collective control over task-related

decisions has motivational benefits for team members that can improve their task performance.

Autonomy motivates team members by increasing their sense of responsibility and

accountability for a project (Cohen et al., 1996; Hackman, 1987; Latham, Winters, & Locke,

1994; Janz et al., 1997). Individuals also become more emotionally committed to groups that

strengthen their sense of control (Lawler, 1992), which can be helpful for building the collective

identity that is often at risk in transnational teams reliant on electronic communication (Shapiro,

Furst, Spreitzer, & Von Glinow, 2000). Further, autonomy signals the apparent endorsement of

senior managers (Langfred, 2000). Such endorsement can increase a team’s status, performance

expectations, and anticipated rewards, motivating members to invest more in its project.

Beyond being motivational, team autonomy also allows the individuals who are closest to

a team’s task to make decisions relevant to that task. To the extent that these individuals have

information about the task that is not available to external managers, their participation in critical

task decisions can result in informational benefits for the team (cf. Langfred & Moye, 2004;

Locke, Alavi, & Wagner, 1998; Miller & Monge, 1986). Additionally, autonomy buffers teams

from excessive external influence by giving them freedom to make decisions that contradict

received wisdom or outsiders’ wishes without fear of negative repercussions. Team members can

make the decisions that they believe best without the need for time-consuming ambassadorial

activities to secure political support or head off possible threats (cf. Ancona & Caldwell, 1992).

For similar reasons, new-product development initiatives and change implementation units often

are given unusual latitude in organizations (Kanter, 1988; Tushman & O’Reilly, 1996).

7

By providing these motivational, informational, and buffering benefits, autonomy enables

as well as encourages team members to make independent decisions that are in the best interests

of their tasks. Indeed, the benefits of autonomy are a central theme not only in theories of team

self-management but also in the literatures on worker empowerment and high-performing

organizations, where autonomy is viewed as contributing to empowerment and effective

performance. For example, Kirkman and Rosen (1999) identified team autonomy as the most

important of four distinct and mutually reinforcing conditions for team empowerment, together

with potency, meaningfulness, and impact. Similarly, in the high-performing organizations

literature, Lawler, Mohrman, and Ledford (1998) argue that organizations are most effective

when they implement “high-involvement work practices” that increase employees’ power as well

as their information, knowledge, and rewards.

Of particular relevance to project teams engaged in knowledge work, studies of team

learning have shown that autonomy can have positive effects on learning behaviors within a

team, including encouraging experimentation, communication, and codification of experience

(Gibson & Vermeulen, 2003), stimulating collective reflection on alternative courses of action

(Edmondson, 2002), and making the team more proactive in seeking to improve its work

processes and find innovative solutions to work problems (Hyatt & Ruddy, 1997; Kirkman &

Rosen, 1999; Wellins, Byham, & Wilson, 1991). Despite these potential benefits, however, the

empirical evidence indicates that team autonomy does not always have positive effects on team

performance (Cohen & Bailey, 1997).

Prior research on why self-managing teams may not perform well has typically taken an

internal perspective, focusing on the nature and structure of relations among team members. For

example, when members have low interdependence, team autonomy can impede rather than

improve performance because they have trouble coordinating with each other to accomplish

8

group tasks (Janz et al., 1997; Langfred, 2005). On the other hand, high interdependence

constrains and frustrates team members with strong preferences for individual autonomy

(Wageman, 1995). Similarly, empowerment is less effective when a team’s members have

negative feelings toward collaboration (Kirkman & Shapiro, 2001) or perceive a lack of trust

within the team (Kirkman & Rosen, 1999). Yet the performance of a self-managing team also

suffers when high trust within the team combines with high individual autonomy to make team

members reluctant to monitor each other (Langfred, 2004). Furthermore, team self-management

can create its own dysfunctional dynamics. For example, the heightened status, performance

expectations, and anticipated rewards in autonomous teams may encourage struggles for power

among team members (Barry, 1991). Self-managing teams that are torn by internal conflict may

restructure themselves to weaken interdependence and reduce individual autonomy, creating a

less than ideal team design (Langfred, 2007). And self-managing teams are also particularly

vulnerable to concertive control, in which the motivation of team members can suffer as peers

monitor each other and impose penalties for norm violations (Barker, 1993).

For project teams in particular, however, an external perspective suggests that the

expected benefits of autonomy may fail to materialize for a previously unexplored reason:

Autonomous teams that can make decisions without having to solicit input from outsiders may

become isolated from their environments, to the detriment of their performance. For these teams,

isolation is the other edge of the sword of independence. For example, “groupthink” among

President Kennedy’s advisors prior to the 1961 Bay of Pigs debacle was exacerbated by their

failure to solicit inputs from outsiders who might have broken through the group’s blind spots

and taboos on divergent opinions (Janis, 1982). Similar problems have since derailed decision

makers in events ranging from the NASA shuttle disasters to the failure of the U.S. intelligence

agencies to anticipate the terrorist strikes of 9/11. Retrospective analyses have traced the roots of

9

these tragedies in part to high autonomy that isolated key decision-making groups from their

environments (Kean & Hamilton, 2004; Vaughan, 1990).

For project teams in more typical organizational settings, isolation can impede strategic

effectiveness and operational efficiency by leading teams to miss valuable opportunities for

learning. First, an isolated team can overlook insights that might improve its strategic

effectiveness. For example, studies of “boundary spanning” have shown that work groups can

obtain useful information for their tasks by scanning their environments for new technical and

market knowledge (e.g.,Ancona & Caldwell, 1992; Tushman, 1977), and innovation research has

shown that brainstorming groups can increase their creativity by seeking ideas and inspiration

from outside sources (Hargadon & Sutton, 1997). Teams that do not solicit knowledge from

outsiders may overlook such opportunities to develop, refine, and test their own ideas. Such

missed opportunities for learning particularly harm a team’s strategic effectiveness, since inputs

about options for serving strategic goals, guidance on how best to serve them, and feedback on

the team’s choices all can help it deliver a project that furthers its organization’s strategic goals.

Second, isolation can impede the operational efficiency of project teams by leading

members to waste scarce time and energy duplicating solutions already achieved and mistakes

already made outside their teams. Even if an autonomous team manages to successfully replicate

existing solutions and avoid past mistakes, ignoring available knowledge reduces its operational

efficiency relative to another team that uses such knowledge. For example, a management

consulting team working for a new client might conduct a benchmarking analysis of competitors

that it could have executed more efficiently by adapting an existing analysis done for another

client in the same industry (Haas & Hansen, 2007). Similarly, a product development team might

take longer to bring a new product to market if it misses an opportunity to use intrafirm networks

to identify and transfer task-related knowledge (Hansen, 1999). Highly autonomous teams are

10

likely to be especially prone to “reinventing the wheel” by inefficiently duplicating rather than

importing existing solutions because the status and performance expectations conveyed by the

endorsement of autonomy encourage the “not-invented-here” syndrome (Katz & Allen, 1982).

The Role of Knowledge Embeddedness

Because isolation can lead autonomous project teams to miss opportunities for learning,

the direct effects of team autonomy on strategic effectiveness and operational efficiency may not

always be as positive as anticipated by team self-management theories, or even positive at all.

However, high knowledge embeddedness—strong knowledge-sharing relations between teams

and their environments—can increase the positive effects of team autonomy by mitigating the

dangers of isolation.

Strong knowledge-sharing relations often accompany close personal ties but do not

necessarily require them. Research on social networks suggests that close ties can facilitate

knowledge exchange: for example, close inter-firm relations encourage fine-grained information

transfer (Granovetter, 1985; Uzzi, 1997), and close ties between business units within firms ease

the transfer of complex knowledge (e.g., Hansen, 1999; Tsai, 2002). For project teams, close ties

with outsiders can similarly increase the willingness and ability of those outsiders to share their

own knowledge with the team members, or to refer them to other experts or document sources

that can answer their questions. Yet strong knowledge-sharing relations can emerge even where

close ties did not previously exist. In many knowledge-intensive firms, a culture of knowledge

sharing enables team members to consult a directory of experts, call someone they have never

met, and expect a prompt and helpful response to their query (e.g., Criscuolo, Salter, & Sheehan,

2007). Further, team members can obtain codified knowledge from people with whom they have

no close personal ties by accessing documents they have uploaded to electronic databases or

11

websites (e.g., Hansen & Haas, 2001). Knowledge embeddedness thus involves direct or

mediated relational exchanges between actors, but does not always rely on close personal ties.

Operationally, I define knowledge embeddedness as the total amount of task-relevant

knowledge that a team’s members obtain from sources outside the team. These sources can be

situated both within and beyond their organization. Sources within the organization include close

colleagues, directories of in-house experts who can provide specialized advice, and company

databases. Sources beyond the organization include personal contacts in other organizations who

can offer advice and information, professional associations and conferences, and websites that

provide access to public data. A team’s level of knowledge embeddedness is a group-level

aggregation of individual behaviors: in a typical project team, some members obtain more

outside knowledge than others, and members obtain knowledge from different sources.

Knowledge from sources outside the team can improve the strategic effectiveness of an

autonomous project team by providing new insights that enable it to develop appropriate and

innovative solutions for clients. For example, transferring “best practices” that have been

developed elsewhere in an organization helps a team to benchmark and improve its work

processes (Szulanski, 1996), and soliciting advice from firm experts helps a team win

competitive bids for new client contracts (Haas & Hansen, 2005). More generally, well-

established knowledge management norms and procedures increase team learning in

multinational organizations by stimulating team members to find new ways to improve their

work and helping them identify and implement novel solutions (Zellmer-Bruhn & Gibson, 2006).

Utilizing knowledge from their environments can also improve the operational efficiency

of autonomous project teams by helping them avoid duplicating past solutions or mistakes. For

example, reusing codified knowledge by downloading documents from firm databases saves time

(Haas & Hansen, 2007), and seeking colleagues’ feedback helps teams to avoid past mistakes

12

without incurring the costs of trial-and-error experimentation themselves (Levitt & March,

1988). Obtaining knowledge from outside the team—high knowledge embeddedness—thus helps

autonomous teams to overcome the dangers of isolation by decreasing duplication as well as

increasing insights, improving both their strategic effectiveness and operational efficiency.

Hypothesis 1. Knowledge embeddedness moderates the influence of team autonomy on the

strategic effectiveness and operational efficiency of project teams.

If knowledge embeddedness improves an autonomous team’s strategic effectiveness and

operational efficiency by providing new insights and decreasing duplication, its benefits will be

greater when the team’s need for outside knowledge is greater. To improve the strategic

effectiveness and operational efficiency of their projects, autonomous teams need outside

knowledge that is different from their own more than they need outside knowledge that is

redundant with their own. Additionally, they need outside knowledge more for highly novel or

highly complex projects than for less novel or complex ones. The benefits of knowledge

embeddedness for an autonomous team thus are likely to depend on the characteristics of the

knowledge obtained as well on the characteristics of the project for which it is obtained.

Knowledge characteristics. Obtaining knowledge that is different from a team’s own

offers greater benefits than obtaining knowledge that overlaps with the team’s own (cf. Burt,

1992; Reagans & Zuckerman, 2001). For example, knowledge about the country-specific

conditions relevant to a project has more value than knowledge about the technical aspects of the

work if a team’s members have high functional expertise but little local market expertise (cf.

Lord & Ranft, 2000). For such a team, insights about the priorities of regional governments, for

instance, can shape decisions about how and where to introduce a new product or service more

13

fundamentally than incremental information about technical product or service requirements that

the team has probably already anticipated.

Similarly, knowledge from sources beyond a team’s own organization can offer greater

benefits than knowledge from sources within the organization if the team members’ external

knowledge networks are more differentiated than their internal knowledge networks (cf. Collins

& Clark, 2003). A team that is seeking ideas about how to best tackle a difficult product design

problem, for example, might deploy several team members’ external networks and eventually

identify a contact who has successfully addressed a parallel design problem in a different

industry (Hargadon & Sutton, 1997). In contrast, using the team members’ overlapping internal

networks to search within the organization is likely to generate the same leads to colleagues with

less diverse experiences, and ultimately lead to the same dead ends. Thus, nonredundant

knowledge (such as country-focused or externally-sourced knowledge) can be expected to offer

project teams more new insights and more opportunities to avoid duplication than redundant

knowledge (such as technically-focused or internally-sourced knowledge):

Hypothesis 2. Knowledge embeddedness more strongly moderates the effects of team

autonomy on the strategic effectiveness and operational efficiency of a project team

if the knowledge obtained is nonredundant.

Project characteristics. Obtaining knowledge from the team’s environment also offers

greater benefits for more novel or complex projects. When a project is highly novel for the team

members involved, relying exclusively on their accumulated past experiences endangers strategic

effectiveness because the team members have little or no experience with similar projects.

Failing to utilize outside knowledge can also impede operational efficiency because the team will

have to figure out how to address novel demands through trial and error, making and rectifying

14

mistakes along the way. For a highly novel project, seeking inputs from sources outside the team

thus offers greater benefits than relying solely on exploiting the team’s own knowledge (March,

1991). In contrast, for a relatively routine project, relying on the team’s knowledge has less

downside since the potential value of new knowledge from outside sources is lower (Haas &

Hansen, 2005).

Likewise, the more complex its project, the more a team endangers its strategic

effectiveness by not soliciting inputs that can help it to identify and address unexpected

interdependencies, consequences, or problems (cf. Tushman, 1978). The benefits of outside

knowledge thus are greater for more complex projects. Operating in isolation also impedes the

team’s operational efficiency because it takes longer to develop solutions for highly complex

projects than for simple ones (Hobday, 2000). Importing tried-and-tested solutions from outside

the team reduces the opportunity costs of reinventing these solutions by trial and error more for

complex than for simple projects. Hence, task-relevant knowledge from sources outside the

teams themselves is particularly beneficial for novel or complex projects:

Hypothesis 3. Knowledge embeddedness more strongly moderates the effects of team

autonomy on the strategic effectiveness and operational efficiency of a project team

if the project is highly novel or highly complex.

DATA AND METHODS

I tested the hypotheses using data collected in a multimethod field study of project teams

at an international financial institution that provides financial and technical assistance to national

and regional governments in developing countries. This multinational organization employs over

10,000 worldwide and supports offices in more than 100 countries. Self-managing project teams

carry out the organization’s financial and technical projects.

15

Qualitative Data

To understand the teams and their projects and develop the survey for this study, I

conducted 70 semistructured interviews lasting one to three hours each over a period of six

months, mostly prior to the collection of quantitative data. With assistance from the senior

managers who sponsored this study, I started by interviewing 20 managers and staff who could

provide an overview of the organization’s functions and operations; interviewees included

people responsible for organization-wide strategy and change management, knowledge

management, project evaluation, human resources, and the staff association. Next, I conducted

18 interviews with leaders and members of project teams based at headquarters to develop

insight into teams’ projects, how teams were structured, how they worked, and the types of

problems they encountered. During 7 further interviews at the organization’s Moscow office, I

obtained further insight into these issues from the perspective of teams based outside the United

States. Last, I conducted 25 interviews as part of detailed case studies of another seven teams,

interviewing the leaders and all available members who were working on the project at the time

to gain detailed insights into what they did day-by-day. After these interviews, I systematically

reviewed my interview notes and transcripts as well as internal memos and project documents

obtained from the teams to identify patterns pertaining to decision-making autonomy, knowledge

utilization, and other characteristics of the teams, their projects, and their environments.

I studied both financial project and technical project teams. The former prepared

development programs for particular countries that were ultimately backed by loans to the client

governments for their implementation, and the latter conducted research and analysis on topics of

relevance to particular client countries that were often used as a basis for subsequent

development programs. For the purposes of my study, the teams that worked on these two types

of projects were substantively similar: a typical team was composed of economists and technical

16

specialists with expertise in domains such as public finance, infrastructure, or engineering who

were brought together for a specific project. Teams rarely worked on more than one project

together; instead, team members were usually assembled by a designated team leader who sought

out available individuals with the required skills as needed during the project. For both financial

and technical projects, the main outputs were detailed reports documenting the project findings

and recommendations. In the quantitative data set described below, the average team included

8.5 members, each of whom spent from one month to four years with that team while also

working on two to ten other projects with other teams. The average age of a team member was

44 years, and 65 percent were men. Seventy percent of the team members were based at the

organization’s U.S. headquarters; they conducted their work internationally by flying in and out

of the client countries several times during their projects.

My interviews and observations indicated that although the project teams were all self-

managing to at least some degree, the level of collective control over task decisions varied across

teams. The determinants of team autonomy were multiple and often idiosyncratic, including

bureaucratic requirements, the decentralized matrix structure of the organization, the distribution

of informal as well as informal authority in the units, the status of the team members, the style of

the senior managers, the limitations of the budgets, and the extent of client participation in the

projects. Yet team autonomy was important because it enabled the teams to make independent

decisions. For example, one team I studied was working on preparing a $50 million slum-

upgrading program for an overpopulated city in an impoverished West African country. This

team had high decision-making autonomy partly because it had a team leader who was well-

respected in the organization and partly because it had secured a $1 million project preparation

grant that gave it control over its resources. This autonomy enabled it to decide on the precise

scope and specifics of its objectives, which focused on improving living conditions for over a

17

million slum-dwellers by expanding their access to housing and basic utilities. In one team

meeting I observed, team members debated at length about whether to extend the project to a

second city or keep it focused in the original city. They decided on the extension and

subsequently made numerous follow-up decisions to implement this strategic redirection,

including hiring additional experts in public finance and sanitation engineering to bolster the

team’s capabilities. In contrast, another team, one working on an education reform project for an

East-Central European ministry, lacked comparable autonomy over its decisions. This team

found that both senior managers and stakeholders in the country were eager to impose their own

objectives on the project. As a result, its mandate expanded from establishing a student loan

program to developing a comprehensive program of investment in both secondary and higher

education, while its resources did not grow accordingly. Ultimately, the viability of even the

original narrow reform program was threatened.

Although autonomy enabled teams to make independent decisions that served the best

interests of their projects, my qualitative data indicated that teams that were very autonomous

could become isolated from their environments. In a team charged with preparing an $85m urban

infrastructure investment plan for a large metropolitan area in Latin America, for example, the

costs of isolation were dramatically illustrated during an important internal review meeting.

Because the team could make its own decisions, the plan was already well advanced when team

members decided to solicit feedback on the design. Just hours before the meeting, one external

expert circulated a memo criticizing fundamental aspects of the design, to the surprise and

consternation of the team members, and the ensuing meeting was chaotic. The expert’s core

concern was that the design contravened lessons about water supply that had been learned from

previous projects; the team members had failed to seek advice and input from either in-house or

18

outside experts who would have made them aware of this. The design had to be radically altered,

and the project was set back by several weeks.

Recognizing that isolated teams could miss valuable opportunities to learn from outside

knowledge, senior managers in the organization had introduced a high-profile knowledge

management initiative several years prior to this study, making substantial investments in

“communities of practice,” “expert directories,” “help desks,” and electronic document

repositories. The organization had won widespread acclaim and awards from independent

business groups for its leadership in knowledge management. Many organization members

acknowledged that taking advantage of these resources could help them in their work. As an

infrastructure specialist overseeing a highway project in Russia noted, “Teams are more effective

when they apply past experience from other teams—it’s like a cross-subsidy.” However, teams

varied in the extent to which they sought out knowledge from sources outside their own borders.

The autonomous Latin American team missed opportunities to learn from others, for example,

but the autonomous West African team spent much of its time soliciting input from diverse

parties with wide-ranging knowledge, including not only managers and other teams within the

organization with experience implementing similar projects in other countries, but also

international experts who specialized in urban poverty issues, government officials who could

help them navigate the local power structure, private sector lobbyists who sensitized them to

corporate concerns, and activists representing city dwellers who would have to be relocated

when their homes were razed. Although many of these parties had vested interests in influencing

the team, understanding their concerns was important for shaping the project, and the

information and advice gained was invaluable for developing its strategic direction and enabling

the team to avoid mistakes. Similarly, an autonomous team working on social service provision

in Romania benefited from outside knowledge: by reviewing past programs and consulting with

19

experts around the world, team members avoided investing extensive effort in a problematic

decentralization approach and instead based their approach on a successful initiative pioneered in

a neighboring country. These examples from the qualitative phase of the study thus provided

initial evidence to support the proposition that knowledge embeddedness helps autonomous

teams perform better.

Quantitative Data

To test the embedded autonomy hypotheses quantitatively, I collected data from three

independent sources: the organization’s project evaluation unit, a survey administered to

members of all the teams that were evaluated, and archival project records.

To evaluate projects, the organization had a dedicated unit of 20 full-time staff who drew

a random sample of financial and technical projects from the full population of projects

completed in the organization each year. A different panel of experts identified by the project

evaluation unit assessed each project. Each customized panel included at least two respected

experts in the project’s area with no prior connections to the project. The panel reviewed project

documents, interviewed the team leader, and completed a detailed questionnaire developed

through organization-wide consultation and refined during previous years of the evaluation

process. Using these inputs, panelists reached a joint evaluation. Although each project was

evaluated by a different expert panel, the project evaluation unit took care to ensure that results

were robust across panels. In addition to providing detailed questions and hands-on guidance to

the panels during evaluations, the unit had tested the inter panel reliability of ratings and

established that different panels were highly likely to rate the same project similarly. Interrater

reliability within a panel of experts was not a concern because each evaluation was collectively

determined by the panelists. The project evaluation unit provided me with the evaluations of the

120 project teams sampled in the year of my study (60 financial and 60 technical teams).

20

I sent a survey to all the members of these teams, after conducting pretests with 52 team

members who were not part of the evaluated sample to refine the survey questions and to ensure

the validity and shared understanding of the items in the organizational context. The survey

directed team members to focus on the particular project that was undergoing evaluation, as

identified on the front page. Because levels of involvement in a project varied, I asked each team

leader to provide a roster distinguishing between core members, according to the leader’s own

definition, and those who were less centrally involved in the project, and I sent a slightly

modified survey to the noncore members. Only 25 of the 1,021 team members surveyed

appeared on more than one team roster, indicating that respondents who participated in more

than one team were unlikely to bias the data. I received complete survey responses from 550

team members (54%). Teams for which fewer than 50 percent of respondents were core

members were excluded (Wageman, Hackman, & Lehman, 2005); 96 teams qualified for the

study (50 financial and 46 technical teams; 80%). Tests for selection bias showed no significant

differences in the overall ratings, project type, region, or division of the 24 excluded teams.

Dependent variables. The two dependent variables in this study were the strategic

effectiveness and operational efficiency of a project as rated by its expert panel. Both variables

used ordinal scales on which 3 was “highly satisfactory,” 2 was “satisfactory,” and 1 was

“marginal or unsatisfactory.” To evaluate strategic effectiveness, the expert panel used ten

questions that were customized for financial and technical projects; Table 1 presents these

questions (modified for confidentiality). To arrive at an overall strategic effectiveness rating, the

experts on a panel took into account their understanding of the project and its distinctive

challenges as well as the scores on these questions. To evaluate operational efficiency, the expert

panel reviewed the time taken for a project, its budget, and the other resources, specifically skill

mix, used. The panel then evaluated the appropriateness of these investments in light of the

21

complexity and context of the project to arrive at an overall rating. Of the 96 projects in the final

data set, 41% received a rating of 3, 51% received a rating of 2, and 8% received a rating of 1 for

strategic effectiveness. For operational efficiency, 27% received a rating of 3, 57% received a

rating of 2, and 16 % received a rating of 1.2

----- insert Table 1 and Table 2 about here -----

Team autonomy. Hackman (1987) identified four categories of critical decisions that

contribute to team autonomy. Through the field interviews, I identified five decision areas

critical in this organizational setting within each of the Hackman categories: (1) decisions about

objectives concerned project initiation, overall priority, boundaries and scope, specific

components, and level of innovation; (2) decisions about resources concerned budget size,

additional funding, level of information input, team training or coaching, and team rewards or

recognition; (3) decisions about task and team design concerned project pacing, feedback

solicitation, quality standards, staffing requirements, and selection of team members; and (4)

decisions about processes concerned setting up and managing missions, level of interaction with

clients and management, and handling conflict. In the survey, team members were asked to

report on their team’s level of influence over each of these decisions, using the items shown in

Table 2. I averaged responses over the 20 decisions to create a measure of team autonomy (α =

.90, ICC = 0.05, p < .10, rwg = 0.85).

Knowledge embeddedness. My interviews indicated that team members in this

organization typically classified the sources outside the team from which they obtained

knowledge into four categories according to geographic boundaries (local/global) and

organizational boundaries (internal/external). Survey items (shown in Table 2) asked the team

members how much relevant knowledge they obtained during their project from each of these

four sources: the organization’s country office (local, internal), the rest of the organization

22

(global, internal), the client country (local, external), and the global community (global,

external). The client country included the client government and stakeholders such as local

NGOs (nongovernmental agencies), businesses, and intended beneficiaries, and the global

community included those who worked on development issues around the world, including

international NGOs, think tanks, and academics. Although the survey questions could not

establish with certainty whether teams actually paid attention to and used the knowledge they

obtained, the emphasis on relevant knowledge encouraged them to recall knowledge that they

had found useful. Team members were asked separately about technical and country knowledge,

which were defined respectively as “knowledge about the technical aspects of the work—the

professional skills, competencies, and expertise relevant to the project” and “knowledge about

the local environment—the country-specific conditions relevant to the project.” For the variable

knowledge embeddedness, I averaged the team members’ responses to the resulting eight items to

create an overall measure of knowledge obtained by the team (α = .81, ICC = 0.06. p< .05, rwg

= 0.69). To test for the interaction effect predicted in Hypothesis 1, I multiplied each team’s

knowledge embeddedness score by its team autonomy score.

Knowledge characteristics. I distinguished less redundant from more redundant

knowledge in two ways: (1) by comparing country knowledge to technical knowledge and (2) by

comparing knowledge from within the organization to knowledge from beyond the organization.

My qualitative research indicated that project teams in this organization typically found

country knowledge to be less redundant than technical knowledge because team members were

usually chosen for functional expertise rather than familiarity with a client country. Indeed,

although the organization had national hiring quotas, it purposefully avoided staffing a project

with nationals of the client country and rotated employees onto projects in different countries

frequently to preserve a professional distance from clients. Consequently, team members usually

23

possessed more technical knowledge than country knowledge. Additionally, information about

economic and social conditions in developing countries that might provide country knowledge is

often very poor or nonexistent, whereas technical knowledge usually builds on formal education

or experience in other countries and so is more abundant. Further, the project teams could not

always identify and access country knowledge as easily as technical knowledge because they

were mostly based in the United States rather than the client countries. Since country knowledge

inputs were harder to acquire as well as scarcer, they were less likely to be redundant than

technical knowledge inputs. This insight from the qualitative data was supported by the survey

data, in which team members reported possessing significantly more technical knowledge than

country knowledge prior to the projects (means = 3.81 and 3.43 respectively, t = 4 .50, p <

.001), as well as seeking less technical knowledge than country knowledge during the projects

(means = 2.91 and 3.08, respectively, t = 6.05, p < .001).

I therefore tested Hypothesis 2 by splitting the overall knowledge embeddedness variable

into two subconstructs that captured country knowledge and technical knowledge separately

(country: α = .70, ICC = 0.08, p < .01), rwg 0.66; technical: α = .72, ICC = 0.06, p < .05, rwg =

0.65). I established the convergent and discriminant validity of these two subconstructs using

two approaches (Venkatraman & Grant; 1986): a multitrait-multimethod matrix analysis of the

group-level measures indicated that the average within-scale correlations for the subconstructs (r

= .48, r = .46) exceeded the average between-scale correlation (r = .27), and a group-level

confirmatory factor analysis on the eight knowledge embeddedness items using maximum-

likelihood estimates indicated that the two-factor structure was superior to a one-factor structure

(∆χ2

1 = 6.10, p < .05). According to Hypothesis 2, the interaction effect between knowledge

embeddedness and team autonomy should be stronger for the less redundant country knowledge

than for the more redundant technical knowledge.

24

My interviews further indicated that knowledge from beyond the organization tended to

be less redundant for project teams than knowledge from within the organization. Many team

members had worked in the organization for multiple years, and most were located at its

headquarters. As a result, the personal networks that they used to seek knowledge within the

organization tended to overlap considerably. The team members also all had access to the same

knowledge intranet within the organization, leading them to the same internal documents and

data when they searched electronically. Seeking knowledge within the organization thus was

likely to yield insights, ideas, and information that overlapped either with the team members’

own prior knowledge or with the inputs obtained by other team members from the same source.

In contrast, the external networks and information resources utilized by the team members were

more diverse, since the organization hired individuals from a wide range of countries,

institutions, and professions, all of which were potential sources of nonredundant knowledge for

teams. The quantitative data were consistent with this expectation: survey respondents reported

obtaining statistically higher levels of knowledge from beyond the organization than from within

it (means = 3.15 and 2.83, respectively, t = –6.24, p < .001), showing that team members sought

more inputs from sources of nonredundant knowledge. I therefore separately averaged the four

items for intrafirm knowledge (obtained from sources within the organization, i.e. the country

office or rest of the organization) and the four items for extrafirm knowledge (obtained from

sources beyond the organization, i.e. the client country or global community) to create two

knowledge embeddedness subconstructs (intrafirm, α = .78. ICC = 0.05, p < .10, rwg = 0.62;

extrafirm, α = .80, ICC = 0.11, p < .01, rwg = 0.60). Convergent and discriminant validity tests

for these two subconstructs indicated that the average within-scale correlations (r = .37, r = .52)

exceeded the average between-scale correlation (r = .29), and the two-factor model fitted the data

better than a one-factor model (∆χ2

1 = 6.70, p < .01). Hypothesis 2 predicts that the interaction

25

effect between knowledge embeddedness and team autonomy should be stronger for the less

redundant extrafirm knowledge than for the more redundant intrafirm knowledge.

Project characteristics. To capture project complexity, I averaged ratings on three survey

items reported in Table 2 (cf. Tushman, 1977) (α = .77, ICC = 0.17, p < .01, rwg = 0.74). The

higher the average team member score on this variable, the higher the complexity of a project.

To capture project novelty, I used two survey items that asked the team members how

much prior knowledge of relevance to their project they had (see Table 2), which were then

reverse-coded and averaged (α = .44, ICC = 0.12, p < .01), and rwg = .55). The higher the average

team member score on this variable, the higher the novelty of the project for the team.

Using median splits, I divided the sample into high- versus low-complexity projects and

into high- versus low-novelty projects. To test Hypothesis 3, I compared the models across the

paired sets of projects to establish whether the embedded autonomy effects were stronger for

more complex than less complex projects and for more novel than less novel projects. An

alternative approach would be to use three-way interactions between knowledge embeddedness,

team autonomy, and the project variables instead of median split comparisons, but the number of

observations in the data set made this infeasible.

Control variables. To account for other influences on strategic effectiveness and

operational efficiency, the models included variables that captured the team size (number of team

members) and the team location (1, “headquarters”; 0, “country office”). Because a team can be

expected to perform better if it is a real team rather than a team in name only, I measured the

extent to which the team was a clearly identifiable interdependent group using two items from

Wageman et al. (2005) reported in Table 2 (real team: α = .66, ICC = 0.20, p < .01, rwg = 0.61).

I also included a measure of members’ satisfaction with their team using four Wageman et al.

items, also reported in Table 2. I assumed that better internal team relations would be reflected in

26

higher team satisfaction and thus used this measure to control for how well the team members

worked together as well as for possible resulting biases that might have affected their responses

to the other survey items (team satisfaction: α = .81, ICC = 0.18, p < .01, rwg = 0.82). To capture

any differences due to the seniority or work experience of team members, I included a measure

of their average organizational tenure in years at the start of the sampled project, as well as their

average nonorganizational tenure measured as years spent in other organizations. Additionally, I

included a measure of the proportion of team members who returned their surveys late, since if

they had found out the results of their project evaluation before responding to the survey, the

ratings might have biased their responses. The project evaluation unit reported that seven days

was a reasonable estimate of the time it usually took to notify a team of the outcome of an

evaluation, so surveys were coded as late if they were returned more than seven days after the

project evaluation was finalized; shorter and longer cut-off periods did not change the results for

this variable, late respondents. Because core members might have had different roles and views

than noncore members, the models also included core respondents, the proportion of survey

respondents in each team who were core team members, as identified in the team member roster.

Using archival records, I also calculated the project duration in logged days from initiation to

completion, since longer projects could be more novel or complex. Finally, the models included

project type (1, “financial”; 0, “technical”) to capture mean differences between the two types of

projects conducted by the teams in the study.3

RESULTS

----- insert Tables 3, 4, and 5 about here -----

Table 3 reports descriptive statistics and correlations. I standardized the main

explanatory variables (by subtracting by the mean and dividing by the standard deviation) to

avoid high levels of multicollinearity with the interaction terms (Neter, Wasserman, & Kutner,

27

1985). This table shows that the two dependent variables are significantly but not highly

correlated (r = .37, p < .01). Additionally, the correlation between team autonomy and

knowledge embeddedness is low (r = .15, n.s.), indicating that these constructs are orthogonal

rather than opposite: knowledge embeddedness does not reduce team autonomy, nor does team

autonomy preclude knowledge embeddedness. The average within-scale correlations for the

two constructs (r = .76, r = .34) exceeded the average between-scale correlation (r = .06), and a

supplementary confirmatory factor analysis indicated that the two-factor model provided a

better fit to the data than a one-factor model, further verifying the convergent and discriminant

validity of these constructs (∆χ2

1 = 181.2, p < .01).

Ordinal logit analysis was used to test the hypotheses because the two dependent

variables were both categorical and ordered (Long, 1997). Tables 4a and 5a show the results for

the strategic effectiveness models; Tables 4b and 5b show the results for the operational

efficiency models. All models include the full set of control variables, although they are not

shown in the tables for parsimony. For the strategic effectiveness outcome, a baseline model

that included only the control variables indicated that teams scored higher when they had more

members (β = 0.22, p < .01) and were real teams rather than teams in name only (β = 1.19, p<

.05). For the operational efficiency outcome, a baseline model indicated no significant effects of

the control variables.

Model 1a in Table 4a shows that the main effect of team autonomy on strategic

effectiveness is positive and significant: teams with more collective control over their task-

related decisions delivered projects that were higher in strategic effectiveness. In contrast,

model 1b in Table 4b indicates that the main effect of team autonomy on operational efficiency

is not significant. In additional models not shown here, I examined whether team autonomy

showed curvilinear effects. A second-order team autonomy variable did not have a negative

28

effect on strategic effectiveness or operational efficiency. Models 2a and 2b show that the main

effects of knowledge embeddedness are not significant for either strategic effectiveness or

operational efficiency. Additional models not shown here further indicated no effect of second-

order knowledge embeddedness variables. Models 3a and 3b replicate the main effect results

when the team autonomy and knowledge embeddedness variables are included simultaneously.

Models 4a and 4b report the results for Hypothesis 1, which proposes that knowledge

embeddedness moderates the effect of team autonomy on strategic effectiveness and operational

efficiency. These models reveal that the interaction between knowledge embeddedness and

team autonomy is positive and significant for both effectiveness and efficiency, supporting

Hypothesis 1. These interaction effects are plotted in Figures 2a and 2b to illustrate the

magnitudes of the implications for teams with varying levels of knowledge embeddedness and

team autonomy. High and low levels of knowledge embeddedness and team autonomy are set at

one standard deviation above and below their mean levels, respectively (Aiken & West, 1991).

The strategic effectiveness and operational efficiency vertical axes range from 1 to 3, giving a

maximum difference of 2 points between high and low effectiveness/efficiency projects. The

plots show that teams with high autonomy and high embeddedness delivered more strategically

effective and operationally efficient projects on average than teams with high autonomy but low

embeddedness (0.42 points and 0.38 points higher, respectively). Teams with high autonomy

and high embeddedness also were more strategically effective and operationally efficient than

teams with low autonomy and high embeddedness (0.74 points and 0.32 points higher,

respectively). However, high team autonomy was not always beneficial: the plots reveal that if

they had low knowledge embeddedness, teams with high autonomy delivered only marginally

more strategically effective projects (0.10 points higher), and less operationally efficient

projects (0.56 points lower) than teams with low autonomy.

29

----- insert Figures 2a and 2b about here -----

Models 5a and 5b and Models 6a and 6b present the results for Hypothesis 2, which

proposes that knowledge embeddedness more strongly moderates the effects of team autonomy

on strategic effectiveness and operational efficiency if the knowledge obtained is nonredundant.

Model 5a shows that for strategic effectiveness, the interaction between autonomy and country

knowledge is substantial and significant, whereas the interaction between autonomy and

technical knowledge is small and not significant. Model 6a shows that the interaction between

autonomy and knowledge from sources within the organization is not significant, whereas the

interaction with knowledge from sources beyond the organization is significant. This model also

indicates that the main effect of knowledge from sources beyond the organization is positive.

Accounting for main and interaction effects, the difference between the country knowledge and

technical knowledge effects is significant (χ2

= 2.89, p < .10), and the difference between the

intrafirm and extrafirm knowledge effects is also significant (χ2

= 3.88, p < .05).

Model 5b and model 6b show a similar pattern of results for operational efficiency: the

interaction effect with autonomy is significant for country knowledge but not for technical

knowledge, and significant for extrafirm knowledge but not intrafirm knowledge. For this

dependent variable, the difference between the country knowledge and technical knowledge

effects is significant (χ 2 = 3.02, p < .10), although the difference between the intrafirm

knowledge and extrafirm knowledge effects is not significant (χ2

= 1.29, n.s.). With the exception

of this one weak result, the findings for both dependent variables generally indicate that high

knowledge embeddedness aided autonomous teams more when the knowledge obtained was less

redundant (country-focused or externally-sourced) rather than more redundant with the team’s

own knowledge (technically-focused or internally-sourced), as Hypothesis 2 predicts.

30

The four sets of paired models presented in Tables 5a and 5b report the results for

Hypothesis 3, which proposes that knowledge embeddedness is especially advantageous for

autonomous teams that work on novel or complex projects. These results show a consistent

pattern: the interactions between knowledge embeddedness and team autonomy are larger and

significant for projects above the median level of project novelty (models 7a[ii] and 7b[ii]), but

smaller and not significant for projects below this level (models 7a[i] and 7b[i]). Likewise, the

interactions are larger and significant for projects above the median level of project complexity

(models 8a[ii] and 8b[ii]), but smaller and not significant below this level (models 8a[i] and

8b[i]). However, using a seemingly unrelated estimation procedure to test for the statistical

significance of the differences in the full models indicates that the paired complexity models are

significantly different from each other for strategic effectiveness but not for operational

efficiency (χ2

= 35.00, p < .01; χ2

= 16.93, n.s.); in contrast, the paired novelty models are

significantly different for operational efficiency but not for strategic effectiveness (χ2

= 25.31, p

< .05; χ2

= 11.80, n.s.). These results thus provide mixed evidence for the hypothesis that the

combination of knowledge embeddedness and team autonomy is more valuable for novel and

complex projects: although novelty increases the benefits of embedded autonomy for operational

efficiency, complexity increases the benefits of embedded autonomy for strategic effectiveness.

DISCUSSION

In an influential political sociology theory of nation states, Evans (1995) coined the term

“embedded autonomy” to argue that well-functioning nation states are those that can act

independently from their constituents but still continuously solicit and consider the opinions of

those constituents. In the very different domain of multinational organizations, the embedded

autonomy perspective developed here offers a similar argument: Self-managing project teams

31

can perform better when their members are independent but informed by knowledge from

sources outside the team.

The findings of the study show that the combination of embeddedness and autonomy

enables project teams to deliver more strategically effective and operationally efficient projects,

where autonomy takes the form of collective control over critical task-related decisions and

embeddedness takes the form of knowledge-sharing relations between teams and their

environments. In the multinational organization studied here, high team autonomy improved

strategic effectiveness but not operational efficiency for project teams. However, teams with high

knowledge embeddedness as well as high autonomy performed better on measures of both

strategic effectiveness and operational efficiency. Unexpectedly, teams with high autonomy but

low knowledge embeddedness were only marginally more strategically effective and less

operationally efficient than teams with low autonomy, indicating that high autonomy could hurt

rather than help teams that were isolated from their environments. Obtaining knowledge from

sources outside a team was particularly beneficial when the team’s need for the knowledge was

great: both strategic effectiveness and operational efficiency benefited more from nonredundant

knowledge than from redundant knowledge, while strategic effectiveness benefited more from

knowledge from sources outside the team when the project was complex whereas operational

efficiency benefited more when the project was novel. These results support a team-centered

theory of knowledge work that views team autonomy as providing independence at the risk of

isolation, and knowledge embeddedness as offsetting this risk by increasing insights and

decreasing duplication of past mistakes and existing solutions.

Converging Evidence and Alternative Explanations

Both the quantitative data and the qualitative insights presented in this study provide

converging evidence for the value of embedded autonomy for project teams. The interaction

32

between knowledge embeddedness and team autonomy was found to be positively and

consistently associated with two substantively different measures of project performance:

strategic effectiveness and operational efficiency. These two measures were significantly but not

highly correlated (r = .39, p < .01). Additionally, prior research has shown a similar, positive

association between the interaction and a third measure of project performance—project

quality— that is different from strategic effectiveness and operational efficiency (r = .69, p <

.01, and r = .45, p < .01, respectively) (Haas, 2006). Although this prior research did not utilize

the concept of embedded autonomy or identify its boundary conditions, the robustness of the

interaction effect over the three measures indicates that the embedded autonomy framework has

some generalizability across multiple performance metrics.

To further explore whether the findings were generalizable across project types as well as

performance metrics, in additional analyses I examined whether the embedded autonomy effect

held for both financial projects and technical projects considered separately. The results

indicated that the positive effect of embedded autonomy on strategic effectiveness was

significant for both financial and technical projects (β = 1.61, p < .05; β = 1.25, p < .05); the

effect on operational efficiency was significant for technical projects though not quite significant

for financial projects (β = 1.52, p < .01; β = 0.59, p = .16). Although further corroboration is

needed, this pattern of results generally suggests that the embedded autonomy framework has

some generalizability across project types too, at least in this organization.

Although the data support the importance of embedded autonomy for project teams,

some of the empirical findings in this study may have alternative explanations. One possibility is

that external factors drove the observed relationships between team autonomy, knowledge

embeddedness, and performance. For example, perhaps some teams were given more autonomy,

obtained more knowledge, and also performed better because their members were more expert or

33

experienced. This possibility would suggest that team member quality should be strongly

correlated with both team autonomy and knowledge embeddedness. However, three measures of

team member quality utilized in this study, their prior knowledge and their organizational and

nonorganizational tenure, revealed only low and mostly nonsignificant correlations with

decision-making autonomy and knowledge embeddedness. A related argument is that novel or

complex projects that required more knowledge were staffed with better teams that were given

more autonomy. But correlations between project novelty or complexity and team autonomy

again were low and nonsignificant. The qualitative data also provided little support for this

argument, since team leaders often complained that they could not strategically staff more

demanding projects with better team members because they had to take whoever was available,

and also because it was hard to predict the difficulty of a project in advance.

A different alternative explanation is that the findings of the study might be a result of

post-evaluation attribution bias. Some projects had undergone the full quality evaluation process

before the surveys were distributed, raising the possibility that the members of these teams knew

the outcome of their evaluations and might have made self-serving attributions in their responses

to the survey (Miller & Ross, 1975). Specifically, team members who believed their team had

performed poorly might have been likely to report both that they had insufficient autonomy and

that they had solicited extensive knowledge inputs, in an effort to make themselves look as

though they had done all they could. The research design allowed me to test such biases,

however, by comparing 19 teams whose members all returned their surveys before their quality

evaluation outcome was announced with 37 teams whose members all returned their surveys at

least seven days after their evaluations had been completed. The tests revealed no significant

differences on the main variables, including team autonomy and knowledge embeddedness, or on

their correlations with strategic effectiveness or operational efficiency. There were also no

34

significant differences in the strategic effectiveness and operational efficiency ratings in these

two sets of teams. Attribution bias thus is not a convincing alternative to the benefits of

embedded autonomy as an explanation for the results of this study.

Implications for Theory and Research

This study extends understanding of self-managing teams by explicitly addressing the

relations between these teams and their environments. Prior research on self-managing teams has

mostly taken an internal perspective on the conditions under which autonomy is beneficial. Only

a few studies have considered the contexts in which self-managing teams work, for example by

examining the role of external leaders in coaching such teams (Manz & Sims, 1987; Wageman,

2001), the effects of organizational reward systems (Wageman, 1995), or corporate strategic

priorities (Zellmer-Bruhn & Gibson, 2006). The present study contributes to this effort by taking

an external perspective on project teams that draws attention to the risk that the independence

conferred by autonomy can lead them to become isolated from their environments. The weak

support of prior studies for the benefits of self-management in project teams thus may be

explained at least in part by the need for such teams to engage with their environments.

Supporting this claim, I found that project teams benefited more from autonomy if they avoided

isolation by seeking outside knowledge. In contrast, teams that isolated themselves from their

environments were harmed rather than helped by high autonomy. The implication is that theory

and research on self-managing project teams must look beyond the teams themselves to

understand how the desired advantages of autonomy can be achieved or derailed.

By demonstrating the importance of knowledge inflows for self-managing project teams,

this study also contributes to research on boundary spanning and intraorganizational networks

showing that teams have much to gain from activities such as “market scanning” and “technical

scouting” (e.g., Ancona & Caldwell, 1992; Hansen, 1999; Tushman, 1977). To this research, the

35

embedded autonomy perspective adds two insights. First, boundary spanning is particularly

critical for highly autonomous teams that otherwise risk isolation. And second, teams that engage

in boundary spanning will not benefit from it unless they can make independent decisions based

on the knowledge obtained; in the absence of such autonomy, these teams may be ineffectual or

even co-opted by outsiders with their own agendas and interests (Haas, 2006).

Additionally, my findings complement and extend theories of team empowerment and

high-performing organizations. These theories have noted that beyond autonomy, information is

one of several other key factor that can enhance empowerment and performance. For example, a

study of 1,150 directors of Fortune 1000 companies showed that boards particularly needed

power and information to be effective (Lawler et al, 2002), and a study of 198 Fortune 1000

firms found that team-enabling practices, including the use of self-managing teams, improved

product and service quality, while information-sharing practices improved customer service

(Gibson, Porath, Benson, & Lawler, 2007). However, prior research has not addressed how the

two key factors of autonomy and information interact, even in studies that have examined

interactions between other team design, process, and context variables (e.g. Janz et al., 1997). In

contrast, this study focuses in on these two key factors, demonstrating their importance for

project teams and offering theoretical explanations and empirical evidence for how they interact.

This study also advances the knowledge-based view of the firm by moving down to the

level where the work of the firm actually happens. Although there is now a substantial body of

research on knowledge flows between national subsidiaries in multinationals and business units

within firms more generally, the actors who exchange and utilize this knowledge are often teams

that work within and across these formal structures. Yet research on how teams obtain and use

knowledge in firms, particularly multinational organizations, is still relatively rare (for examples,

see Cummings, 2003; Hansen & Lovas, 2004). By focusing on project teams, this study

36

highlights a concern that can otherwise be easily overlooked in a knowledge-based view of the

firm: more knowledge flows are not necessarily better. Instead, the advantages of more

knowledge flows depend on the characteristics of the organizational actors involved. If these

actors are project teams, then understanding team-level characteristics such as team autonomy is

critical for understanding when and why knowledge delivers the strategic advantages proposed in

the knowledge-based view. In drilling down to the team level, this study thus increases the

capacity of the knowledge-based view to explain variance in performance outcomes within and

between organizations.

The focus on project teams as the unit of analysis also makes the embedded autonomy

perspective particularly relevant for advancing theory and research on multinational management

because it provides insight into the fundamental challenge of combining local differentiation

with global integration at the level of the task units that actually carry out the everyday work of

many multinationals. Subsidiary autonomy is a central concept in theories of multinational

organizational design (e.g. Birkinshaw, Hood, & Jonsson, 1998; Nohria & Ghoshal, 1997), but

extant theory and research offer little explanation of variations in the performance of critical

tasks within a subsidiary that has a given level of autonomy. Rather than focusing on subsidiary

autonomy, this study identifies team autonomy as a critical locus of differentiation that enables

appropriate decisions to be made and highlights the tension between independence and isolation

that exists for teams as well as subsidiaries. Additionally, rather than focusing on cross-

subsidiary knowledge flows, this study emphasizes that knowledge flows to teams are a critical

integrating mechanism. The embedded autonomy perspective thus addresses the macro-level

challenges of managing multinationals from the micro-level perspective of their project teams

and shifts attention toward the micro-level conditions that enhance team effectiveness.

Limitations and Future Directions

37

The growing role of specialist expertise in the global economy increasingly makes

autonomy critical for effective decision making in organizations while simultaneously making

isolation an unviable strategy (Child & McGrath, 2001). The multinational organization studied

here was an excellent context in which to investigate how project teams manage this tension

because their tasks demand both autonomy and knowledge embeddedness. This combination was

found to improve team performance across both performance metrics and project types. Still,

since this study was conducted in a single organizational setting, the extent to which the theory

and results hold in other settings should be further explored. For example, the benefits of

embedded autonomy may depend on organizational culture, since obtaining knowledge requires

investments in search and transfer that may be prohibitive in cultures that encourage hoarding

rather than sharing (Boisot, 1998). The embedded autonomy framework also would benefit from

validation in different national as well as organizational contexts (cf. Gibson, Zellmer-Bruhn, &

Schwab, 2003): for example, the extent of independence as well as isolation experienced by

autonomous teams may vary across national contexts, even within the same organization.

Although this study focused particularly on project teams in multinational organizational

settings, the theoretical framework of embedded autonomy is relevant and applicable to domestic

organizational settings too. However, whether the framework applies beyond the domain of

project teams engaged in knowledge work to other types of teams engaged in other types of work

remains a question. Much prior research on self-managing teams has focused on blue-collar

workers (Janz et al., 1997), whose labor-intensive tasks may not benefit as much from

knowledge from outside sources. Similarly, self-managing teams that are stable over time may

develop internal interaction patterns such as transactive memory systems that fundamentally

shape their effectiveness (Lewis, 2004). For such teams, an internal perspective may prove

sufficient for understanding the conditions under which autonomy improves performance.

38

Other questions raised by this study arise from the limitations of the data, which did not

allow for direct examination of internal team relations that could further illuminate the findings.

The relative impacts of internal and external relations on the success of self-managing project

teams might be explored to compare their explanatory power. Integrating these perspectives,

though, offers more interesting avenues to explore: for example, the benefits of embedded

autonomy may be amplified in teams with effective internal relations but reduced in teams with

ineffective internal relations, such as conflict among subgroups (cf. Gibson & Vermeulen, 2003)

or distrust among dispersed members (cf. Shapiro et al., 2002). Additionally, further unpacking

of the concepts of team autonomy and knowledge embeddedness could be worthwhile. Future

research could distinguish autonomy relative to different external stakeholders, for instance, to

establish whether the benefits of independence and the risks of isolation depend on who else is

involved in decision making. Likewise, knowledge embeddedness could be examined using

network methods to see whether different structural configurations of connections to individual

knowledge providers offer different informational benefits to teams (cf. Reagans & Zuckerman,

2001). Finally, following the precedent set by the wide-ranging embeddedness literature (e.g.,

Baum & Dutton, 1996; Zukin & DiMaggio, 1990), future research could extend beyond the

concept of knowledge embeddedness to examine the implications of other types of relations

between self-managing teams and their environments, such as social or political relations.

In conclusion, self-managing project teams in multinational organizations face a tension

between independence and isolation. The team-centered theory of knowledge work developed

here maintains that they can address this tension by combining decision-making autonomy with

knowledge embeddedness. As these teams strive to deliver projects that are strategically

effective and operationally efficient, embedded autonomy can help them to make decisions that

are well-informed as well as independent.

39

REFERENCES

Aiken, L. S., & West, S. G. 1991. Multiple regression: Testing and interpreting interactions.

Newbury Park, CA: Sage.

Ancona, D. G. 1987. Groups in organization: Extending laboratory models. In C. Hendrick (Ed.),

Annual review of personality and social psychology: group processes and intergroup

processes: 207-231. Palo Alto, CA: Annual Reviews.

Ancona, D. G., & Caldwell, D. F. 1992. Bridging the boundary: External activity and

performance in organizational teams. Administrative Science Quarterly, 37: 634-665.

Barker, J. R. 1993. Tightening the iron cage: Concertive control in self-managing teams.

Administrative Science Quarterly, 38: 408-437.

Barry, D. 1991. Managing the bossless team: Lessons in distributed leadership. Organizational

Dynamics, 20: 31-47.

Bartlett, C. A., & Ghoshal, S. 1989. Managing across borders: The transnational solution. Boston:

Harvard Business School Press.

Baum, J. A. C. & Dutton, J. E. (Eds.) 1996. Advances in strategic management: The embeddedness of

strategy, vol. 13. Greenwich, CT: JAI Press.

Birkinshaw, J., Ghoshal, S., Markides, C. C., Stopford, J. & Yip, G. (Eds.). 2003. The future of

the multinational company. New York: Wiley.

Birkinshaw, J., Hood, N., & Jonsson, S. 1998. Building firm-specific advantages in multinational

corporations: The role of subsidiary initiative. Strategic Management Journal, 19: 221-242.

Boisot, M. 1998. Knowledge assets: Securing competitive advantage in the information

economy. Oxford, U.K.: Oxford University Press.

Burt, R. S. 1992. Structural holes: The social structure of competition. Cambridge, MA:

Harvard University Press.

Child, J., & McGrath, R. G. 2001. Organizations unfettered: Organizational form in an

information-intensive economy. Academy of Management Journal, 44: 1135-1148.

Cohen, S. G., & Bailey, D. E. 1997. What makes teams work: Group effectiveness research from

the shop floor to the executive suite. Journal of Management, 23: 239-290.

Cohen, S. G., & Ledford, G. E. 1994. The effectiveness of self-managing teams: A quasi-

experiment. Human Relations, 47: 13-34.

Cohen, S. G., Ledford, G. E., & Spreizer, G. M. 1996. A predictive model of self-managing work

team effectiveness. Human Relations, 49: 643-676.

Collins, C. J., & Clark, K. D. 2003. Strategic human resource practices, top management team

social networks, and firm performance: The role of human resource practices in creating

organizational competitive performance. Academy of Management Journal, 46: 740-751.

40

Cordery, J. L, Mueller, W. S., & Smith, L. M. 1991. Attitudinal and behavioral effects of

autonomous group working: A longitudinal field study. Academy of Management Journal,

34: 464-476.

Criscuolo, P., Salter, A., & Sheehan, T. 2007. Making knowledge visible: Using expert yellow

pages to map capabilities in professional services firms. Research Policy, 36: 1603-1619.

Cummings, J. N. 2003. Work groups, structural diversity, and knowledge sharing. Management

Science, 50: 352-364.

Dacin, M. T., Ventresca, M. J., & Brent, B. D. 1999. The embeddedness of organizations:

Dialogue and directions. Journal of Management, 25: 31-356.

Earley, P. C., & Gibson, C. B. 2002. Multinational work teams: A new perspective. Mahwah,

NJ: Erlbaum.

Edmondson, A. C. 2002. The local and variegated nature of learning in organizations: A group-

level perspective. Organization Science, 13: 128-146.

Evans, P. 1995. Embedded autonomy: States and industrial transformation. Princeton, NJ:

Princeton University Press.

Gibson, C. B., & Vermeulen, F. 2003. A healthy divide: Subgroups as a stimulus for team

learning. Administrative Science Quarterly, 48: 202-239.

Gibson, C. B., Porath, C. L., Benson, G. S., & Lawler, E. E. 2007. What results when firms

implement practices: The differential relationship between specific practices, firm financial

performance, customer service, and quality. Journal of Applied Psychology, 92: 1467-1480.

Gibson, C. B., Zellmer-Bruhn, M., & Schwab, D. S. 2003. Team effectiveness in multinational

organizations: Development and evaluation across contexts. Group and Organization

Management, 28: 444-474.

Granovetter, M. 1985. Economic action and social structure: The problem of embeddedness.

American Journal of Sociology, 91: 481-510.

Grant, R. M. 1996. Prospering in dynamically competitive environments: Organizational

capability as knowledge integration. Organization Science, 7: 375-387.

Gupta, A.K., & Govindarajan, V. 2000. Knowledge flows within multinational corporations.

Strategic Management Journal, 21: 473-496.

Haas, M. R. 2006. Knowledge gathering, team capabilities, and project performance in

challenging work environments. Management Science, 52(8): 1170-1184.

Haas, M. R., & Hansen, M. T. 2005. When using knowledge can hurt performance: The value of

organizational capabilities in a management consulting company. Strategic Management

Journal, 26: 1-24.

Haas, M. R., & Hansen, M. T. 2007. Different knowledge, different benefits: A productivity

perspective on knowledge sharing in organizations. Strategic Management Journal, 28:

1133-1153.

41

Hackman, J. R. 1987. The design of work teams. In J. Lorsch (Ed.), Handbook of organizational

behavior. Englewood Cliffs, NJ: Prentice-Hall.

Hansen, M. T. 1999. The search-transfer problem: The role of weak ties in sharing knowledge

across organization subunits. Administrative Science Quarterly, 44: 82-111.

Hansen, M.T., & Haas, M. R. 2001. Competing for attention in knowledge markets: Electronic

document dissemination in a management consulting company. Administrative Science

Quarterly, 46: 1-28.

Hansen, M. T., & Lovas, B. 2004. How do multinational companies leverage technological

competencies? Moving from single to interdependent explanations. Strategic Management

Journal, 25: 801-822.

Hargadon, A., & Sutton, R. I. 1997. Technology brokering and innovation in a product

development firm. Administrative Science Quarterly, 42: 716-749.

Hobday, M. 2000. The project-based organisation: An ideal form for managing complex

products and systems? Research Policy, 29: 871-893.

Hyatt, D. E., & Ruddy, T. M. 1997. An examination of the relationship between work group

characteristics and performance: Once more into the breech. Personnel Psychology, 50: 553-585.

Janis, I. L. 1982. Groupthink (2nd ed). Boston: Houghton Mifflin.

Janz, B. D., Colquitt, J. A., & Noe, R. A. 1993. Knowledge worker team effectiveness: The role

of autonomy, interdependence, team development, and contextual support variables.

Personnel Psychology, 50: 877-904.

Kanter, R. M. 1988. When a thousand flowers bloom: Structural, collective, and social

conditions for innovation in an organization. In B. M. Staw & L. L. Cummings (Eds.),

Research in organizational behavior, vol. 10: 169-211. Greenwich, CT: JAI Press.

Katz, R., & Allen, T. J. 1982. Investigating the not invented here (NIH) syndrome: A look at the

performance, tenure, and communication patterns of 50 R&D project groups. R&D

Management, 12: 7-20.

Kean, T. H., & Hamilton, L. 2004. The 9/11 commission report: Final report of the national

commission on terrorist attacks upon the United States. New York: St. Martin’s Press.

Kirkman, B. L., & Rosen, B. 1999. Beyond self-management: Antecedents and consequences of

team empowerment. Academy of Management Journal, 42: 58-74.

Kirkman, B. L., & Shapiro, D. L. 2001. The impact of cultural values on job satisfaction and

organizational commitment in self-managing work teams: The mediating role of employee

resistance. Academy of Management Journal, 44: 557-569.

Kogut, B., & Zander, U. 1993. Knowledge of the firm and the evolutionary theory of the

multinational corporation. Journal of International Business Studies, 24: 625-645.

Kogut, B., & Zander, U. 1996. What firms do? Coordination, identity, and learning.

Organization Science, 7: 502-518.

42

Kostova, T. 1999. Transnational transfer of strategic organizational practices: A contextual

perspective. Academy of Management Journal, 24: 308-324.

Langfred, C. W. 2000. The paradox of self-management: Individual and group autonomy in

work groups. Journal of Organizational Behavior, 21: 563-585.

Langfred, C. W. 2004. Too much of a good thing? Negative effects of high trust and individual

autonomy in self-managing teams. Academy of Management Journal, 47: 385-399.

Langfred, C. W. 2005. Autonomy and performance in teams: The multilevel moderating effect of

task interdependence. Journal of Management, 31: 513-529.

Langfred, C. W. 2007. The downside of self-management: A longitudinal study of the effects of

conflict on trust, autonomy, and task interdependence in self-managing teams. Academy of

Management Journal, 50: 885-900.

Langfred, C. W., & Moye, N. A. 2004. Effects of task autonomy on performance: An extended

model considering motivational, informational, and structural mechanisms. Journal of

Applied Psychology, 89: 934-945.

Latham, G. P., Winters, D. C., & Locke, E. A. 1994. Cognitive and motivation effects of

participation: A mediator study. Journal of Organizational Behavior, 15: 49-63.

Lawler, E. J. 1992. Affective attachments to nested groups: A choice-process theory. American

Sociological Review, 57: 327-339.

Lawler, E. E., Finegold, D. L., Benson, G. S., & Conger, J. A. 2002. Corporate boards: Keys to

effectiveness. Organizational Dynamics, 30: 310-324.

Lawler, E. E., Mohrman, S. A., & Ledford, G. E. 1998. Creating high-performance

organizations: Practice and results of employee involvement, total quality management,

and reengineering programs in Fortune 1000 corporations. San Francisco: Jossey-Bass.

Levitt, B., & March, J. G. 1988. Organizational learning. Annual Review of Sociology, 14: 319-338.

Lewis, K. 2004. Knowledge and performance in knowledge-worker teams: A longitudinal study

of transactive memory systems. Management Science, 50: 1519-1533.

Lipnak, J., & Stamps, J. 1997. Virtual teams: Reaching across space, time, and organizations

with technology. New York: John Wiley & Sons, Inc.

Locke, E. A., Alavi, M., & Wagner, J. A. 1997. Participation in decision making: An

information exchange perspective. In G. R. Ferris (Ed.), Research in personnel and human

resources management, vol.15: 293-331. Greenwich, CT: JAI Press.

Long, J. S. 1997. Regression models for categorical and limited dependent variables. Thousand

Oaks, CA: Sage.

Lord, M. D., & Ranft, A. L. 2000. Organizational learning about new international markets:

Exploring the internal transfer of local market knowledge. Journal of International

Business Studies, 31: 573-589.

43

Manz, C. C., & Sims, H. P. 1987. Leading workers to lead themselves: The external leadership

of self-managing work teams. Administrative Science Quarterly, 32: 106-126.

Manz, C. C., & Sims, H. P. 1993. Business without bosses: How self-managing teams are

building high-performing companies. New York: Wiley.McEvily, B., & Zaheer, A. 1999.

Bridging ties: A source of firm heterogeneity in competitive capabilities. Strategic

Management Journal, 20: 1135-1156.

Miller, K. I., & Monge, P. R. 1986. Participation, satisfaction, and productivity: A meta-analytic

review. Academy of Management Journal, 29: 727-753.

Miller, S. T., & Ross, M. 1975. Self-serving biases in the attribution of causality: Fact or fiction?

Psychological Bulletin, 82: 93-118.

Monteiro, L. F., Arvidsson, N., & Birkinshaw, J. 2008. Knowledge flows within multinational

corporations: Explaining subsidiary isolation and its performance implications. Organization

Science, 19: 90-107.

Neter, J. W., Wasserman, S., & Kutner, M. H. 1985. Applied linear statistical models (2nd ed.).

Homewood, IL: Irwin.

Nohria, N., & Ghoshal, S. 1997. The differentiated network: Organizing multinational

corporations for value creation. San Francisco: Jossey-Bass.

Pearce, J. A., & Ravlin, E. C. 1987. The design and activation of self-regulating work groups.

Human Relations, 40: 751-782.

Prahalad, C.K., & Doz, Y. L. 1987. The multinational mission. New York: Free Press.

Reagans, R., & Zuckerman, E. Z. 2001. Networks, diversity, and productivity: The social capital

of corporate R&D teams. Organization Science, 12: 502-517.

Roth, K., & Kostova, T. 200. The use of the multinational corporation as a research context.

Journal of Management, 29: 883-902.

Rugman, A. M., & Verbeke, A. 2001. Subsidiary-specific advantages in multinational

enterprises. Strategic Management Journal, 22: 237-250.

Schulz, M. 2003. Pathways of relevance: Exploring inflows of knowledge into subunits of

multinational corporations. Organization Science, 14: 440-459.

Shapiro, D. L., Furst, S. A., Spreitzer, G. M., & Von Glinow, M. A. 2002. Transnational teams

in the electronic age: Are team identity and high performance at risk? Journal of

Organizational Behavior, 23: 455-467.

Shapiro, D. L., Von Glinow, M. A., & Cheng, J. L. (Eds.). Managing multinational teams: Global

perspectives. City: JAI/Elsevier Press.

Snow, C. C., Snell, S. A., Davison, S. C., & Hambrick, D.C. 1996. Use transnational teams to

globalize your company. Organizational Dynamics, 24: 50-67.

Szulanski, G. 1996. Exploring internal stickiness: Impediments to the transfer of best practice within

44

the firm. Strategic Management Journal, 17(winter special issue): 27-43.

Szulanski, G., Cappetta, R., & Jensen, R. J. When and how trustworthiness matters: Knowledge

transfer and the moderating effect of causal ambiguity. Organization Science, 15: 600-613.

Tsai, W. 2001. Knowledge transfer in intraorganizational networks: Effects of network position

and absorptive capacity on business unit innovation and performance. Academy of

Management Journal, 44: 996-1004.

Tsai, W. 2002. Social structure of “coopetition” within a multiunit organization: Coordination,

competition, and intraorganizational knowledge sharing. Organization Science, 13: 179-190.

Tushman, M. L. 1977. Communication across organizational boundaries: Special boundary roles

in the innovation process. Administrative Science Quarterly, 22: 581-606.

Tushman, M. L. 1978. Technical communication in R&D laboratories. The impact of project

work characteristics. Academy of Management Journal, 21: 624-645.

Tushman, M. L., & O’Reilly, C. A. 1996. Ambidextrous organizations: managing evolutionary

and revolutionary change. California Management Review, 38(4): 8-30.

Uhl-Bien, M., & Graen, G. B. 1998. Individual self-management: Analysis of professionals’ self-

managing activities in functional and cross-functional work teams. Academy of Management

Journal, 41: 340-350.

Uzzi, B. 1996. The sources and consequences of embeddedness for the economic performance of

organizations: The network effect. American Sociological Review, 61: 674-698.

Vaughan, D. 1990. Autonomy, interdependence, and social control: NASA and the space shuttle

Challenger. Administrative Science Quarterly, 35: 225-257.

Venkatraman, N., & Grant, J. H. 1986. Construct measurement in organizational strategy

research: A critique and proposal. Academy of Management Review, 11: 71-87.

Wageman, R. 1995. Interdependence and group effectiveness. Administrative Science

Quarterly, 40: 145-180.

Wageman, R. 2001. How leaders foster self-managing team effectiveness: Design choices vs.

hands-on coaching. Organization Science, 12: 559-577.

Wageman, R., Hackman, J.R., & Lehman, E. 2005. Team diagnostic survey: Development of an

instrument. Journal of Applied Behavioral Science, 41: 373-398.

Wellins, R. S., Byham, W.C., & Wilson, J. M.. 1991. Empowered teams: Creating self-

managing working groups and the improvement of productivity and participation. San

Francisco: Jossey-Bass.

Zellmer-Bruhn, M. E., & Gibson, C. B. 2006. Multinational organization context: Implications

for team learning and performance. Academy of Management Journal, 49: 501-518.

Zukin, S., & DiMaggio, P. 1990. Structures of capital: The social organization of the economy.

New York: Cambridge University Press.

45

FIGURE 1a

Interaction Effect of Team Autonomy and Knowledge Embeddedness on Strategic Effectiveness a

FIGURE 1b

Interaction Effect of Team Autonomy and Knowledge Embeddedness on Operational Efficiency a

a To illustrate the direction and magnitude of effects, low values were set at one standard deviation below the mean,

high values were set at one standard deviation above the mean, and the plots were constructed using OLS regression.

46

TABLE 1

Project Evaluation Questions for Strategic Effectiveness

Variable Questions

Strategic effectiveness

(financial projects)

To what extent are key development objectives addressed by the project? Clearly

linked to achieving strategic benchmarks? Clarity and realism of objectives?

Appropriateness of the approach? Approach adequately reflects lessons of

experience? Adequacy of knowledge and strategy underpinning the program?

Strength of client country ownership? Appropriateness and realism of program

conditions? Appropriate partnership with other key partners? Appropriateness and

realism of measures for achieving reforms?

Strategic effectiveness

(technical projects)

Appropriate objectives for the project clearly defined? Issues to be addressed clearly

identified? Objectives and issues adjusted in light of changing circumstances, if

necessary? Supports achieving one or more specific strategic objectives? Likely to

make a contribution to the organization’s stock of knowledge? Comparative

advantage of the organization in doing this work? Fit with other work being done

by the organization? Prospects for action on the issues addressed? Appropriate

audiences clearly defined? Expected impact and how to achieve it clearly defined?

TABLE 2

Survey Items for Multi-item Scales

Variable Survey Items

Team autonomy How was influence over [a list of 20 decisions about team objectives, resources,

design, and processes] distributed between the team itself (including the team leader)

and others outside the team (including senior managers, the client country, and the

development community)?*

Team embeddedness During the course of this project, how much relevant [i] country knowledge [ii]

technical knowledge did you get from (a) the country office (b) the rest of the

organization (c) the client country (d) the development community?*

Project complexity To what extent did the project require complex approaches and solutions?* To what

extent did the work depart from the usual work of a routine financial/technical

project?* To what extent did the techniques or skills or information needed for the

project change during the course of the project?*

Project novelty Prior to the project, how much relevant [i] country knowledge [ii] technical

knowledge did you personally have?*

Real team How clear was the membership of your team?* Was the team’s work performed

mainly by individuals or by the team as a whole?*

Team satisfaction Working together energized and uplifted members of this team.

There was a lot of unpleasantness among members of this team (reverse-coded).

Members of this team were getting better and better at working together.

The longer we worked together as a team, the less well we did (reverse-coded).

* Questions with customized five-point response scales. All other items used a scale ranging from 1, “strongly disagree,” to

5, “strongly agree.”

47

TABLE 3

Descriptive Statistics and Bivariate Correlations a

Variable Mean s.d. Min Max 1 2 3 4 5 6 7 8 9

1. Strategic effectiveness 1.68 0.62 1 3

2. Operational efficiency 1.89 0.65 1 3 .37

3. Project type 0.52 0.50 0 1 -.00 .22

4. Project duration 5.61 0.75 3.76 7.27 -.02 -.09 -.16

5. Project complexity 3.55 0.48 1.67 4.67 .07 .09 .16 .08

6. Project novelty 3.62 0.54 2.00 4.60 -.01 .02 -.10 .03 .17

7. Organizational tenure 8.46 4.40 1.00 25.50 -.06 .15 .12 -.14 .22 .00

8. Nonorganizational tenure 4.98 3.63 0.00 15.20 .14 -.05 .17 -.00 .12 .30 -.28

9. Late respondents 0.56 0.41 0.00 1.00 -.06 .02 .40 .11 -.09 -.14 .09 .08

10. Core respondents 0.76 0.22 0.17 1.00 .03 .17 .14 -.10 .16 .02 .14 -.08 .03

11. Real team 3.82 0.56 2.38 5.00 .09 .12 .19 -.10 -.00 .08 -.07 -.09 .05

12. Team size 8.48 4.15 2 23 .30 .10 .29 .00 .23 -.04 -.13 .21 -.07

13. Team location 0.80 0.40 0 1 .05 .08 .31 -.02 .19 -.14 .02 -.11 .02

14. Team satisfaction 4.04 0.37 2.75 4.75 -.01 .21 .07 -.16 .05 .14 .02 .03 -.10

15. Team autonomyb 3.62 0.31 2.75 4.32 .24 .05 .04 .03 .14 .11 .05 -.05 -.15

16. Knowledge embeddednessb 3.00 0.43 1.91 3.98 .13 .01 .00 -.02 .22 .19 -.02 .18 -.04

17. Knowledge embeddedness (country k.)b 3.08 0.47 1.88 4.17 .16 .06 .09 .01 .26 .19 -.04 .25 -.02

18. Knowledge embeddedness (technical k.)b 2.91 0.44 1.75 3.92 .05 -.02 -.06 -.05 .15 .19 -.00 .08 -.08

19. Knowledge embeddedness (intrafirm k.)b 2.83 0.44 1.81 3.75 .02 -.03 .06 -.00 .17 .21 .04 .09 .03

20. Knowledge embeddedness (extrafirm k.)b 3.15 0.54 1.67 4.30 .19 .03 -.05 -.03 .20 .12 -.06 .21 -.11

Variable 10 11 12 13 14 15 16 17 18 19

11. Real team .36

12. Team size -.10 -.23

13. Team location -.05 .19 .21

14. Team satisfaction .38 .50 -.19 .06

15. Team autonomy1 .03 .11 .02 -.04 .26

16. Knowledge embeddednessb .05 .15 .06 -.08 .26 .15

17. Knowledge embeddedness (country k.)b .01 .18 .08 .00 .27 .15 .94

18. Knowledge embeddedness (technical k.)b .08 .12 .04 -.14 .25 .12 .95 .80

19. Knowledge embeddedness (intrafirm k.)b .07 .14 .09 -.07 .16 .03 .84 .77 .82

20. Knowledge embeddedness (extrafirm k.)b .02 .11 .03 -.07 .28 .23 .90 .86 .83 .52

a n = 96. If r < .16, p < .10; if r > .20, p < .05; if r > .26, p < .01

b Variable was standardized in analyses by subtracting the mean and dividing by the standard deviation.

48

TABLE 4a. Ordinal Logit Analysis Results for Strategic Effectiveness (n=96) †

Model 1a Model 2a Model 3a Model 4a Model 5a Model 6a

Team autonomy a

0.93***

(0.27)

0.93***

(0.27)

1.00***

(0.29)

1.04***

(0.30)

0.91***

(0.29)

Knowledge embeddedness a

0.27

(0.23)

0.24

(0.24)

0.24

(0.26)

Knowledge embeddedness (country k.)a

0.64

(0.46)

Knowledge embeddedness (technical k.)a

-0.34

(0.45)

Knowledge embeddedness (intrafirm k.)a

-0.35

(0.28)

Knowledge embeddedness (extrafirm k.)a

0.61**

(0.30)

Knowledge embeddedness × Team autonomy

0.78**

(0.31)

Knowledge embeddedness (country k.) × Team autonomy 1.11**

(0.57)

Knowledge embeddedness (technical k.) × Team autonomy -0.19

(0.49)

Knowledge embeddedness (intrafirm k.) × Team autonomy

0.14

(0.31)

Knowledge embeddedness (extrafirm k.) × Team autonomy

0.65**

(0.29)

df 13 13 14 15 17 17

Log-likelihood -69.78 -76.03 -69.26 -65.64 -64.22 -63.95

LL χ2 ratio testb 13.84*** 1.34 14.90*** 22.13*** 24.87*** 25.51***

Pseudo-R2 0.18 0.11 0.19 0.23 0.25 0.25

TABLE 4b. Ordinal Logit Analysis Results for Operational Efficiency (n=96) †

Model 1b Model 2b Model 3b Model 4b Model 5b Model 6b

Team autonomya

-0.09

(0.23)

-0.08

(0.23)

-0.13

(0.24)

-0.15

(0.24)

-0.15

(0.24)

Knowledge embeddednessa

-0.11

(0.23)

-0.11

(0.23)

-0.20

(0.24)

Knowledge embeddedness (country k.)a

0.52

(0.43)

Knowledge embeddedness (technical k.)a

-0.61

(0.43)

Knowledge embeddedness (intrafirm k.)a

-0.33

(0.26)

Knowledge embeddedness (extrafirm k.)a

0.09

(0.28)

Knowledge embeddedness × Team autonomy

0.76***

(0.27)

Knowledge embeddedness (country k.) × Team autonomy 0.87*

(0.50)

Knowledge embeddedness (technical k.) × Team autonomy 0.02

(0.46)

Knowledge embeddedness (intrafirm k.) × Team autonomy

0.25

(0.27)

Knowledge embeddedness (extrafirm k.) × Team autonomy

0.59**

(0.26)

df 13 13 14 15 17 17

Log-likelihood -83.66 -83.61 -83.55 -79.12 -77.59 -78.52

LL χ2 ratio testb 0.13 0.23 0.35 9.21** 12.27** 10.41*

Pseudo-R2 0.07 0.07 0.07 0.12 0.14 0.13

•p < .10, ••p < .05, •••p < .01; two-tailed test for variable coefficients; † Standard errors are in parentheses; a Variable is standardized by subtracting the mean and dividing by the standard deviation;b Compared to baseline model with control variables only.

49

TABLE 5a

Ordinal Logit Analysis Results for Strategic Effectiveness in

High v. Low Novelty Projects and High v. Low Complexity Projects (n=96) †

Model 7a[i] Model 7a[ii] Model 8a[i] Model 8a[ii]

Median split criteria Project Novelty Project Complexity

Low High Low High

Team autonomy a 0.79**

(0.51)

1.99***

(0.72)

1.61**

(0.63)

3.05***

(0.97)

Knowledge embeddedness a 0.09

(0.39)

1.04*

(0.59)

0.61

(0.38)

-0.66

(0.59)

Knowledge embeddedness × Team autonomy 0.27

(0.50)

1.49**

(0.59)

0.76

(0.55)

1.32**

(0.65)

Number of observations 48 47 48 47

dfbb 14 14 14 14

Log likelihood -33.21 -25.45 -26.34 -20.48

LL χ2 ratio test c 4.80 23.58*** 10.09** 31.71***

Pseudo R2 0.14 0.43 0.33 0.55

TABLE 5b

Ordinal Logit Analysis Results for Operational Efficiency in

High v. Low Novelty Projects and High v. Low Complexity Projects (n=96) †

Model 7b[i] Model 7b[ii] Model 8b[i] Model 8b[ii]

Median split criteria Project Novelty Project Complexity

Low High Low High

Team autonomy a -0.59

(0.45)

0.32

(0.35)

-0.72

(0.44)

0.30

(0.40)

Knowledge embeddedness a -0.14

(0.40)

0.18

(0.39)

0.06

(0.35)

-0.69

(0.42)

Knowledge embeddedness * Team autonomy 0.67

(0.53)

0.62*

(0.38)

0.53

(0.47)

0.97**

(0.48)

Number of observations 47 46 47 46

Degrees of freedom b 14 14 14 14

Log-likelihood -29.18 -37.63 -35.86 -34.33

LL χ2 ratio test c 3.70 5.03 5.86 8.66**

Pseudo R2 0.29 0.21 0.21 0.23

•p < .10, ••p < .05, •••p < .01; two-tailed test for variable coefficients † n = 96. Standard errors are in parentheses

a Variable is standardized by subtracting the mean and dividing by the standard deviation b Models include all control variables except project novelty in Models 7a and 7b and project complexity in Models 8a and 8b c Compared to the same models including control variables only

50

1 Team autonomy neither precludes nor implies individual autonomy within a team; instead it

refers to the extent to which the team as a unit can make decisions without having its authority

superseded by external parties such as other work units, senior managers, or clients (Langfred,

2000; Manz & Sims, 1993; Uhl-Bien & Graen, 1998).

2 In supplementary analyses, I examined whether the results of my analyses were sensitive to

alternative specifications of the dependent variables, since relatively small numbers of projects

received the lowest ratings. First, I created alternative dichotomous measures of strategic

effectiveness and operational efficiency that were coded 1 if a project was rated “highly

satisfactory” or 0 if the project was rated either “satisfactory” or “marginal/unsatisfactory.” The

results using these alternative dichotomous measures were the same as the results for the original

categorical measures for Hypothesis 1, and substantively similar though weaker for Hypotheses

2 and 3. Second, because the relatively small number of projects that received the lowest rating

was a particular concern for the strategic effectiveness dependent variable, I created an

alternative continuous measure of strategic effectiveness by summing a project’s scores on the

ten questions that the expert panels used to inform their overall strategic effectiveness ratings (α

= .89 for financial projects, α = .82 for technical projects). This continuous measure was highly

correlated with the categorical measure of strategic effectiveness (r = .88), but not identical to it

because, as noted, the experts took into account their full understanding of a project and its

challenges as well as the scores on the questions. Results based on the alternative continuous

measure with finer variation were the same as results with the original categorical measure for

Hypotheses 1, 2, and 3. Having established that the results were robust to these alternative

specifications of the dependent variables, I retain and report the original categorical measures

because the project evaluation unit that generated the project ratings viewed the three categories

as meaningfully different from each other and the categorical ratings as more accurate

assessments than the continuous ratings. 3 I conducted one-tailed t-tests to see whether there were statistically significant differences

between the 50 financial projects and the 46 technical projects on the variables in this study. The

results showed only that some of the control variables differed by project type, notably the real

team, team size, late respondents, and project duration variables. To address these differences, I

multiplied each of these variables by the project type variable and included the interaction terms

as additional control variables. However, they did not change the results, and so these additional

variables were not included in the models presented here.

THE DOUBLE-EDGED SWORDS OF AUTONOMY ANDEXTERNAL KNOWLEDGE: ANALYZING TEAM

EFFECTIVENESS IN A MULTINATIONAL ORGANIZATION

MARTINE R. HAASUniversity of Pennsylvania

Extending the differentiation-integration view of organizational design to teams, Ipropose that self-managing teams engaged in knowledge-intensive work can performmore effectively by combining autonomy and external knowledge to capture thebenefits of each while offsetting their risks. The complementarity between havingautonomy and using external knowledge is contingent, however, on characteristics ofthe knowledge and the task involved. To test the hypotheses, I examined the strategicand operational effectiveness of 96 teams in a large multinational organization. Find-ings provide support for the theoretical model and offer implications for research onteam ambidexterity and multinational management as well as team effectiveness.

In many organizations, self-managing teams con-duct knowledge-intensive work such as designingnew products, developing innovative technologies,and delivering professional services to clients (e.g.,Hackman, 2002; Manz & Sims, 1993; Mohrman,Cohen, & Mohrman, 1995). The extensive researchon team effectiveness offers two seemingly unre-lated perspectives with useful but controversial in-sights for such teams. Research on team self-management has presented the argument that au-tonomy, in the form of collective control over crit-ical task-related decisions, can enable teams to per-form more effectively (e.g., Cohen & Ledford, 1994;Langfred, 2000; Wellins, Byham, & Wilson, 1991).However, the empirical evidence for this view hasbeen mixed: autonomy appears to be advantageousfor stable, full-time work teams but is not necessar-ily so for temporary project teams, which are com-mon in knowledge-intensive work settings (Cohen& Bailey, 1997). In parallel, research on team“boundary spanning” has drawn attention to theimportance of teams’ external rather than internalinteractions (e.g., Gladstein, 1984). Such studieshave suggested that teams can perform more effec-tively if they obtain and use external knowledge, in

the form of task-related information, know-how,and feedback from sources outside the teams (e.g.,Ancona & Caldwell, 1992; Hansen, 1999; Reagans,Zuckerman, & McEvily, 2004). Yet external knowl-edge does not always help teams to perform moreeffectively, even in knowledge-intensive work set-tings, and it sometimes hurts (e.g., Cummings,2004; Haas & Hansen, 2005). These twin tensions inthe team effectiveness literature suggest that al-though having autonomy and gaining externalknowledge both have possible benefits for teams,both also expose teams to risks that can preventrealization of their full potential. For self-managingteams engaged in knowledge-intensive work, thequestion that arises is, When do the benefits ofthese potentially favorable conditions outweigh therisks?

To address this question, I integrate the separateperspectives on team effectiveness offered in theresearch on team self-management and teamboundary spanning by applying the differentiation-integration view of organizational design to teams.At the organization level, differentiation refers tothe extent to which business units adapt their ac-tivities to their own environments; integration re-fers to the extent to which they coordinate theiractivities with each other (Lawrence & Lorsch,1967). Empirically, business units are often viewedas more differentiated if they have more decision-making autonomy (e.g., Birkinshaw, Hood, & Jons-son, 1998; Garnier, 1982), and as more integrated ifthey obtain and use more knowledge from otherunits (e.g., Gupta & Govindarajan, 2000; Tsai,2001). A fundamental principle of organizationaldesign is that differentiation and integration arecomplementary, so that firms perform more suc-

I gratefully acknowledge the valuable contributions ofAssociate Editor Wenpin Tsai and the anonymous re-viewers for this article. I also thank the managers andteam members who participated in this study, and JulianBirkinshaw, Stuart Bunderson, Chris Collins, Amy Ed-mondson, Mauro Guillen, Richard Hackman, MortenHansen, Dave Harrison, Witold Henisz, Tatiana Kostova,Jay Lorsch, Beta Mannix, Mitchell Orenstein, Madan Pil-lutla, Phanish Puranam, Nancy Rothbard, and Freek Ver-meulen for their helpful inputs on earlier versions.

� Academy of Management Journal2010, Vol. 53, No. 5, 989–1008.

989

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download or email articles for individual use only.

cessfully if highly differentiated business units arealso highly integrated (Lawrence & Lorsch, 1967;Nohria & Ghoshal, 1997). Applying this comple-mentarity principle to teams, and in keeping withmy earlier work (Haas, 2006), I propose that com-bining the two conditions of autonomy and exter-nal knowledge use can increase team effectivenessmore than either alone, because the benefits of eachoffset the other’s risks.

A second fundamental principle of organization-al design, however, is that the optimal conditionsfor business unit performance are contingent onsituation (Lawrence & Lorsch, 1967). A similar con-tingency principle can be expected to apply at theteam level. To identify when the complementaritybetween autonomy and external knowledge is ad-vantageous for self-managing teams engaged inknowledge-intensive work—and when it is not—Idraw on organization design research to identifytwo sets of contingencies with particular relevancefor such teams: knowledge-based contingencies(e.g., Birkinshaw, Nobel, & Ridderstrale, 2002) andtask-based contingencies (e.g., Galbraith, 1973;Tushman, 1979).1 By focusing attention on the con-tingent complementarity of autonomy and externalknowledge use, applying organizational designprinciples to teams helps resolve the tensions inprior research on team effectiveness.

The setting in which I tested hypotheses is amultinational organization whose teams operateworldwide. Examining teams in multinational or-ganizations is useful for extending research onteam effectiveness because the complexity of suchsettings highlights the importance of both differen-tiation and integration at the team level (cf. Roth &Kostova, 2003). A multinational organization cancreate value by combining autonomous subsidiar-ies with cross-subsidiary flows of knowledge toform a “differentiated network” (Nohria & Ghoshal,1997). Similarly, its teams may perform more effec-tively if they combine autonomy with use of exter-nal knowledge. Additionally, many multinationalsrely on teams to carry out much of their work,making team effectiveness an important issue formultinational management research (e.g., Earley &Gibson, 2002). I examine two dimensions of teameffectiveness that were viewed as critical in theorganization studied here, as they are in many mul-tinationals (cf. Gibson, Zellmer-Bruhn, & Schwab,2003): strategic effectiveness, which refers to the

extent to which a team delivered project outputsthat furthered the organization’s strategic goals,and operational effectiveness, which refers to howappropriately a team utilized available resources indelivering project outputs.

DIFFERENTIATION AND INTEGRATIONIN TEAMS

The Double-Edged Sword of Team Autonomy:Independence and Isolation

Theories of team self-management suggest thatautonomy motivates teams to make independentdecisions that serve the best interests of their tasks,by giving them a greater sense of responsibility andaccountability for their work (e.g., Cohen & Led-ford, 1994; Cordery, Mueller, & Smith, 1991; Janz,Colquitt, & Noe, 1997) and signalling managementendorsement (Langfred, 2000). Team autonomyalso allows those closest to tasks to make criticaltask decisions (Hackman, 2002) without having tocompromise to secure support from parties withtheir own agendas, such as senior managers orpowerful clients (cf. Ancona & Caldwell, 1992).

Nevertheless, empirical studies have not pro-vided compelling evidence that autonomy im-proves the performance of teams that work onproject-based tasks (Cohen & Bailey, 1997). A pos-sible explanation that has been underexplored instudies of team self-management is that autono-mous teams may become isolated from their envi-ronments, to the detriment of their performance onsuch tasks. Organizational design theorists havelong recognized that delegating autonomy to busi-ness units is risky because units that are able tomake decisions without external input may over-look or resist courses of action that are preferable totheir organization as a whole (Galbraith, 1973; Law-rence & Lorsch, 1967). Research on “groupthink”has suggested that a similar dynamic may impedethe effectiveness of small groups of decision mak-ers with high autonomy (Janis, 1982). For autono-mous teams in contemporary organizations, more-over, project-based tasks can create a sense of “timefamine” that leads them to believe that time spentsoliciting input from outsiders is wasted (Perlow,1999). The status conveyed to a team by the en-dorsement of autonomy also encourages the “not-invented-here” syndrome, characterized by unwill-ingness to adopt ideas from outside the team (Katz& Allen, 1982).

The resulting isolation has implications for bothstrategic and operational effectiveness: teams maymiss opportunities to learn about options for serv-ing their organization’s strategic goals, or they maywaste time replicating solutions that could have

1 Other contingencies commonly studied in organiza-tional design research include firm technology and envi-ronmental dynamism (Lawrence, 1993), but these are lessrelevant at the level of teams.

990 OctoberAcademy of Management Journal

been more efficiently imported from outside. Theconsequences for new-product development tasks,for example, may be less creativity (Hargadon &Sutton, 1997) or slower time-to-market (Hansen,1999). Because such risks of isolation reduce thebenefits of independence, more autonomy does notnecessarily result in better team performance.

The Double-Edged Sword of External Knowledge:Information and Influence

In contrast to research on team self-management,which has emphasized the benefits of autonomywithout directly addressing teams’ interactionswith their environments, research on team bound-ary spanning has tended to take an external per-spective, highlighting the benefits of knowledgefrom the teams’ environment, including expert net-works and document repositories both inside andoutside their organization (e.g., Ancona & Caldwell,1992; Collins & Clark, 2003; Reagans et al., 2004).According to this research, obtaining and usingexternal knowledge can improve task outcomes byhelping teams make more informed decisions. Forexample, soliciting advice from experts to help wincompetitive bids for new client contracts can en-hance strategic effectiveness (Haas & Hansen,2005), and transferring codified “best practices”developed elsewhere to help teams benchmark andimprove their work processes can increase opera-tional effectiveness (Szulanski, 1996).

Such studies have paid little attention, however,to the reality that external knowledge is often asource of influence as well as information (Pfeffer,1981; Spekman, 1979). Many organizations are con-tested terrain (Edwards, 1979), characterized bycompeting coalitions (Cyert & March, 1963), “turfwars” (Brown, Lawrence, & Robinson, 2005), andentrenched commitments to particular ideologicalperspectives (Carlile, 2002). In such environments,actors often hold conflicting views about what in-puts are appropriate and important for a task andhow they should be used (Pettigrew, 1973). Conse-quently, knowledge providers may attempt to in-fluence teams that seek inputs from them, throughdirect demands for support of their agendas, selec-tive presentation of information, or subtle empha-sis on favored solutions (Feldman, 1988).

Although influence attempts do not always ac-company external knowledge and sometimes sen-sitize teams to important concerns, problems oftenarise. For example, teams may expend valuabletime and energy managing organizational politics(cf. Ancona & Caldwell, 1992); damaging conflictsmay arise between team members who advocatecompeting views as a result of pressures from out-

siders (cf. Jehn, 1995); or outsiders may co-opt aproject to forward their own agendas (cf. Selznick,1949). Because external knowledge poses influencerisks as well as offering information benefits, ob-taining more such knowledge does not necessarilyimprove team performance.

Complementarity and Contingencies

If autonomy and external knowledge each conveysrisks as well as benefits to teams, how can the risks beminimized? Recognizing that the differentiation-inte-gration view of organizational design has addressed aparallel question at the business unit level, in thehypotheses that follow I apply its principles ofcomplementarity and contingency to teams. The fulltheoretical model is shown in Figure 1.

According to research on organizational design,business units require substantial decision-makingautonomy if they are to develop and deliver locallydifferentiated products or services; in multina-tional organizations, for example, national subsid-iaries need autonomy to ensure local responsive-ness (e.g., Bartlett & Ghoshal, 1989; Birkinshawet al., 1998; Garnier, 1982). The risk of delegatingauthority to a business unit, however, is that it maybecome cut off from the rest of an organization(Lawrence & Lorsch, 1967; Nohria & Ghoshal,1997). To mitigate this risk, cross-unit integrationmechanisms are necessary to facilitate coordina-tion and promote learning (e.g., Martinez & Jarillo,1989). In particular, researchers have emphasizedthat cross-unit flows of knowledge serve as a criti-cal integration mechanism (e.g., Gupta & Govin-darajan, 2000; Kogut & Zander, 1993; Tsai, 2001).By enabling business units to combine differentia-tion with integration, autonomy and externalknowledge use thus can serve as complementaryconditions for effective performance.

Classic organizational design research focuses onbusiness units as the loci of differentiation andintegration, but the work of many contemporaryorganizations is carried out by project teams thatoperate within or across business units. For theseteams, balancing differentiation and integration is ateam-level challenge, rather than one that is re-solved for them at the business-unit level. Apply-ing the complementarity principle to such teamssuggests that they can perform more effectively bycombining autonomy with use of external knowl-edge. These conditions are complementary becausethe benefits of each condition offset the risks of theother. The information benefits provided by exter-nal knowledge reduce the isolation risks created byautonomy: autonomous teams are less isolated ifthey obtain more external knowledge. The inde-

2010 991Haas

pendence benefits of autonomy reduce the influ-ence risks created by external knowledge: teamsthat obtain external knowledge are less vulnerableto influence if they are more autonomous. Conse-quently, teams with high levels of both autonomyand external knowledge can make decisions thatare both independent and well-informed. In con-trast, teams that have high levels of autonomy butlack external knowledge can suffer from insuffi-cient information, while teams that use high levelsof external knowledge but lack autonomy can sufferfrom insufficient independence. Stated formally,autonomy and the use of external knowledge mod-erate each others’ effects on team performance, insuch a way that the effect of either condition ismore positive if the other is also present. Hence:

Hypothesis 1. The effects of autonomy and ex-ternal knowledge on team performance inter-act positively.

Although the complementarity principle high-lights the mutual advantages of autonomy and ex-ternal knowledge, the contingency principle of or-ganizational design is a recognition that high levelsof differentiation and integration may not be opti-mal in all situations (Lawrence & Lorsch, 1967). Forexample, the characteristics of a business unit’sknowledge can affect the level of autonomy it needsto carry out R&D activities (Birkinshaw et al., 2002).Additionally, the characteristics of its tasks can

influence its external knowledge needs, because ofthe associated information-processing demands (Gal-braith, 1973; Tushman, 1979). Such knowledge andtask contingencies at the business unit level suggestthat the needs for autonomy and external knowledgemay be similarly contingent at the team level.

For self-managing teams engaged in knowledge-intensive work, the complementarity between au-tonomy and external knowledge depends onwhether using external knowledge creates influ-ence risks and having autonomy creates isolationrisks. If external knowledge does not create influ-ence risks, autonomy is not needed; if autonomydoes not create isolation risks, external knowledgeis not needed. When these risks are high, in con-trast, the potential for complementarity exists. Inthe four hypotheses that follow, I identify knowl-edge and task contingencies that increase the risksof influence or isolation for teams, resulting in astronger positive interaction between autonomyand external knowledge. Hypotheses 2 and 3 ad-dress situations in which the risks of influence aregreater because of characteristics of the content orsource of knowledge. Hypotheses 4 and 5 addresssituations in which the risks of isolation are greaterbecause of characteristics of a team’s task.

Knowledge contingencies. The risks of influenceassociated with external knowledge are greaterwhen the content of that knowledge is scarce thanthey are when it is common, because teams are

FIGURE 1Theoretical Modela

a The shaded area shows causal mechanisms. These were not observed.

992 OctoberAcademy of Management Journal

more dependent on the providers of scarce knowl-edge content (Pfeffer & Salancik, 1978). They can-not as easily avoid influence attempts by switchingto another provider (cf. Haunschild & Beckman,1998), nor can they as readily judge the accuracy ortrustworthiness of the content since little basis forcomparison exists (cf. Szulanski, Cappetta, &Jensen, 2004). Since the providers of scarce knowl-edge thus have more power to push their own agen-das, strategic effectiveness is more vulnerable tothe risks of external influence. Operational effec-tiveness is also endangered as these knowledgeproviders have more power to demand that teamsspend time addressing their demands. Autonomyenables teams to act more independently in the faceof such pressures, however, since they can resistefforts by knowledge providers to impose inappro-priate agendas on the teams’ projects and avoidextensive debate over every decision. Because theindependence benefits of autonomy are more criti-cal when the influence risks of external knowledgeare higher:

Hypothesis 2. The effects of autonomy and ex-ternal knowledge on team performance inter-act more positively when the knowledge con-tent is scarce rather than common.

The influence risks associated with externalknowledge are also greater when the knowledgesources are nonorganizational (i.e., based outsidethe organization in which a team works) rather thanorganizational (i.e., based inside the organization),because the agendas of organizational outsiders areless likely to be aligned with the goals of the teamthan those of insiders (Barnard, 1938). Althoughagendas often diverge within organizations (Cyert &March, 1963), the superordinate identity, sharedinterests, and social norms created by organizationmembership typically increase cooperation amongmembers (Schein, 1992). Consequently, efforts topush projects in directions that are suboptimal forthe team’s strategic effectiveness, or to impose time-consuming demands that impede its operational ef-fectiveness, are less likely to occur when the pro-viders of external knowledge are organizationalinsiders rather than outsiders. Also, team memberstend to value knowledge from outsiders more thanthat from insiders (Menon & Pfeffer, 2003), increas-ing their vulnerability to outsiders’ agendas. Sinceknowledge from nonorganizational sources carriesgreater influence risks, the independence benefitsof autonomy are more valuable when teams obtainsuch knowledge:

Hypothesis 3. The effects of autonomy and ex-ternal knowledge on team performance inter-

act more positively when the knowledgesources are nonorganizational rather thanorganizational.

Task contingencies. The isolation risks createdby autonomy are greater for more uncertain tasks,which are characterized by higher novelty or com-plexity (Galbraith, 1973; Tushman, 1979). Themore novel a task for the team members involved init, the more isolation endangers their strategic ef-fectiveness, because they have little experiencewith similar tasks to exploit as they develop, select,and pursue strategic options. Isolation also poses agreater threat to a team’s operational effectivenesswhen the task is more novel, because the team’smembers must rely on inefficient trial-and-errorprocesses that often involve making and rectifyingmistakes (cf. Levitt & March, 1988). The more com-plex the task, similarly, the more teams endangertheir strategic effectiveness by not soliciting inputsthat can help them to anticipate, identify, and ad-dress potential problems or unexpected conse-quences of their decisions (Tushman, 1978). Isola-tion also impedes operational effectiveness morefor complex tasks because it takes longer to developsolutions for such tasks, whereas importing exist-ing solutions can reduce the opportunity costs ofreinventing them (Hobday, 2000). Because externalknowledge offers information benefits that can off-set the heightened isolation risks associated withautonomy in situations of high task novelty or com-plexity, the complementarity between autonomyand external knowledge will be greater in suchsituations:

Hypothesis 4. The effects of autonomy and ex-ternal knowledge on team performance inter-act more positively when task uncertainty ishigh rather than low.

Finally, the extent of this complementarity alsodepends on task pressures. For self-managingteams engaged in knowledge-intensive work, taskpressures commonly take two forms: time pressureand client pressure. Time pressure tends to begreater when teams work on tasks of shorter dura-tion, owing to tight deadlines (Perlow, 1999).Shorter task duration increases the risks of isola-tion because autonomous teams face more tempta-tion to make critical task decisions swiftly andavoid spending valuable time consulting with out-siders. Client pressure tends to be greater whenteams work on tasks for larger clients, as these areoften very powerful stakeholders in a project (Mint-zberg, 1983). Because autonomous teams may dis-appoint or alienate these powerful stakeholders ifthey make decisions without sufficient external

2010 993Haas

consultation, the risks of isolation are higher. Ulti-mately, both strategic and operational effectivenesscan suffer in situations characterized by high taskpressure, if autonomous teams fail to take advan-tage of external knowledge that could improve thequality of their work or save them time (Haas &Hansen, 2007). Since the information benefits ofexternal knowledge are more critical when the iso-lation risks of autonomy are higher:

Hypothesis 5. The effects of autonomy and ex-ternal knowledge on team performance inter-act more positively when task pressures arehigh rather than low.

DATA AND METHODS

Research Setting

I tested the hypotheses using quantitative datacollected in a field study at a multinational organ-ization with more than 10,000 employees and 100offices worldwide. To develop an understanding ofthe research setting prior to collecting these data, Iconducted semistructured interviews lasting one tothree hours each with 50 team members involvedin projects around the world, as well as with 20managers and staff responsible for project evalua-tion, strategy and change management, knowledgemanagement, and human resources. I systemati-cally reviewed my interview notes as well as inter-nal memos and project documents to gain insightinto the nature of the work in this organization, andto prepare and refine the survey instrument.

The organization is a prominent international de-velopment agency whose clients are national andregional governments in developing countries. Istudied financial teams that designed large-scaleinvestment programs and technical teams that pro-vided high-level research and analysis on develop-ment issues for these clients. The main outputs forboth types of teams were detailed reports document-ing their recommendations. The teams typically in-cluded economists and technical specialists in fieldssuch as public finance, infrastructure, and engineer-ing. In the study sample described below, the averageteam included 8.5 members, each of whom spentfrom one month to four years with that team whilealso working on two to ten projects with other teams.Their average age was 44 years, and 66 percent weremen. The teams in the sample conducted projects inAfrica, Central Asia, East Asia, Europe, Latin Amer-ica, the Middle East, and South Asia.

Although all the teams were self-managing tosome extent, their levels of decision-making auton-omy varied according to factors that included thedistribution of informal as well as formal authority

in the organization, the status of the team members,and the style of the senior managers to whom ateam was accountable. Because the work wasknowledge-intensive, requiring high levels of ex-pertise and experience, the organization had in-vested in expert directories, help desks, and docu-ment databases to help teams access knowledge fortheir projects. Many team members recognized thattaking advantage of such resources both inside andoutside the organization could be useful, yet teamsvaried in the extent to which they obtained andused external knowledge in their work.

Quantitative Data

I collected quantitative data from three indepen-dent sources: the organization’s project evaluationunit, a team member survey, and archival projectrecords. To evaluate projects, the organization hadestablished a unit of 20 full-time staff who drew arandom sample of financial and technical projectsfrom the full population of projects completed eachyear. This project evaluation unit then assembled acustomized panel of experts to assess each selectedproject. Each panel included at least two respectedexperts in the project’s area with no prior connec-tions to the project.

The panels reviewed the project documents, inter-viewed the team leader, and completed a detailedevaluation protocol. Although a different expertpanel evaluated each project, the project evaluationunit took care to ensure that the ratings based on theseinputs were robust across panels: in addition to pro-viding detailed guidance to the panels during evalu-ations, the unit regularly tested the interpanel reli-ability of the ratings to confirm that different panelswere highly likely to rate the same project similarly.2

The project evaluation unit provided the ratings forthe 120 teams sampled in the year of this study (60financial and 60 technical teams).

I obtained official team rosters from the firm’sdatabases and asked the team leaders to distinguishbetween those they defined as core members andthose they defined as noncore members. After pre-testing the survey questions with 52 members ofteams that were not part of the evaluated sample, Isent surveys to all 1,021 core and noncore membersof the 120 teams (including the leaders, who werecore members) as soon as possible after theirprojects were selected for evaluation. The respon-dents were directed to focus on the project underevaluation, as identified on the front page of the sur-

2 Interrater reliability within panels was not a concernbecause the panelists evaluated their projects jointly.

994 OctoberAcademy of Management Journal

vey, but they were anonymous within teams exceptfor one question that identified them as core or non-core members. Only 18 of the team members sur-veyed appeared on more than one team roster, indi-cating that respondents who participated in morethan one team were unlikely to bias the data. I re-ceived survey responses from 550 team members(54%). After excluding teams for which fewer than 50percent of the respondents were core members (Hack-man, 2002: 47), 96 teams qualified for the study (80percent; 50 financial and 46 technical teams, with485 member respondents). Tests for selection biasshowed no significant differences in the effectivenessratings, project types, regions, or divisions of the 24teams that did not qualify.

Dependent variables. The dependent variablesin this study were (1) the strategic effectiveness of ateam and (2) the operational effectiveness of a team,as rated by its independent expert panel. These twodimensions of team effectiveness were viewed askey indicators of performance in this organization,as their inclusion in the project evaluation processattested. Both variables used ordinal scales onwhich 3 was “highly satisfactory,” 2 was “satisfac-tory,” and 1 was “marginal or unsatisfactory.”These scales were based on criteria developed bythe project evaluation unit through a multiyear pro-cess of consultation and refinement. To evaluatestrategic effectiveness, each expert panel used a setof ten questions.3 To evaluate operational effective-ness, each expert panel assessed the appropriate-ness of the time taken for the project, its budget,

and the other resources used, particularly skill mix,in light of the nature and context of the project. Toarrive at overall effectiveness ratings, the panelstook into account their full understanding ofprojects and their distinctive challenges as well asthe scores on these questions. Of the 96 teams inthe data set, 41 percent received a rating of 3, 51percent received a 2, and 8 percent received a 1 onstrategic effectiveness; and 27 percent received arating of 3, 57 percent received a 2, and 16 percentreceived a 1 on operational effectiveness.4

Team autonomy. Hackman (1987, 2002) devel-oped an authority matrix that identifies four levelsof team self-management based on the extent towhich teams have control over the critical deci-sions related to their tasks. Working from this ma-trix, I examined four categories of critical task-related decisions that contribute to team autonomy:(1) managing work processes, (2) managing the de-sign of the task or team, (3) managing resources,and (4) managing the objectives of the task or team.Within each of these four categories, I drew on myinterview data to identify 5 specific decisions thatwere critical in this organization. The resulting 20decisions focused on (1) setting up and managingsite visits, interactions with clients and manage-ment, handling conflict; (2) project pacing, feed-back solicitation, quality standards, staffing re-quirements, selection of team members; (3) budgetsize, additional funding, information inputs, teamtraining, team rewards; and (4) project initiation,overall priority, boundaries/scope, specific compo-

3 For financial projects, the ten questions were: “Towhat extent does the project . . . Address key develop-ment objectives? Clearly link to achieving strategicbenchmarks? Demonstrate clarity and realism of objec-tives? Establish appropriateness of approach? Ade-quately reflect lessons of experience? Show adequacy ofknowledge and strategy underpinning the program? In-volve strong client country ownership? Provide appro-priate and realistic program conditions? Establish appro-priate partnership with other key partners? Provideappropriate and realistic measures for achieving re-forms?”

For technical projects, the ten questions were: “Towhat extent does the project . . . Clearly define appropri-ate objectives? Clearly identify issues to be addressed?Enable objectives and issues to be adjusted in light ofchanging circumstances, if necessary? Support achievingone or more specific strategic objectives? Make a likelycontribution to the organization’s stock of knowledge?Fit with the organization’s comparative advantage? Fitwith other work being done by the organization? Haveprospects for actions on the issues addressed? Clearlydefine appropriate audiences? Clearly define expectedimpact and how to achieve it?”

4 In supplementary analyses, I examined whether myfindings were sensitive to alternative specifications ofthe dependent variables, since relatively small numbersof projects received the lowest ratings. First, I createdalternative dichotomous measures coded 1 if a projectwas rated “highly satisfactory” or 0 if the project wasrated “satisfactory” or “marginal/unsatisfactory”; the re-sults were substantively the same. Second, because therelatively small number of projects that received the low-est rating was a particular concern for the strategic effec-tiveness measure, I created an alternative continuousmeasure by summing a project’s scores on the ten ques-tions that the expert panels used to inform their overallratings of strategic effectiveness (� � .89 for financialprojects, � � .82 for technical projects). This continuousmeasure was highly correlated with but not identical tothe categorical measure (r � .88); the results were thesame. I also examined whether excluding specific itemsfrom this continuous measure changed the results, but itdid not. Having established the robustness of my find-ings, I report the results using the original categoricalmeasures because the project evaluation unit viewedthese as best capturing meaningful differences betweenprojects.

2010 995Haas

nents, level of innovation. In the survey, each teammember was asked about his or her team’s level ofautonomy over all 20 decisions, as follows: “Howwas influence over [decision] distributed betweenthe team itself (including the leader) and othersoutside the team (including managers, the clientcountry, and the development community)?” (ratedon a scale from 1, “The team had very little influ-ence,” to 5, “The team had almost all the influ-ence”). To construct the team autonomy variable, Iaveraged the responses to these 20 items for eachteam member and then within teams (� � .90,ICC �.05, p � .10, rwg � .85).

External knowledge. My interviews indicatedthat team members in this organization typicallyclassified the sources from which they obtainedexternal knowledge into four categories. Two ofthese sources were inside the organization’s bound-aries: (1) the country office and (2) the rest of theorganization. The country office referred to the or-ganization’s local office in the client country; therest of the organization referred to its other globaloffices. Two of the sources were outside the organ-ization’s boundaries: (3) the client country and (4)the global community. The client country referredto the national or regional government and otherlocal stakeholders such as nongovernmental agen-cies (NGOs) and businesses; the global communityreferred to international NGOs, think tanks, aca-demics, and others working on development issuesglobally.

For each of these four sources, I asked the teammembers, “During the course of this project, howmuch relevant (a) technical knowledge (b) countryknowledge did you gather from [this source]?” (1,“very little”; 5, “a lot”).5 Technical knowledge wasdefined as “knowledge about the technical aspectsof the work—the professional skills, competencies,and expertise relevant to the project.” Countryknowledge was defined as “knowledge about thelocal environment—the country-specific condi-tions relevant to the project.” Both types of knowl-edge were required for every project in this organ-ization: for example, an infrastructure project inRussia required technical expertise in engineeringas well as information about the workings of localgovernment ministries; a social services project inArgentina required knowledge about best practicesin service provision and also about the particular

needs of the target population. To construct theexternal knowledge variable, I averaged the re-sponses to the eight survey items for each teammember and then within teams (� � .85, ICC � .06,p � .05, rwg � .69). To test for the interaction effectstated in Hypothesis 1, I multiplied each team’sautonomy score by its external knowledge score,after standardizing to avoid high multicollinearity(Neter, Wasserman, & Kutner, 1985).

Knowledge characteristics. To distinguish be-tween knowledge content that was relatively scarceversus relatively common, I constructed separatemeasures of external country knowledge (� � .70,ICC � .08, p � .01, rwg � .66) and external technicalknowledge (� � .72, ICC � .06, p � .05, rwg � .65).My interview data indicated that teams in this or-ganization typically found country knowledge to bescarcer than technical knowledge, for three rea-sons. First, team members were usually chosen fortechnical expertise rather than client countryfamiliarity.6

Second, country knowledge in the form of reli-able information on economic and social condi-tions is often very limited in developing countries,whereas technical knowledge usually builds on for-mal education or experience in other countries andso is more abundant. Third, team members couldnot always identify and access country knowledgeas easily as technical knowledge because they weremostly based at the U.S. headquarters rather than inthe client countries, and accordingly they tended tobe more deeply entrenched in professional thannational knowledge-sharing networks. I establishedthe convergent and discriminant validity of the twofour-item measures using two approaches (Ven-katraman & Grant, 1986): a multitrait-multimethodmatrix analysis indicated that the average within-scale correlations of the group-level measures (r �.48, r � .46) exceeded their average between-scalecorrelation (r � .27), and a group-level confirma-tory factor analysis on the eight items using maxi-mum-likelihood estimates indicated that the two-factor structure was superior to a one-factorstructure (��2

1 � 6.10, p � .05). I tested Hypothesis2 by comparing the interaction effects between au-tonomy and country knowledge (relatively scarce)with those between autonomy and technicalknowledge (relatively common).

Task characteristics. To examine the effects oftask uncertainty, I measured both task novelty and

5 These questions could not establish with certaintywhether teams actually used the knowledge they ob-tained, but the emphasis on relevant knowledge encour-aged them to recall knowledge that they had founduseful.

6 In keeping with this, the survey respondents re-ported possessing less country than technical knowledgeprior to their projects (means � 3.43 and 3.81, respec-tively; t � 4.50, p � .001).

996 OctoberAcademy of Management Journal

task complexity. Items for both were rated from 1,“very little,” to 5, “a lot.” To capture task novelty, Iused two survey items: “Prior to the start of thisproject, how much relevant (a) technical knowl-edge (b) country knowledge did you personallyhave?” After reverse-coding, the higher the averageteam member score on this variable (� � .44, ICC �.12, p � .01, rwg � .55), the higher the novelty of thetask for a team.7 To capture task complexity, I usedthree survey items (cf. Tushman, 1978): “To whatextent did the project require complex approachesand solutions?”; “To what extent did the work de-part from the usual work of a routine financial/technical project?”; “To what extent did the tech-niques or skills or information needed for theproject change during the course of the project?”The higher the average team member score on thisvariable (� � .77, ICC � .17, p � .01, rwg � .74), thehigher the complexity of the task.

To examine the effects of task pressure, I con-structed measures of both time and client pressure.To capture time pressure, I used archival budgetdata to calculate the logged number of days fromproject initiation to completion, task duration. Inmy interviews, team members reported that timepressure was greater when task duration wasshorter because it was more difficult to ensure thatthe standard components required of any projectwere adequately covered. For example, all financialprojects, from narrowly focused student loan pro-grams to large-scale housing sector investments,required full stakeholder, environmental, and im-plementation readiness assessments. These re-quirements could make it hard to satisfactorilycomplete even tasks that were relatively straightfor-ward (i.e., low in novelty or complexity) in a shorttime. To capture client pressure, I used national eco-nomic data to construct a logged measure of clientcountry size, task client, by calculating the country’sgross national product (GNP) as a percentage of re-gional GNP (because the organization managed itsprojects by regions). The interviewees reported thatlarger countries’ governments tended to be morepowerful clients who tried to exert more sway overthe direction and details of the work conducted forthem; they also served as regional role models forprojects that might be taken to other countries. Thus,client pressure was typically higher in projects con-ducted for China or Brazil, for example, than inprojects for smaller countries in their regions.

To test Hypothesis 4, I used median splits todivide the sample into relatively high- versus low-

novelty tasks and relatively high- versus low-com-plexity tasks and then compared the models foreach paired set of tasks to establish whether theinteraction effect between autonomy and externalknowledge was stronger when task uncertainty washigh rather than low. Similarly, to test Hypothesis5, I used median splits to divide the sample intorelatively short- versus long-duration tasks and intorelatively large- versus small-client tasks and thencompared the models for each paired set of tasks toestablish whether the interaction effect was stron-ger when task pressures were high rather than low.8

Control variables. To account for other possibleinfluences on strategic and operational effective-ness, the models included team size (number ofteam members) and team location (1, “headquar-ters”; 0, “country office”). I also constructed a mea-sure of team satisfaction using four items withscales ranging from 1, “strongly disagree,” to 5,“strongly agree” (Wageman, Hackman, & Lehman,2005): “Working together energized and upliftedmembers of this team”; “There was a lot of unpleas-antness among the members of this team” (reverse-coded); “Members of this team were getting betterand better at working together”; “The longer weworked together as a team, the less well we did”(reverse-coded) (� � .81, ICC � .18, p � .01, rwg �.82). Assuming that better internal relations amongthe team members would be reflected in higherteam satisfaction, I used this measure to control forhow well the team members worked together, aswell as for possible satisfaction-driven biases thatmight have affected their responses to the othersurvey items. To capture differences due to senior-ity or work experience, I included the team mem-bers’ average organizational tenure in years at thestart of the project, as well as their average nonor-ganizational tenure, as years spent in other organ-izations. I also included late respondents, the per-centage of team members who returned theirsurveys after the results of their project evaluationhad been announced, since the outcome of theirevaluation might have influenced their responses.Because core members might have had differentroles and views than noncore members, the modelsalso included core respondents, the proportion ofsurvey respondents in each team who were coremembers. Finally, I included project type (1, “fi-

7 The main effect of this variable also serves as a con-trol for the team’s level of knowledge prior to the project.

8 I conducted sensitivity checks to see whether reallo-cation of borderline tasks affected the results, but theydid not. An alternative approach would be to examinethree-way interactions between autonomy, externalknowledge, and the task variables, but the number ofobservations in the data set made this infeasible.

2010 997Haas

nancial”; 0, “technical”) to capture differences be-tween the two types of projects.9

RESULTS

Table 1 reports descriptive statistics and correla-tions. As shown, the two dependent variables aresignificantly but not highly correlated (r � .37).Additionally, the correlation between autonomyand external knowledge is low and not significant(r � .15), indicating that the two constructs areorthogonal rather than necessarily related or alter-native choices. The average within-scale correla-tions for these two survey constructs (r � .76, r �.34) exceeded the average between-scale correla-tion (r � .06), and a confirmatory factor analysisindicated that the two-factor model provided a bet-ter fit to the data than a one-factor model, verifyingtheir convergent and discriminant validity (��2

1 �181.2, p � .01).

Because both dependent variables were categor-ical and ordered, I used ordinal logit analysis to testthe hypotheses (Long, 1997).10 Tables 2a, 3a, and 4ashow the strategic effectiveness models; Tables 2b,3b, and 4b show the operational effectiveness mod-els. All models included the full set of control andtask variables (not shown); the only consistentlysignificant results on these were that teams scoringhigher on strategic effectiveness had more members(b � 0.16–0.18, p � .05), and teams scoring higheron operational effectiveness tended to serve largerclients and work on financial projects (b � 0.29–0.32, p � .05; b � 1.22–1.99, p � .05).

In Tables 2a and 2b, models 1a and 1b show thatthe main effect of autonomy is positive and signif-icant for strategic effectiveness but not for operationaleffectiveness, and models 2a and 2b show that themain effect of external knowledge is also positive andsignificant for strategic but not operational effective-ness. In additional analyses, I examined whether au-tonomy or external knowledge use showed curvilin-ear effects, but quadratic terms did not have anegative effect on either strategic or operational effec-tiveness. Models 3a and 3b replicate the main effect

results when the autonomy and external knowledgevariables are included together.

Models 4a and 4b report the results for Hypoth-esis 1, which proposes a positive interaction effectbetween autonomy and external knowledge. Thesemodels show that the interaction effect is positiveand significant for both strategic and operationaleffectiveness, supporting Hypothesis 1. The resultsare plotted in Figure 2 to illustrate the magnitudesof the effects for teams with varying levels of au-tonomy and external knowledge. High and low lev-els of each are set at one standard deviation aboveand below their mean levels (Aiken & West, 1991).The vertical axes range from 1 to 3, giving a maxi-mum difference of 2 points between high- and low-effectiveness projects. The plots show that teamswith high autonomy and high external knowledgeuse delivered substantially more strategically andoperationally effective projects on average thanteams with low autonomy and high external knowl-edge use (difference of 0.74 points � 37% and 0.32points � 16%) or high autonomy and low externalknowledge use (difference of 0.42 points � 21%and 0.38 points � 19%).

Models 5a and 5b and models 6a and 6b presentthe results for Hypothesis 2, which proposes thatthe autonomy-knowledge interaction is more posi-tive if knowledge content is scarce rather than com-mon, and Hypothesis 3, stating that the autonomy-knowledge interaction is more positive ifknowledge sources are nonorganizational ratherthan organizational. Model 5a shows that for stra-tegic effectiveness, there is a significant, positiveinteraction between autonomy and country knowl-edge, which is relatively scarce, whereas there is anegative and less significant interaction betweenautonomy and technical knowledge, which is rela-tively common. Model 6a shows that autonomy hasa significant, positive interaction with nonorgani-zational knowledge but not with organizationalknowledge. These models also show positive maineffects of country and nonorganizational knowl-edge. Comparing models with and without equalityconstraints on the effects of these variables, I foundthat the difference between the country knowledgeand technical knowledge effects is significant (�2 �8.59, p � .05), as is the difference between theorganizational and nonorganizational knowledgeeffects (�2 � 3.88, p � .05). Models 5b and 6b showa similar pattern of results for operational effective-ness: the interaction with autonomy is marginallysignificant for country but not technical knowledgecontent, and significant for nonorganizational butnot organizational knowledge sources. The differ-ence between the country and technical knowledgeeffects is marginally significant (�2 � 2.96, p � .10),

9 I also ran models that included controls for projectregions and divisions; these did not affect the results.

10 Since interaction effects in nonlinear regressionmodels can be problematic to interpret (Ai & Norton,2003), I also generated the marginal effects for the inter-action terms and ran the models using an ordinary leastsquares specification instead. These two alternative ap-proaches both generated the same pattern of results forthe variable coefficients and for the statistical tests ofcoefficient differences.

998 OctoberAcademy of Management Journal

TA

BL

E1

Des

crip

tive

Sta

tist

ics

and

Biv

aria

teC

orre

lati

onsa

Var

iabl

eM

ean

s.d

.M

in.

Max

.1

23

45

67

89

1011

1213

1415

1617

1819

1.S

trat

egic

effe

ctiv

enes

s1.

680.

621

32.

Op

erat

ion

alef

fect

iven

ess

1.90

0.67

13

.37

3.T

eam

auto

nom

yb3.

620.

312.

754.

32.2

4.0

64.

Ext

ern

alkn

owle

dge

b3.

000.

431.

913.

98.1

3.0

1.1

55.

Ext

ern

alco

un

try

know

led

geb

3.08

0.47

1.88

4.17

.16

.06

.15

.94

6.E

xter

nal

tech

nic

alkn

owle

dge

b2.

910.

441.

753.

92.0

5�

.02

.12

.95

.80

7.E

xter

nal

orga

niz

atio

nal

know

led

geb

2.83

0.44

1.81

3.75

.02

�.0

3.0

3.8

4.7

7.8

2

8.E

xter

nal

non

orga

niz

atio

nal

know

led

geb

3.15

0.54

1.67

4.30

.19

.03

.23

.90

.86

.83

.52

9.T

ask

nov

elty

1.38

0.54

0.40

3.00

.01

.02

�.1

1�

.19

�.1

9�

.19

�.2

1�

.12

10.

Tas

kco

mp

lexi

ty3.

550.

481.

674.

67.0

7.1

0.1

4.2

2.2

6.1

5.1

7.2

0�

.17

11.

Tas

kd

ura

tion

c35

4.49

258.

5143

1,44

2�

.02

�.0

6.0

3�

.02

.01

�.0

5�

.00

�.0

3�

.03

.08

12.

Tas

kcl

ien

tc0.

140.

210.

010.

76.1

0.1

8.0

2�

.01

�.0

4.0

4.0

3�

.03

�.1

6.0

7.0

213

.T

eam

size

8.48

4.15

223

.30

.10

.02

.06

.08

.04

.09

.03

.04

.23

.00

�.0

814

.T

eam

loca

tion

0.80

0.40

01

.05

.08

�.0

4�

.08

.00

�.1

4�

.07

�.0

7.1

4.1

9�

.02

�.1

5.2

115

.T

eam

sati

sfac

tion

4.04

0.37

2.75

4.75

�.0

1.2

0.2

6.2

6.2

7.2

5.1

6.2

8�

.14

.05

�.1

6�

.08

�.1

9.0

616

.O

rgan

izat

ion

alte

nu

re8.

464.

401.

0025

.50

�.0

6.1

4.0

5�

.02

�.0

4.0

0.0

4�

.06

�.0

0.2

2�

.14

�.0

2�

.13

.02

.02

17.

Non

orga

niz

atio

nal

ten

ure

16.9

37.

052.

5036

.50

.02

.03

.14

.05

.14

�.0

2.0

3.0

6�

.31

.22

�.0

3.1

4.2

0.1

3.0

1�

.05

18.

Lat

ere

spon

den

ts0.

560.

410.

001.

00�

.06

.00

�.1

5�

.04

�.0

2�

.08

.03

�.1

1.1

4�

.09

.11

.05

�.0

7.0

2�

.10

.09

.13

19.

Cor

ere

spon

den

ts0.

760.

220.

171.

00.0

3.1

7.0

3.0

5.0

1.0

8.0

7.0

2�

.02

.16

�.1

0�

.09

�.1

0�

.05

.38

.14�

.03

.03

20.

Tas

kty

pe

0.52

0.50

01

�.0

0.2

0.0

4.0

0.0

9�

.06

.06

�.0

5.1

0.1

6�

.16

�.0

8.2

8.3

1.0

7.1

2.3

5.4

0.1

4

an

�96

.If

r�

.16,

p�

.10;

ifr

�.2

0,p

�.0

5;if

r�

.26,

p�

.01.

bV

aria

ble

was

stan

dar

diz

edin

anal

yses

bysu

btra

ctin

gth

em

ean

and

div

idin

gby

the

stan

dar

dd

evia

tion

.c

Su

mm

ary

stat

isti

csar

efr

omth

eu

nlo

gged

vari

able

;co

rrel

atio

ns

are

from

the

logg

edva

riab

le.

but the difference between the organizational andnonorganizational knowledge interaction effects isnot (�2 � 1.95, n.s.). The findings for Hypotheses 2and 3 thus indicate strong support for strategiceffectiveness and a weaker but similar pattern ofsupport for operational effectiveness.

The four sets of paired models presented inTables 3a and 3b report the results for Hypothesis4, which proposes that the autonomy-knowledgeinteraction effect is more positive when task uncer-tainty (i.e., novelty or complexity) is high ratherthan low. The models show that for both strategicand operational effectiveness, the interaction ispositive and significant for high-novelty tasks(7a[ii] and 7b[ii]) but not for low-novelty tasks(7a[i] and 7b[i]). For strategic effectiveness, the in-teraction effect is also positive and marginally sig-nificant for high-complexity tasks (8a[ii]) but notfor low-complexity tasks (8a[i]). For operational ef-fectiveness, however, the interaction is not signifi-cant for either high- or low-complexity tasks (8b[ii]and 8b[i]). Since testing the hypothesis required

comparing coefficients across models, I tested forthe statistical significance of these differences us-ing the seemingly unrelated estimation algorithmin Stata 10. The tests indicated that the pairednovelty models are significantly different fromeach other for operational but not strategic effec-tiveness (�2 � 22.00, p � .05; �2 � 12.70, n.s); incontrast, the paired complexity models are signifi-cantly different from each other for strategic but notoperational effectiveness (�2 � 26.06, p � .05; �2 �16.80, n.s.). These results provide partial supportfor Hypothesis 4: the complementarity between au-tonomy and external knowledge is greater for opera-tional effectiveness under conditions of high tasknovelty, and greater for strategic effectiveness un-der conditions of high task complexity.

Finally, the results for Hypothesis 5, which pro-poses that the autonomy-knowledge interaction ef-fect is more positive when task pressure is highrather than low, are shown in Tables 4a and 4b. Thefour paired sets of models comparing relativelyhigh-pressure tasks (i.e., those with shorter dura-

TABLE 2AResults of Ordinal Logit Analysis for Strategic Effectivenessa

Variables Model 1a Model 2a Model 3a Model 4a Model 5a Model 6a

Team autonomyb 0.72* (0.29) 0.73* (0.30) 0.80* (0.32) 0.95** (0.35) 0.74* (0.32)External knowledgeb 0.43† (0.25) 0.44† (0.26) 0.55† (0.30)External country knowledgeb 0.95† (0.53)External technical knowledgeb �0.05 (0.50)External organizational

knowledgeb�0.25 (0.31)

External nonorganizationalknowledgeb

0.82* (0.35)

Team autonomy � externalknowledge

0.77* (0.34)

Team autonomy � externalcountry knowledge

2.28** (0.72)

Team autonomy � externaltechnical knowledge

�1.05† (0.55)

Team autonomy � externalorganizational knowledge

0.34 (0.37)

Team autonomy � externalnonorganizationalknowledge

0.66* (0.34)

df c 13 13 14 15 17 17Log-likelihood �59.65 �61.47 �58.19 �55.21 �50.48 �52.10Log-likelihood � 2 ratio testd 6.42* 2.78† 9.34** 15.30** 24.70** 21.52**Pseudo-R2 0.13 0.11 0.15 0.20 0.27 0.24

a n � 96. Standard errors are in parentheses.b Variable was standardized by subtracting the mean and dividing by the standard deviation.c Models include all control and task variables.d Compared to baseline model with control variables only.

† p � .10* p � .05

** p � .01Two-tailed test for variable coefficients.

1000 OctoberAcademy of Management Journal

tions or larger clients) with relatively low-pressuretasks show a similar pattern for the two dependentvariables: the interaction effects are positive andsignificant for tasks with shorter durations (9a[i]and 9b[i]), but not for tasks with longer durations(9a[ii] and 9b[ii]). The interaction effects are alsolarger and more significant for tasks with largerclients (10a[ii] and 10b[ii]) than for tasks withsmaller clients (10a[i] and 10b[i]). Using the seem-ingly unrelated estimation algorithm to test for thestatistical significance of these differences indi-cated, however, that the paired task duration mod-els are marginally significantly different from eachother for operational but not strategic effectiveness(�2 � 20.16, p � .10; �2 � 9.21, n.s.), whereas thepaired task client models are significantly differentfrom each other for strategic but not operationaleffectiveness (�2 � 23.78, p � .05; �2 � 12.65, n.s.).Thus, in partial support of Hypothesis 5, thecomplementarity between autonomy and externalknowledge is greater for operational effectivenesswhen task durations are shorter, and greater forstrategic effectiveness when task clients are larger.

DISCUSSION

Extending organizational design principles toself-managing teams engaged in knowledge-inten-sive work, this study has shown that teams’ auton-omy and use of external knowledge provide com-plementary conditions for team effectiveness. Inthe multinational organization studied here, teamswith high levels of both autonomy and externalknowledge delivered more strategically and opera-tionally effective projects than teams with highautonomy but low external knowledge or high ex-ternal knowledge but low autonomy. The comple-mentarity between autonomy and external knowl-edge use depended, however, on characteristics ofthe knowledge and the task. The combination im-proved both strategic and operational effectivenesswhen the content of knowledge was scarce (countryknowledge) but not when it was common (technicalknowledge), and it improved both types ofeffectivenesswhentheknowledgesourceswerenonor-ganizational but not when they were organizational.The combination of autonomy and external knowl-

TABLE 2BResults of Ordinal Logit Analysis for Operational Effectivenessa

Variables Model 1b Model 2b Model 3b Model 4b Model 5b Model 6b

Team autonomyb �0.24 (0.28) �0.24 (0.28) �0.24 (0.28) �0.29 (0.29) �0.26 (0.29)External knowledgeb 0.03 (0.25) 0.04 (0.25) 0.09 (0.26)External country knowledgeb 0.74 (0.49)External technical knowledgeb �0.51 (0.48)External organizational

knowledgeb�0.15 (0.29)

External nonorganizationalknowledgeb

0.27 (0.31)

Team autonomy � externalknowledge

1.00** (0.32)

Team autonomy � externalcountry knowledge

1.05† (0.60)

Team autonomy � externaltechnical knowledge

0.12 (0.52)

Team autonomy � externalorganizational knowledge

0.23 (0.32)

Team autonomy � externalnonorganizationalknowledge

0.87** (0.33)

df c 13 13 14 15 17 17Log-likelihood �70.53 �70.90 �70.51 �64.81 �63.21 �64.00Log-likelihood �2 ratio testd 0.76 0.02 0.80 12.2** 15.4** 13.82**Pseudo-R2 0.10 0.10 0.10 0.17 0.19 0.18

a n � 96. Standard errors are in parentheses.b Variable was standardized by subtracting the mean and dividing by the standard deviation.c Models include all control and task variables.d Compared to baseline model with control variables only.

† p � .10** p � .01Two-tailed test for variable coefficients.

2010 1001Haas

edge use also improved strategic effectiveness morefor tasks with higher complexity or client pressure,but it improved operational effectiveness more fortasks with higher novelty or time pressure. Thus, thecomplementarity between autonomy and externalknowledge use was contingent on situation.

Further Evidence

The findings of this study indicate substantial sup-port for the theoretical model, but some of the empir-ical results may have alternative explanations that

require consideration. One possibility is that exoge-nous factors could account for the observed relation-ships between autonomy, external knowledge, andteam performance. For example, perhaps some teamswere given more autonomy, obtained more knowl-edge, and also performed better because their mem-bers were more expert. However, this explanationwould suggest that team member expertise should bestrongly correlated with both autonomy and externalknowledge, but the correlations with three measuresof expertise utilized in this study—prior knowledge(which was higher when task novelty was lower),

TABLE 3AResults of Ordinal Logit Analysis for Strategic Effectiveness in Less versus More Uncertain Tasksa

Variables

Task Novelty Task Complexity

Low: Model 7a[i] High: Model 7a[ii] Low: Model 8a[i] High: Model 8a[ii]

Team autonomyb 1.71† (0.40) 1.85** (0.62) 1.99** (0.74) 2.70** (0.77)External knowledgeb 0.19 (0.39) 0.91* (0.46) 0.38 (0.38) 0.47 (0.52)Team autonomy � external knowledge 0.19 (0.53) 1.01* (0.48) 0.84 (0.57) 1.04† (0.61)

n 48 47 48 47dfc 13 13 13 133Log-likelihood �32.68 �29.03 �25.13 �26.12Log-likelihood � 2 ratio testd 7.48† 19.32** 14.22** 24.98**Pseudo-R2 0.29 0.35 0.36 0.42

a Standard errors are in parentheses.b Variable was standardized by subtracting the mean and dividing by the standard deviation.c Models include all the control variables and task variables.d Compared to the same models including control variables only.

† p � .10* p � .05

** p � .01Two-tailed test for variable coefficients.

TABLE 3BResults of Ordinal Logit Analysis for Operational Effectiveness in Less versus More Uncertain Tasksa

Variables

Task Novelty Task Complexity

Low: Model 7b[i] High: Model 7b[ii] Low: Model 8b[i] High: Model 8b[ii]

Team autonomyb �0.70† (0.42) 0.31 (0.33) �0.47 (0.41) 0.37 (0.39)External knowledgeb 0.06 (0.38) 0.26 (0.38) 0.13 (0.35) �0.68† (0.42)Team autonomy � external knowledge 0.43 (0.47) 0.58† (0.34) 0.76 (0.48) 0.62 (0.40)

n 47 46 47 46dfc 13 13 13 13Log-likelihood �41.72 �41.72 �39.35 �35.21Log-likelihood � 2 ratio testd 3.34 5.26 5.90 6.98†

Pseudo-R2 0.18 0.24 0.19 0.21

a Standard errors are in parentheses.b Variable was standardized by subtracting the mean and dividing by the standard deviation.c Models include all the control variables and task variables.d Compared to the same models including control variables only.

† p � .10Two-tailed test for variable coefficients.

1002 OctoberAcademy of Management Journal

organizational tenure, and nonorganizational ten-ure—were low. Another possibility is that more novelor complex tasks that required more knowledge werestaffed with better teams that were given more auton-omy, but the correlations between autonomy and tasknovelty or complexity were also low. Similarly, per-haps tasks with shorter durations or larger clientswere assigned to better teams with more autonomy,but again, the correlations between autonomy andtask duration or client size were low.

A different possibility is that the findings of thestudy might be a result of postevaluation attribution

bias. Some teams had undergone the full project eval-uation process before the surveys were distributed,raising the possibility that the members of theseteams knew the outcome of their evaluations andmade self-serving attributions in their responses tothe survey (Miller & Ross, 1975). The research designallowed me to test for such biases by comparing 19teams whose members all returned their surveys be-fore their project evaluations were completed with 37teams whose members all returned their surveys atleast seven days after their evaluations were com-pleted, by which time the results would have been

TABLE 4AResults of Ordinal Logit Analysis for Strategic Effectiveness in Less versus More Pressured Tasksa

Variables

Task Duration Task Client

Short: Model 9a[i] Long: Model 9a[ii] Small: Model 10a[i] Large: Model 10a[ii]

Team autonomyb 0.79† (0.48) 0.82 (0.58) 1.61† (0.84) 0.53** (0.49)External knowledgeb 0.22 (0.43) 0.81† (0.47) 1.92* (0.79) �0.03 (0.39)Team autonomy � external knowledge 0.77† (0.42) 0.36 (0.68) 1.04 (0.82) 1.31* (0.53)

n 43 37 40 40df c 13 13 13 13Log-likelihood �29.35 �25.56 �19.30 �25.24Log-likelihood � 2 ratio testd 8.86* 6.10 7.68† 9.54*Pseudo-R2 0.22 0.17 0.26 0.43

a Standard errors are in parentheses.b Variable was standardized by subtracting the mean and dividing by the standard deviation.c Models include all the control variables and task variables.d Compared to the same models including control variables only.

† p � .10* p � .05

** p � .01

TABLE 4BResults of Ordinal Logit Analysis for Operational Effectiveness in Less versus More Pressured Tasksa

Variables

Task Duration Task Client

Short: Model 9b[i] Long: Model 9b[ii] Small: Model 10b[i] Large: Model 10b[ii]

Team autonomyb �0.30 (0.43) �0.13 (0.55) �0.09 (0.63) �0.05 (0.41)External knowledgeb 0.11 (0.41) 0.23 (0.43) 0.23 (0.52) 0.13 (0.33)Team autonomy � external knowledge 1.57** (0.50) 0.10 (0.66) 1.09† (0.68) 1.57** (0.43)

n 42 37 39 40df c 13 13 13 13Log-likelihood �29.33 �28.03 �27.57 �31.78Log-likelihood � 2 ratio testd 13.78** 0.34 7.62† 8.24*Pseudo-R2 0.34 0.15 0.21 0.22

a Standard errors are in parentheses.b Variable was standardized by subtracting the mean and dividing by the standard deviation.c Models include all the control variables and task variables.d Compared to the same models including control variables only.

† p � .10* p � .05

** p � .01

2010 1003Haas

announced. The tests showed no significant differ-ences in the main variables, including autonomy andexternal knowledge, or in their correlations with stra-tegic or operational effectiveness. There were also nosignificant differences in the effectiveness ratings forthe two sets of teams. Attribution bias thus is not aconvincing alternative explanation for the results ofthis study.

Beyond these empirical checks, the study providessome evidence for the robustness of the theoreticalmodel across performance metrics as well as projecttypes. The combination of autonomy and externalknowledge was found here to be positively associatedwith two substantively different measures of teamperformance, strategic and operational effectiveness,which were significantly but not highly correlated.Prior research has also shown a similar, positive as-sociation with a third measure of team performance—project quality—that is different from strategic andoperational effectiveness (r � .65, p � .01; r � .45, p �.01) (Haas, 2006a). Additionally, examining financialand technical projects separately indicated that theinteraction between autonomy and external knowl-edge was positively and significantly associated withstrategic effectiveness for both project types (b � 1.61,p � .05; b � 1.25, p � .05), and with operationaleffectiveness for technical projects though not for fi-nancial projects (b � 1.52, p � .01; b � 0.59, p � .16).

Theoretical and Practical Implications

Team effectiveness. For self-managing teams, themixed evidence of prior research on the effects ofautonomy is concerning as well as perplexing. Previ-ous studies have typically examined why self-man-aging teams may not perform well by focusing ontheir internal interactions, which may suffer from, forexample, insufficient interdependence (Langfred,

2005), negative feelings toward collaboration (Kirk-man & Shapiro, 2001), or rigid rule enforcement(Barker, 1993). Only a few studies have explicitlyconsidered the importance of teams’ external interac-tions, for example by examining the role of outsidecoaches (Manz & Sims, 1993), reward systems (Wage-man, 1995), and corporate strategic priorities(Zellmer-Bruhn & Gibson, 2006). With the presentstudy I contribute to this effort by arguing that auton-omy is a double-edged sword that offers indepen-dence but can lead teams to become isolated fromtheir environments. Supporting this argument, Ifound that teams benefited more from autonomy ifthey avoided isolation by seeking external knowl-edge. Recognizing the importance of such externalinteractions for knowledge-intensive work thus offersa promising way to increase the robustness of theoriesof team self-management. Further, managers who im-plement work practices based on team self-manage-ment are likely to be better able to realize their poten-tial if they recognize that autonomy carries risks ofisolation that these teams should try to avoid.

By drawing attention to the influence risks of ex-ternal knowledge, this study also addresses the mixedfindings of prior research on team boundary span-ning. Previous studies have focused mostly on thetechnical, social, and cognitive barriers to knowledgesharing, such as search and transfer problems (Han-sen, 1999), arduous relationships (Szulanski, 1996),and team members’ cosmopolitan versus local orien-tations (Haas, 2006b). The value of knowledge shar-ing may also be reduced by political problems, how-ever, since knowledge is a double-edged sword thatcan be used to influence as well as to inform (e.g.,Pfeffer, 1981). For research on teams conductingknowledge-intensive work, as well as the broader lit-eratures on knowledge sharing (e.g., Argote, McEvily,& Reagans, 2003) and the knowledge-based view of

FIGURE 2Interaction Effects of Team Autonomy and External Knowledgea

a To illustrate the direction and magnitude of effects, low values were set at one standard deviation below the mean, high values wereset at one standard deviation above the mean, and the plots were constructed using OLS regression.

1004 OctoberAcademy of Management Journal

multinational enterprises (e.g., Kogut & Zander,1993), the implication is that teams may not benefitfrom external knowledge unless they can make inde-pendent decisions based on this knowledge. For man-agers, recognizing that knowledge sharing carries in-fluence risks suggests that investing in knowledgemanagement infrastructure such as document data-bases and “communities of practice” may yield lowerthan expected returns because the knowledge that isshared through these initiatives may be biased, mis-leading, or intended to persuade rather than assistusers.

Multinational management. In focusing on theperformance of teams in a multinational organization,this study also provides a team-level analogue to ex-isting theories of organizational design that helps toexplain variation neglected by these theories, espe-cially as they apply to multinational management.Research built on the differentiation-integration viewhas typically focused on large business units, yet thecritical tasks of many organizations today are carriedout by relatively small teams located in one or severalunits, such as cross-functional teams. Multinationalmanagement research, in particular, has usually fo-cused on subsidiary autonomy and cross-subsidiaryknowledge flows (e.g., Birkinshaw et al., 1998; Kogut& Zander, 1993; Nohria & Ghoshal, 1997) rather thanon smaller units that operate in or across nationalsubsidiaries. In contrast, this study identifies teamautonomy as a critical locus of differentiation in or-ganizations and highlights knowledge flows to teamsas a critical integrating mechanism. By demonstratingthe contingent complementarity between autonomyand external knowledge at the team level, this studyextends the principles of the differentiation-inte-gration view to explain performance variationwithin and among national subsidiaries, as well asbusiness units more generally. Additionally, focus-ing on teams enables practitioners to address thecentral macrolevel strategic challenge of multina-tional management—to “think global, act local”(Prahalad & Doz, 1987)—at the micro level wherethis mandate is implemented.

Team ambidexterity. Finally, this study offers aview of conditions for team effectiveness that is rele-vant to the growing literature on organizational am-bidexterity. Research in this area has highlighted thepotential value of structures and processes that en-able organizations to engage in exploration and ex-ploitation simultaneously, rather than separately orsequentially (e.g., Gibson & Birkinshaw, 2004; Tush-man & O’Reilly, 1996). This research has typicallyoffered a macro perspective on how ambidexteritycan be achieved: for example, by focusing some busi-ness units on exploitation and others on exploration,then integrating these in a firm’s senior team (Smith &

Tushman, 2005). Yet, increasingly, ambidexterity isimportant at lower levels of organizations. For exam-ple, new-product development, technological inno-vation, and professional service delivery all requireteams to simultaneously exploit their existing capa-bilities and explore new approaches and opportuni-ties. The concept of “team ambidexterity” may provevaluable, therefore, for understanding how ambidex-terity can be achieved in organizations. Since auton-omy allows teams to exploit their capabilities, andexternal knowledge allows them to explore new ap-proaches and opportunities, the combination of theseconditions can be viewed as facilitating ambidexter-ity. Further, the contingent value of these comple-mentary conditions shows that team ambidexterity,like organizational ambidexterity, may not always benecessary, and provides insight for researchers andmanagers into the boundary conditions under whichit is (or is not) advantageous.

Directions for Future Research

Alongside further exploration of the theoreticaland practical implications of this study, investiga-tion of the extent to which its findings hold in othersettings could also usefully be pursued, since thepresent research was conducted in one organiza-tion. For example, the benefits and risks of obtain-ing external knowledge may depend on an organi-zation’s culture, since identifying and securinguseful knowledge may be more costly in culturesthat encourage hoarding rather than sharing(Boisot, 1998). The findings could also benefit fromdetailed cross-national comparisons, since, for ex-ample, the level of independence experienced byautonomous teams may vary with national contexteven in the same multinational organization (cf.Gibson et al., 2003). Future research might also ex-plore whether the current theoretical model appliesbeyond the domain of self-managing teams engagedin knowledge-intensive work. For example, muchprior research on team autonomy has focused onblue-collar work groups that conduct labor-intensivework, which may be less vulnerable to the risks ofisolation or influence (cf. Janz et al., 1997).

Other questions raised by this study arise fromlimitations of the data, which did not allow forexamination of internal team relations that couldincrease or decrease the advantages of autonomyand external knowledge, such as “transactive mem-ory systems” (e.g., Lewis, 2004) or subgroup con-flict (e.g., Gibson & Vermeulen, 2003). Further un-packing the concepts of autonomy and externalknowledge could also be worthwhile: for example,future research could examine autonomy relative todifferent stakeholders to establish whether the ben-

2010 1005Haas

efits of independence and risks of isolation dependon who else is involved in decision making, andexternal knowledge providers could be examinedvia network methods to see whether different con-figurations of providers affect the information ben-efits and influence risks of obtaining externalknowledge.

Conclusion

In an influential political sociology theory of na-tion states, Evans (1995) coined the phrase “embed-ded autonomy” to argue that well-functioning na-tion states are those whose institutions solicit andconsider the opinions and concerns of their constit-uents but are able to resist excessive pressures fromthose constituents. The combination of autonomyand external knowledge that this study has shownto be valuable invokes a similar vision of the con-ditions that promote effective performance for self-managing teams engaged in knowledge-intensivework. As such teams strive to perform effectively,embedded autonomy—in the form of externalknowledge use combined with control over criticaltask decisions—can enable them to avoid the dan-gers of excessive isolation or influence and to makedecisions that are both informed and independent.

REFERENCES

Ai, C., & Norton, E. C. 2003. Interaction terms in logit andprobit models. Economics Letters, 80: 123–129.

Aiken, L. S., & West, S. G. 1991. Multiple regression:Testing and interpreting interactions. NewburyPark, CA: Sage.

Ancona, D. G., & Caldwell, D. F. 1992. Bridging theboundary: External activity and performance in or-ganizational teams. Administrative Science Quar-terly, 37: 634–665.

Argote, L., McEvily, B., & Reagans, R. 2003. Managingknowledge in organizations: An integrative frame-work and review of emerging themes. ManagementScience, 49: 571–582.

Barker, J. R. 1993. Tightening the iron cage: Concertivecontrol in self-managing teams. Administrative Sci-ence Quarterly, 38: 408–437.

Barnard, C. I. 1938. The functions of the executive.Boston: Harvard University Press.

Bartlett, C. A., & Ghoshal, S. 1989. Managing acrossborders: The transnational solution. Boston: Har-vard Business School Press.

Birkinshaw, J., Hood, N., & Jonsson, S. 1998. Buildingfirm-specific advantages in multinational corpora-tions: The role of subsidiary initiative. StrategicManagement Journal, 19: 221–242.

Birkinshaw, J., Nobel, R., & Ridderstrale, R. 2002. Knowl-edge as a contingency variable: Do the characteristicsof knowledge predict organizational structure? Or-ganization Science, 13: 274–289.

Boisot, M. 1998. Knowledge assets: Securing competi-tive advantage in the information economy. Ox-ford, U.K.: Oxford University Press.

Brown, G., Lawrence, T. B., & Robinson, S. L. 2005.Territoriality in organizations. Academy of Man-agement Review, 30: 577–594.

Carlile, P. 2002. A pragmatic view of knowledge andboundaries: Boundary objects in new product devel-opment. Organization Science, 13: 442–455.

Cohen, S. G., & Bailey, D. E. 1997. What makes teamswork: Group effectiveness research from the shopfloor to the executive suite. Journal of Management,23: 239–290.

Cohen, S. G., & Ledford, G. E. 1994. The effectiveness ofself-managing teams: A quasi-experiment. HumanRelations, 47: 13–34.

Collins, C. J., & Clark, K. D. 2003. Strategic human re-source practices, top management team social net-works, and firm performance. Academy of Manage-ment Journal, 46: 740–751.

Cordery, J. L., Mueller, W. S., & Smith, L. M. 1991.Attitudinal and behavioral effects of autonomousgroup working: A longitudinal field study. Academyof Management Journal, 34: 464–476.

Cummings, J. N. 2004. Work groups, structural diversity,and knowledge sharing. Management Science, 50:352–364.

Cyert, R. M., & March, J. G. 1963. A behavioral theory ofthe firm. Cambridge, MA: Blackwell.

Earley, P. C., & Gibson, C. B. 2002. Multinational workteams: A new perspective. Mahwah, NJ: Erlbaum.

Edwards, R. C. 1979. Contested terrain: The transfor-mation of the workplace in the twentieth century.New York: Basic Books.

Evans, P. 1995. Embedded autonomy: States and indus-trial transformation. Princeton, NJ: Princeton Uni-versity Press.

Feldman, S. P. 1988. Secrecy, information, and politics:An essay in organizational decision making. HumanRelations, 41: 73–90.

Galbraith, J. R. 1973. Designing complex organizations.Boston: Addison-Wesley.

Garnier, G. H. 1982. Context and decision-making auton-omy in the foreign affiliates of U.S. multinationalcorporations. Academy of Management Journal,25: 893–908.

Gibson, C. B., & Birkinshaw, J. 2004. The antecedents,consequences, and mediating role of organizationalambidexterity. Academy of Management Journal,47: 209–226.

1006 OctoberAcademy of Management Journal

Gibson, C. B., & Vermeulen, F. 2003. A healthy divide:Subgroups as a stimulus for team learning. Admin-istrative Science Quarterly, 48: 202–239.

Gibson, C. B., Zellmer-Bruhn, M., & Schwab, D. S. 2003.Team effectiveness in multinational organizations:Development and evaluation across contexts. Groupand Organization Management, 28: 444–474.

Gladstein, D. 1984. Groups in context: A model of taskgroup effectiveness. Administrative Science Quar-terly, 29: 499–517.

Gupta, A. K., & Govindarajan, V. 2000. Knowledge flowswithin multinational corporations. Strategic Man-agement Journal, 21: 473–496.

Haas, M. R. 2006a. Knowledge gathering, team capabilities,and project performance in challenging work environ-ments. Management Science, 52: 1170–1184.

Haas, M. R. 2006b. Acquiring and applying knowledge intransnational teams: The roles of cosmopolitans andlocals. Organization Science, 17: 313–332.

Haas, M. R., & Hansen, M. T. 2005. When using knowl-edge can hurt performance: The value of organiza-tional capabilities in a management consulting com-pany. Strategic Management Journal, 26: 1–24.

Haas, M. R., & Hansen, M. T. 2007. Different knowledge,different benefits: A productivity perspective onknowledge sharing in organizations. Strategic Man-agement Journal, 28: 1133–1153.

Hackman, J. R. 1987. The design of work teams. In J.Lorsch (Ed.), Handbook of organizational behav-ior: 315–342. Englewood Cliffs, NJ: Prentice-Hall.

Hackman, J. R. 2002. Leading teams: Setting the stagefor great performances. Boston: HBS Press.

Hansen, M. T. 1999. The search-transfer problem: Therole of weak ties in sharing knowledge across organ-ization subunits. Administrative Science Quar-terly, 44: 82–111.

Hargadon, A., & Sutton, R. I. 1997. Technology brokeringand innovation in a product development firm. Ad-ministrative Science Quarterly, 42: 716–749.

Haunschild, P. R., & Beckman, C. M. 1998. When dointerlocks matter?: Alternative sources of informa-tion and interlock influence. Administrative Sci-ence Quarterly, 43: 815–845.

Hobday, M. 2000. The project-based organisation: Anideal form for managing complex products and sys-tems? Research Policy, 29: 871–893.

Janis, I. 1982. Groupthink (2nd ed.). Boston: Houghton-Mifflin.

Janz, B. D., Colquitt, J. A., & Noe, R. A. 1997. Knowledgeworker team effectiveness: The role of autonomy, in-terdependence, team development, and contextualsupport variables. Personnel Psychology, 50: 877–904.

Jehn, K. 1995. A multimethod examination of the bene-fits and detriments of intragroup conflict. Adminis-trative Science Quarterly, 40: 245–282.

Katz, R., & Allen, T. J. 1982. Investigating the not in-vented here (NIH) syndrome: A look at the perfor-mance, tenure, and communication patterns of 50R&D project groups. R&D Management, 12: 7–20.

Kirkman, B. L., & Shapiro, D. L. 2001. The impact ofcultural values on job satisfaction and organizationalcommitment in self-managing work teams: The me-diating role of employee resistance. Academy ofManagement Journal, 44: 557–569.

Kogut, B., & Zander, U. 1993. Knowledge of the firm andthe evolutionary theory of the multinational corpo-ration. Journal of International Business Studies,24: 625–645.

Langfred, C. W. 2000. The paradox of self-management:Individual and group autonomy in work groups.Journal of Organizational Behavior, 21: 563–585.

Langfred, C. W. 2005. Autonomy and performance inteams: The multilevel moderating effect of task inter-dependence. Journal of Management, 31: 513–529.

Lawrence, P. R. 1993. The contingency approach to or-ganizational design. In R. T. Golembiewski (Ed.),Handbook of organizational behavior: 9–18. NewYork: Marcel Dekker.

Lawrence, P. R., & Lorsch, J. W. 1967. Organization andenvironment: Managing differentiation and inte-gration. Boston: HBS Press.

Levitt, B., & March, J. G. 1988. Organizational learning. InW. R. Scott & J. Blake (Eds.), Annual review of sociol-ogy, vol. 14: 319–338. Palo Alto, CA: Annual Reviews.

Lewis, K. 2004. Knowledge and performance in knowl-edge-worker teams: A longitudinal study of transac-tive memory systems. Management Science, 50:1519–1533.

Long, J. S. 1997. Regression models for categorical andlimited dependent variables. Thousand Oaks, CA:Sage.

Manz, C. C., & Sims, H. P. 1993. Business without bosses:How self-managing teams are building high-per-forming companies. New York: Wiley.

Martinez, J. I., & Jarillo, J. C. 1989. The evolution ofresearch on coordination mechanisms in multina-tional corporations. Journal of International Busi-ness Studies, 20: 489–514.

Menon, T., & Pfeffer, J. 2003. Valuing internal versusexternal knowledge: Explaining the preference foroutsiders. Management Science, 49: 497–513.

Miller, S. T., & Ross, M. 1975. Self-serving biases in theattribution of causality: Fact or fiction? Psychologi-cal Bulletin, 82: 93–118.

Mintzberg, H. 1983. Power in and around organiza-tions. Englewood Cliffs, NJ: Prentice-Hall.

Mohrman, S. A., Cohen, S. G., & Mohrman, A. M., Jr. 1995.Designing team-based organizations: New forms forknowledge work. San Francisco: Jossey-Bass.

Neter, J. W., Wasserman, S., & Kutner, M. H. 1985.

2010 1007Haas

Applied linear statistical models. Homewood, IL:Irwin.

Nohria, N., & Ghoshal, S. 1997. The differentiated net-work: Organizing multinational corporations forvalue creation. San Francisco: Jossey-Bass.

Perlow, L. 1999. The time famine: Toward a sociology ofwork time. Administrative Science Quarterly, 44:57–81.

Pettigrew, A. M. 1973. The politics of organizationaldecision-making. London: Tavistock.

Pfeffer, J. 1981. Power in organizations. Marshfield, MA:Pitman.

Pfeffer, J., & Salancik, G. 1978. The external control oforganizations: A resource dependency perspec-tive. New York: HarperCollins.

Prahalad, C. K., & Doz, Y. L. 1987. The multinationalmission: Balancing local demands and global vi-sion. New York: Free Press.

Reagans, R., Zuckerman, E. W., & McEvily, B. 2004. Howto make the team: Social networks vs. demographyas criteria for designing effective teams. Administra-tive Science Quarterly: 49: 101–133.

Roth, K., & Kostova, T. 2003. The use of the multinationalcorporation as a research context. Journal of Man-agement, 29: 883–902.

Schein, E. 1992. Organizational culture and leadership(2nd ed.). San Francisco: Jossey-Bass.

Selznick, P. 1949. The TVA and the grass roots: A studyin the sociology of formal organization. Berkeley:University of California Press.

Smith, W. K., & Tushman, M. 2005. Managing strategiccontradictions: A top management model for man-aging innovation streams. Organization Science,16: 522–536.

Spekman, R. E. 1979. Influence and information: An explo-ration of the boundary role person’s basis of power.Academy of Management Journal, 22: 104–117.

Szulanski, G. 1996. Exploring internal stickiness: Imped-iments to the transfer of best practice within thefirm. Strategic Management Journal, 17(winterspecial issue): 27–43.

Szulanski, G., Cappetta, R., & Jensen, R. J. 2004. Whenand how trustworthiness matters: Knowledge trans-

fer and the moderating effect of causal ambiguity.Organization Science, 15: 600–613.

Tsai, W. 2001. Knowledge transfer in intraorganizationalnetworks: Effects of network position and absorptivecapacity on business unit innovation and performance.Academy of Management Journal, 44: 996–1004.

Tushman, M. L. 1978. Technical communication in R&Dlaboratories. The impact of project work characteris-tics. Academy of Management Journal, 21: 624–645.

Tushman, M. L. 1979. Work characteristics and subunitcommunication structure: A contingency analysis.Administrative Science Quarterly, 24: 82–97.

Tushman, M. L., & O’Reilly, C. A. 1996. Ambidextrousorganizations: managing evolutionary and revolu-tionary change. California Management Review,38(4): 8–30.

Venkatraman, N., & Grant, J. H. 1986. Construct measure-ment in organizational strategy research: A critiqueand proposal. Academy of Management Review,11: 71–87.

Wageman, R. 1995. Interdependence and group effective-ness. Administrative Science Quarterly, 40: 145–180.

Wageman, R., Hackman, J. R., & Lehman, E. 2005. Teamdiagnostic survey: Development of an instrument.Journal of Applied Behavioral Science, 41: 373–398.

Wellins, R. S., Byham, W. C., & Wilson, J. M. 1991.Empowered teams: Creating self-managing work-ing groups and the improvement of productivityand participation. San Francisco: Jossey-Bass.

Zellmer-Bruhn, M. E., & Gibson, C. B. 2006. Multina-tional organization context: Implications for teamlearning and performance. Academy of Manage-ment Journal, 49: 501–518.

Martine R. Haas (mrhaas@wharton.upenn.edu) is an as-sociate professor of management at the Wharton Schoolof the University of Pennsylvania. Her research focuseson collaboration in knowledge-intensive organizations,with particular emphasis on team effectiveness, knowl-edge sharing, and global work. She received her Ph.D. inorganizational behavior from Harvard University.

1008 OctoberAcademy of Management Journal