You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6575587

The Development of Collective Efficacy in Teams: A Multilevel and


Longitudinal Perspective

Article  in  Journal of Applied Psychology · February 2007


DOI: 10.1037/0021-9010.92.1.17 · Source: PubMed

CITATIONS READS
199 2,930

3 authors:

Kevin Tasa Simon Taggar


York University Wilfrid Laurier University
25 PUBLICATIONS   732 CITATIONS    36 PUBLICATIONS   1,910 CITATIONS   

SEE PROFILE SEE PROFILE

Gerard Seijts
The University of Western Ontario
68 PUBLICATIONS   2,230 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Political Skill: Teams and Negotiation View project

Goal Setting View project

All content following this page was uploaded by Kevin Tasa on 25 October 2017.

The user has requested enhancement of the downloaded file.


Journal of Applied Psychology Copyright 2007 by the American Psychological Association
2007, Vol. 92, No. 1, 17–27 0021-9010/07/$12.00 DOI: 10.1037/0021-9010.92.1.17

The Development of Collective Efficacy in Teams:


A Multilevel and Longitudinal Perspective

Kevin Tasa Simon Taggar


McMaster University Wilfrid Laurier University

Gerard H. Seijts
University of Western Ontario

The authors developed and tested a longitudinal multilevel model of collective efficacy formation. In 50
self-managing student teams, they investigated the effects of individual-level and team-level factors on
observed behaviors and the subsequent development of collective efficacy for mastering a complex team
task. Self-efficacy for teamwork, task-relevant knowledge, and collective efficacy predicted individual
teamwork behaviors (rated by peers). Aggregated measures of teamwork behavior were related to
subsequent collective efficacy, which was significantly related to final team performance.

Keywords: collective efficacy, team performance, multilevel, self-efficacy, task-relevant knowledge

Teams have become a common feature in many organizations vicarious experience, verbal persuasion, and physiological and
(Cohen & Bailey, 1997). A growing body of research suggests that affective states. However, there is strong reason to believe that the
collective efficacy, which refers to a team’s shared perception of its development of efficacy beliefs may not be isomorphic at the
capability to successfully perform a specific task (Bandura, 1997), individual and team levels. Several researchers have suggested that
influences team outcomes. Locke (1991) theorized that efficacy the factors that shape the evolution of collective efficacy may be
beliefs, in conjunction with goals, constitute the “motivational different from the antecedents of self-efficacy (Chen & Bliese,
hub,” which represents the processes that most directly affect 2002; Gibson, 1999; Gully et al., 2002). Hence, there is a need to
action. According to Bandura (1997), collective efficacy can “in- initiate research that explores the determinants of collective efficacy.
fluence the type of future [people] seek to achieve, how they In this article, we build on Gibson’s (1999) observation that
manage their resources, the plans, and strategies they construct, collective efficacy develops partly through exchanges of informa-
how much effort they put into their group endeavor, their staying tion and observed behaviors within the team. Specifically, our
power when collective efforts fail to produce quick results or model of collective efficacy emergence, depicted in Figure 1,
encounter forcible opposition, and their vulnerability to discour- states that, over time, collective efficacy is influenced by the
agement” (p. 418). Research consistently shows that efficacy be- aggregate amount of teamwork behaviors in a team. As we explain
liefs contribute significantly to the level of motivation and perfor- below, teamwork behaviors refer to individual activities that con-
mance (Bandura & Locke, 2003). tribute to team process. We expect that teams with higher levels of
Although scholars have demonstrated the link between collec- aggregate teamwork behavior will have higher levels of collective
tive efficacy and team performance (Gully, Incalcaterra, Joshi, & efficacy. Thus, our model begins by exploring the factors that
Beaubien, 2002), little is yet known about the factors responsible influence teamwork behaviors at the individual level. We then
for the development of collective efficacy. This is unfortunate assess how these behaviors, when aggregated to the team level,
because a better understanding of the antecedents of collective influence collective efficacy beliefs and subsequent team perfor-
efficacy can help managers and team members more accurately mance. The model is multilevel in nature, in response to Klein and
shape this important shared perception. Kozlowski’s (2000) call for researchers to focus on team research
Bandura (1997) stated that determinants of efficacy beliefs at multiple levels. The model proposes cross-level effects (Rous-
“operate in much the same way at the collective level as they do at seau, 1985), whereby team-level constructs influence individual
the individual level” (p. 478). According to this perspective, col- behavior, which in turn, influences team-level perceptions and
lective efficacy is influenced by enactive mastery experience, performance (Lindsley, Brass, & Thomas, 1995).
In our theorizing we followed Morgeson and Hofmann’s (1999)
guidelines for investigating the emergence of constructs at the
Kevin Tasa, DeGroote School of Business, McMaster University, Ham- collective level. We start at the most elementary unit of analysis in
ilton, Ontario, Canada; Simon Taggar, School of Business and Economics,
any social system—the individual behavioral act (Parsons, 1951).
Wilfrid Laurier University, Waterloo, Ontario, Canada; Gerard H. Seijts,
Richard Ivey School of Business, University of Western Ontario, London,
In our case, these individual acts refer to teamwork behavior.
Ontario, Canada. Moreover, as we will explain, self-efficacy for teamwork and
Correspondence concerning this article should be addressed to Kevin task-relevant knowledge may be determinants of the frequency
Tasa, DeGroote School of Business, McMaster University, Hamilton, with which an individual performs teamwork behaviors. Next, we
Ontario L8S 4M4, Canada. E-mail: tasa@mcmaster.ca consider context. In our longitudinal model, initial collective effi-
17
18 TASA, TAGGAR, AND SEIJTS

H1a Individual
Individual Teamwork
Teamwork
Level Self-efficacy
Behavior
H1b

Task-relevant Additive
knowledge Aggregation

Team-level
Teamwork
Behavior
H1c
Team
H2
Level

Collective Collective H3
Initial Team
Efficacy Efficacy
Performance Performance
T1 T2

Time Period Week 2 Week 5 Week 7 Week 10

Figure 1. Longitudinal multilevel model linking efficacy beliefs, behavior, and team performance. The time
period shown in the figure corresponds to the period in which the measures were assessed. H ⫽ hypothesis; T ⫽
time.

cacy and performance feedback can be considered group-level ture, Stevens and Campion (1994) developed a typology of generic
contextual variables that affect individual behavior, but also affect or transferable knowledge, skills, and abilities (KSAs) necessary
the nature of subsequent team interaction. Multilevel theorists refer for teamwork that focused specifically on the individual team
to this type of relationship as a top-down relationship (Klein & member. They identified two broad categories of teamwork KSAs:
Kozlowski, 2000). Lastly, we consider the role collective efficacy interpersonal KSAs (i.e., conflict resolution, collaborative
plays in team performance. problem-solving, and communication) and self-management
KSAs (e.g., goal-setting and performance management, and plan-
Theoretical Background and Hypotheses ning and task coordination). Although possessing the knowledge,
skills and abilities to perform in a team may be important, it is the
Marks, Mathieu, and Zaccaro (2001) suggested that emergent translation of the KSAs into actual behaviors that ultimately con-
states such as collective efficacy are products of team experiences tributes to team performance. Thus, in this research we rely on
and become new inputs to subsequent processes and outcomes. Stevens and Campion’s typology as a guiding framework for
The reciprocal interplay among team interactions and emergent assessing the visible teamwork behaviors performed by individual
states at the collective level is likely to start with the observable team members. (The behaviors we used are presented in the
behaviors of individual team members. Seeing one’s teammates Appendix.)
perform behaviors that are generally accepted as helpful with As mentioned above, the KSAs identified by Stevens and Cam-
respect to team performance should instill a sense of confidence pion (1994) are fairly generic, or transportable (Cannon-Bowers,
about the team’s capability. Thus, we suggest that collective effi- Tannenbaum, Salas, & Volpe, 1995). Accordingly, behaviors such
cacy is partly shaped by individual behaviors. as problem solving and performance management should contrib-
Teamwork behaviors contribute to the team’s task and goal ute to team performance to varying degrees across a range of team
requirements. We draw attention to individual behaviors, rather tasks. Stevens and Campion also suggested that the generic nature
than the group-level notion of team processes, because individual- of the teamwork KSAs would make them applicable to semiau-
level behaviors are directly observable by other members of a team tonomous, or self-managing, teams, the type of team used in the
and thus are likely to be used by members when assessing the present study.
overall capabilities of the team. Therefore, it is important to
understand the conditions that give rise to the performance of
individual teamwork behaviors. Antecedents of Teamwork Behavior
At least three factors should be related to the extent to which an
Teamwork Behavior
individual engages in teamwork behavior. These factors are task-
What are the individual teamwork behaviors that contribute to relevant knowledge, self-efficacy for teamwork behaviors, and the
team performance? After an extensive review of the teams litera- level of collective efficacy in the team.
DEVELOPMENT OF COLLECTIVE EFFICACY 19

Task-relevant knowledge. One factor that is likely to influence using longitudinal designs support the reciprocal relationship be-
whether an individual displays teamwork behavior is his or her tween feedback and efficacy perceptions at the team level (e.g.,
level of task-relevant knowledge. Marks et al. (2001) argued that Seijts, Latham, & Whyte, 2000; Tasa & Whyte, 2005). In addition,
member expertise is utilized through group processes to determine this relationship might be most predictive when performance feed-
emergent states such as collective efficacy. Similarly, research has back is directed at the team as a whole. For example, when a team
shown that a group has an ability to recognize the task-relevant wins or loses a game or contest, this feedback is likely to directly
expertise residing in the group (Henry, 1995), and that groups with affect collective efficacy because it speaks to whether the team as
a high level of task-relevant knowledge outperform those with low a whole can succeed. Thus, we expect that initial team-level
levels of task-relevant knowledge (LePine, Hollenbeck, Ilgen, & performance feedback is related to collective efficacy, which, in
Hedlund, 1997). Although it is far from clear how expertise is turn, crosses levels and has a direct influence on individual team-
recognized and linked to team performance, it is likely that team work behavior.
members who know more about the team’s task and goal require- The discussion to this point leads to the following hypothesis:
ments will contribute more to discussions about the team’s task
Hypothesis 1: An individual’s teamwork behavior is predicted by (a)
and goals. Therefore, we posit that an individual’s task-relevant
his or her level of teamwork self-efficacy, (b) his or her level of
knowledge is related to individual teamwork behavior.
task-relevant knowledge, and (c) the initial level of collective efficacy
Although knowledge relevant to the completion of the particular in the team.
task in question may motivate a team member’s participation,
social cognitive theory (Bandura, 1997) indicates that two percep-
Teamwork Behavior and Collective Efficacy
tions should be related to a member’s motivation to engage in
teamwork behavior. These are the individual’s perceptions of his Hypothesis 1 describes the relationship between two individual-
or her ability to perform generic teamwork behaviors (i.e., self- level independent variables, a team-level independent variable,
efficacy for teamwork), and perceptions regarding the team’s and the dependent variable of individual behaviors. As Figure 1
possession of the resources necessary for completing the task (i.e., shows, we further suggest that collective efficacy has emergent
the level of collective efficacy of the team). properties that are rooted in the teamwork behaviors displayed by
Self-efficacy for teamwork behavior. Self-efficacy refers to a each team member. Thus, to understand the evolution of collective
person’s belief that he or she can successfully perform the specific efficacy perceptions, we must understand the factors that drive
behaviors necessary for successful task completion (Bandura, individual behavior. In this sense, our model combines top-down
1986). Research has consistently found that self-efficacy beliefs (macro to micro) and bottom-up (micro to macro) perspectives of
determine whether an individual will choose to engage in certain collective efficacy emergence. Next, we describe how between-
behavior, and, if so, how much effort is expended on that behavior group differences in teamwork behaviors may affect the formation
(Bandura, 1997). Therefore, consistent with Bandura’s (1986, of subsequent collective efficacy and team performance.
1997) social cognitive theory, self-efficacy for generic teamwork In forming a self-efficacy belief, a person assesses the availabil-
behavior should predict the generic teamwork behaviors displayed ity of specific resources (e.g., skill level and knowledge of task-
by a team member. For example, a team member with a high sense relevant strategies) and the constraints that apply (e.g., conflicting
of self-efficacy for actively engaging in problem-solving tasks job demands, time) when performing the task (Gist & Mitchell,
should work at constructively solving complex problems, more 1992). Taggar and Seijts (2003) suggested that the process of
than would a team member with low self-efficacy in this dimen- assessing personal and situational resources and constraints de-
sion. scribed by Gist and Mitchell is also relevant at the team level.
Initial collective efficacy. Finally, we consider the contextual Taggar and Seijts theorized that during team interaction, individ-
role that early perceptions of collective efficacy play in determin- uals collect information not only about their own capabilities
ing the teamwork behavior of individual team members. Mowday around teamwork behavior and assumed roles (e.g., leader, facil-
and Sutton (1993) characterized context as “stimuli and phenom- itator, critical evaluator), but also about other team members’
ena that surround and thus exist in the environment external to the task-relevant competencies. Task-relevant behaviors are often
individual, most often at a different level of analysis” (p. 198). demonstrated in a public forum and hence can be readily incorpo-
Because collective efficacy is a comprehensive judgment that is rated into an individual’s assessment of collective efficacy.
derived from knowledge about a team’s capabilities, resources, We expect that in self-managing teams the development of
and constraints, it may exert a top-down influence on individual collective efficacy is, in part, an emergent process that is derived
teamwork behaviors. The extent to which individuals believe their from the observed behaviors and interactions that occur among
team has the capability to produce results or attain goals may affect team members. For example, if a team member observes that his or
the degree to which they themselves will be motivated to demon- her peers are actively participating in team-related activities such
strate the behaviors that are required for effective team perfor- as managing deadlines, coordinating activities, collaboratively
mance. Lindsley et al. (1995) suggested that such a cross-level solving problems, and assuming leadership roles, then that team
effect is possible because efficacious teams create a context in member’s collective efficacy belief should theoretically increase.
which constructive individual behaviors are expected. This view is consistent with Gibson’s (1999) conjecture regarding
It should also be noted that performance feedback is usually the formation of collective efficacy through an exchange and
positively associated with collective efficacy (e.g., Gully et al., assimilation of information. She explained that “group-efficacy
2002). A central tenet of social cognitive theory is that enactive forms as group members collectively acquire, store, manipulate,
mastery (e.g., feedback on performance attainments) and efficacy and exchange information about each other and about their task,
perceptions are reciprocally related (Bandura, 1997). Past studies context, process, and prior performance” (p. 138). Conversely, the
20 TASA, TAGGAR, AND SEIJTS

absence of such teamwork behaviors would cast doubt in the they worked on a longitudinal simulation over a 10-week period.
minds of the team members about the group’s capability to pull The simulation, called the Human Resource Management Simula-
together and successfully complete the task. tion, was developed by Smith and Golden (2001), and it began
When team members consider the overall capabilities of their following the third week of the course. The pedagogical purpose of
team, they may partly base their perceptions of collective efficacy the simulation is to create a realistic context in which students
on beliefs about their teammates’ teamwork-behavior levels. Team must translate course content, covering such issues as selection and
members’ collective efficacy may increase (or decrease) if they reward systems, into team solutions and decisions for improving
observe their colleagues demonstrating (or failing to demonstrate) the performance of the HRM department. All teams were told by
performance-relevant teamwork behaviors. This assessment of the course instructor that their performance on the simulation
other people’s capabilities is one of the antecedents of collective would largely be determined by the extent to which they were able
efficacy that does not function as an antecedent of self-efficacy. to discern linkages among the various HRM topics covered in the
Hypothesis 2: At the team level, aggregated teamwork behavior is course. The course content that was most relevant to the simulation
positively related to subsequent collective efficacy. was presented either prior to group formation (e.g., strategic HRM,
selection and HRM planning) or during the first few weeks of the
Collective Efficacy and Team Performance simulation (e.g., compensation, training and performance ap-
praisal). Furthermore, participants were told that the algorithms
A meta-analysis by Gully et al. (2002) showed that the relation- which determine performance on the simulation are carefully mod-
ship between collective efficacy and team performance was posi- eled from current theories of human resources and management. A
tive and significant (␳ ⫽ .41). According to social cognitive 25-page player’s manual containing information on the nature of
theory, efficacy is a primary determinant of the extent to which the company, industry, and simulation procedures reinforced these
individuals or teams are likely to put forth the effort required to instructions.
perform successfully (Bandura, 1986). Therefore, to ensure that
Teams assumed the role of the human resource management
our model is consistent with past research, we hypothesized that
department in a manufacturing organization and were required to
the level of collective efficacy, assessed after the team members
make weekly operational decisions on issues such as wage in-
had been working together for several weeks, would predict final
creases or decreases, hiring, fringe benefits, performance ap-
team performance.
praisal, and training. Each weekly set of decisions corresponded to
Hypothesis 3: Collective efficacy is positively related to subsequent one quarter of a simulated fiscal year. Teams were given an equal
team performance. yearly budget and thus had to make their decisions under con-
straints of finite resources. Although the simulation permits a team
Method to spend in excess of their budget in a given quarter, teams were
penalized for running a deficit in the last quarter of a given fiscal
Participants
year. Each team performed the simulation for 8 simulated quarters
Participants were 191 business students enrolled in four sections (or 2 fiscal years) over the 10-week period. No simulation deci-
of an upper-level Human Resource Management (HRM) course. sions were made during Week 5, due to midterm exams, or Week
All four sections were taught by the same instructor, who was 7, due to other classroom activities.
blind to the study’s hypotheses. As part of the course requirements, For each of the weekly decision periods, the teams had two
participants were assigned to teams (N ⫽ 50) of 3 or 4 individuals. primary tasks. The first task was to make choices about allocating
To maximize the breadth of within-team expertise and reduce the resources to each of the HRM-specific issues in the simulation.
potential for self-selection bias, we formed teams by randomly Teams were required to submit their decisions on a standardized
drawing names from major-specific lists (e.g., finance, marketing, form. These decisions were then entered into a computer program
engineering) in sequence. In total, 77% of the participants were by the course instructor. The second task was to complete a weekly
enrolled in the business program, and 23% were in a combined “incident report.” Each incident report required the team to make
engineering and management program. The average age of the choices among alternative options on one-time issues such as how
participants was 21.7 years (SD ⫽ 1.1), and 59% of the partici- to design a new job-analysis program or the format of a new
pants were male. compensation scheme. Each incident had discrete options that
Although participants had to work on the simulation as part of could be chosen, and the choice had to be justified in a one-page
a course requirement, participation in the study was voluntary. written summary. These incidents were graded by the course
Participants received a 2% bonus mark in exchange for their instructor and returned for developmental purposes, and had no
self-reported data. The response rates to the questionnaires that we effect on simulation performance.
distributed at Weeks 2 and 7 were, in both instances, 90%. None Performance feedback was returned within three days so that
of the 50 teams had more than one team member who did not each team had ample time to prepare for the next decision period.
respond to either survey. The minimum group size was three
Feedback was primarily quantitative, including such indicators as
persons because the response rate in the three-person groups was
employee morale, product quality, productivity, employee turn-
100%.
over, budget surpluses or deficits, and the number of grievances
filed in the simulation. For each indicator, the teams were provided
Simulation and Procedures with the industry average. More detail on how the omnibus mea-
A primary objective of the study was to assess the development sure of team performance was assessed and presented to the
of collective efficacy over time. We therefore studied the teams as participants is described below in the Measures section.
DEVELOPMENT OF COLLECTIVE EFFICACY 21

A team’s performance on the HRM simulation accounted for performance management domains of the Stevens and Campion
10% of each team member’s final course grade. Individual team (1994) typology.
members had an incentive to actively participate because teams To assess the level of inter-rater agreement on the behavioral
had an opportunity to complete, at the end of the simulation, an ratings, we calculated the rwg coefficient (James, Demaree, &
anonymous peer evaluation justifying why some members should Wolf, 1984) for each participant on each of the 12 behavioral
not share the grade equally. There were no teams that chose to do items. The average coefficient across all ratings and participants
so. One-third of the class time in each week (50 minutes) was was .84, with a range of .38. There is debate regarding an appro-
given to participants to work on the simulation. During a session priate cutoff point for rwg coefficients. In the present study, over
at the end of the course in which we described the nature of the 95% of all rwg coefficients were above the commonly used .70
study, participants suggested, by an informal show of hands, that threshold (see Castro, 2002). We examined the cases in which the
they spent between 30 and 90 minutes per week working on the rwg score was below .70 and found that low scores, usually just
simulation outside of class time. below .70, were not specific to a few teams or individuals. We
The simulation was appropriate for this study for several rea- therefore did not omit low-scoring individuals or teams from the
sons. First, consistent with Wood’s (1986) criteria of task com- sample. In conclusion, team members appeared to agree on which
plexity, the simulation is highly complex in terms of component, individuals demonstrated high levels of the behaviors of interest
dynamic, and coordinative complexity. For example, team perfor- (and which individuals did not).
mance is not increased solely through effort and persistence. For Teamwork self-efficacy. Teamwork self-efficacy was mea-
teams to do well, they had to repeatedly monitor the relationship sured after the 2nd week of group formation. Gist and Mitchell
between changes to a multitude of simulation-specific indicators (1992) argued that efficacy measurement must be tailored to the
and subsequent feedback from the simulation with respect to domain being studied. Therefore, the measure assessed each team
performance. Second, since groups usually outperform individuals member’s degree of confidence in his or her capabilities to per-
on heuristic problem-solving tasks (Laughlin, Bonner & Miner, form the 12 specific behaviors included in the teamwork behavior
2002; Kerr & Tindale, 2004), the longitudinal nature of the sim- scale. Participants were asked to rate the extent to which they felt
ulation provided greater opportunity for team members to question confident that they could exhibit each of those teamwork behav-
assumptions, reframe problems, think about ideas and problems iors. Responses were made on 7-point Likert-type scales. Scale
using novel approaches, stimulate avenues of thought that other scores could range from 1 (not at all) to 7 (very much so). The
group members may not otherwise pursue, and successively im- coefficient alpha for the scale was .80.
prove ideas (Bass & Avolio, 1994; Meyer & Schvaneveldt, 1971; Task-relevant knowledge. Because the course content was rel-
Sosik, Kahai, & Avolio, 1998; Torrance, 1988). Finally, the sim- evant to the performance of the simulation, knowledge of the
ulation enacts realistic and frequently occurring scenarios, in course content was used to represent task-relevant knowledge. As
which groups must make iterative decisions in the context of finite part of the course requirements, participants completed a multiple-
resources. The use of realistic tasks or simulations increases the choice exam covering topics in HRM at Week 5 of the simulation
extent to which results are generalizable to settings beyond the (Week 8 of the course). A measure of task-relevant knowledge was
classroom or laboratory (Locke, 1986). created using the score, ranging from 0 to 100, on the exam.
Collective efficacy. Collective efficacy was assessed twice, at
Weeks 2 and 7 of the simulation. We measured collective efficacy at
Measures
Week 2, rather than at the time of group formation, to allow individ-
Teamwork behaviors. Measures of teamwork behavior were uals to become familiar with their team members, their team roles,
taken near the end of the simulation (Week 7 of the simulation) and the nature of the simulation. Consistent with Bandura’s (1997)
because by this time the participants had attained experience with recommendations, we developed a measure of collective efficacy that
the task and had also had ample time to observe their fellow team assessed both magnitude and strength with regard to achieving pro-
members. Peer appraisals were used for assessing behaviors. Par- gressively more difficult team performance levels. Participants were
ticipants were asked to rate the frequency with which each of their provided with nine performance levels, defined by placement within
fellow team members engaged in 12 teamwork behaviors during the top 45, 40, 35, 30, 25, 20, 15, 10, or 5 teams participating in the
the simulation. Scale scores could range from 1 (almost never) to simulation (e.g., “I believe that the team can finish the simulation in
7 (almost always). This procedure is consistent with Neuman and at least the top 10 teams”). Collective efficacy magnitude was opera-
Wright (1999) and Taggar and Seijts (2003). Item scores were tionalized as the total number of “Yes” answers to the nine items
averaged to derive a composite behavioral measure for each indi- described above (“Yes” ⫽ 1, and “No” ⫽ 0). Collective efficacy
vidual. The coefficient alpha for the behavioral measure was .93. strength was the sum of the rating scores across the nine performance
We drew the 12 items (see Appendix) from a 46-item behavioral levels. The ratings were made in terms of a continuous 100-point scale
observation scale (BOS) developed by Taggar and Brown (2001). (0 ⫽ “no confidence at all,” and 100 ⫽ “complete confidence”).
The 46 items are critical incidents of actually observed perfor- Correlations between the magnitude and strength scores were r ⫽ .85,
mance-relevant team member behaviors displayed in student p ⬍ .01, at Week 2 and r ⫽ .89, p ⬍ .01, at Week 7. Thus, the sums
teams. We assembled the 12-item list based on three consider- of the magnitude and strength scores were independently standard-
ations. First, a 46-item peer-appraisal BOS could have led to rater ized, then summed, to create the measure of collective efficacy.
fatigue because members of most teams would have had to make Team performance. Teams were informed at the outset by the
138 ratings (3 peers ⫻ 46 items). Second, not all items of the course instructor that their performance on the simulation would
Taggar and Brown measure were relevant for the task used in the be determined by combining their rankings on 10 indicators. The
present study. Finally, the 12 items represent the interpersonal and 10 indicators were Absenteeism, Turnover Percentage, Accident
22 TASA, TAGGAR, AND SEIJTS

Rate, Morale, Productivity, Quality Index, Unit Cost, Percentage Teamwork behaviors. Chan (1998) argued that an additive ag-
of Female Employees, Percentage of Minority Employees, and gregation model (using the mean of individual-level variables as a
Grievances. Each indicator had its own index, and each team group-level variable) is appropriate when the theoretical interest is the
started the simulation at the same point on each indicator. For magnitude of an effect at the group level. Following Chan, we used
example, the Morale index began at a score of 50 and could go as the average of individual teamwork behaviors to represent team-level
high as 100. The Turnover index was expressed as a percentage, teamwork behavior. This is because we assumed that the amount of
with higher percentages reflecting poorer results. The 10 indicators teamwork behavior exhibited by each individual member increases
were, on average, highly correlated (r ⫽ .56). The simulation the collective pool of that behavior. Characterizing a team as high in
creates an omnibus measure of performance that can range be- teamwork behavior would mean that, taking the team as a unit,
tween 0 and 100, based on equal weighting of the 10 indicators. members would display high amounts of teamwork behavior. This
This measure was reported to each team, along with the overall does not imply that all team members display high teamwork behav-
average score, after each weekly decision period. This method ior, but merely that there are at least some members whose scores
drew a participant’s attention to the overall performance of the elevate the average for the team (e.g., Neuman & Wright, 1999, p.
HRM department, rather than performance on only one or two 384). As suggested by Bliese (2000), we computed intraclass corre-
indicators. We refer to the omnibus Week 2 measure as initial lations (ICCs) to determine the reliability of the team-level teamwork
performance and the Week 10 measure as team performance. behavior measure. The ICC(1) coefficient, which represents the de-
In addition to the omnibus measure of performance, a rank gree of variability in responses at the individual level that is attributed
ordered measure of performance was publicly presented to all the to team membership, was .30. The ICC(2) coefficient, which repre-
teams during class time (using an overhead projector) at Weeks 2 sents the reliability of the team-level means, was .65. These results
and 10. During Week 2, this rank ordered measure was presented support the aggregation of teamwork behavior.
several days before team members were asked to assess their Collective efficacy. According to Chan (1998), a direct con-
perceptions of collective efficacy. This measure was helpful in sensus model is a composition model in which individual ratings
forming accurate efficacy perceptions because it provided stronger of a higher level phenomenon are used to represent the higher level
social comparison information regarding performance than an construct. To be theoretically valid, direct consensus models must
overall average (see Wood & Bandura, 1989). The rank measure contain a suitable degree of within-group agreement (Chan, 1998;
correlated highly with the omnibus measure at both time periods Gibson, Randel & Earley, 2000). The ICC(1) coefficients for the
(r ⫽ .82, p ⬍ .01 at Week 2; r ⫽ .83, p ⬍ .01 at Week 10). collective efficacy measures were .38 at Week 2 and .46 at Week
7, and the ICC(2) coefficients were .71 and .77, respectively.
Data Aggregation Issues These results show that collective efficacy displayed an appropri-
ate degree of within-group agreement relative to between-group
Two variables in this study, teamwork behavior and collective
variance, and thus support aggregation of individual-level data to
efficacy, were measured at the individual level and subsequently
the team level (Klein & Kozlowski, 2000).
aggregated to represent a group-level construct. To justify our
procedure for treating variables measured at the individual level as
having group-level properties, we relied on Chan’s (1998) typol- Results
ogy of composition models. Below we explain for each variable
the relevance of Chan’s typology to our data, and the procedures Table 1 presents descriptive statistics and bivariate correlations
we used to illustrate the group-level properties of the aggregated among study variables at the individual and team levels. Each of
measures. the reported correlations is in the expected direction.

Table 1
Descriptive Statistics, Correlations, and Reliability Coefficients

Variables M SD 1 2 3 4 5

Individual-level measures

1. Teamwork self-efficacy 5.23 0.72 (.80)


2. Teamwork behavior 4.81 0.92 .25** (.93)
3. Task-relevant knowledge 71.74 9.29 .20* .24** —

Team-level measures

1. Average teamwork behavior 4.81 0.64 —


2. Collective efficacy (Week 2) 0.01 1.45 .33* —
3. Collective efficacy (Week 7) 0.05 1.52 .44** .53** —
4. Initial performance (Week 2) 54.55 10.04 .20 .40** .48** —
5. Team performance (Week 10) 58.40 12.32 .29* .29* .57** .33* —

Note. For individual-level measures, n ⫽ 175; for team-level measures, n ⫽ 50. Numbers in parentheses are
coefficient alphas.
*
p ⬍ .05. ** p ⬍ .01.
DEVELOPMENT OF COLLECTIVE EFFICACY 23

Because the dependent variable in Hypothesis 1 was at the As shown in Figure 2, we tested these hypotheses within the
individual level, and the independent variables were both individ- context of a structural equation model based on the team-level
ual and team level, we analyzed the data using hierarchical linear relationships proposed in Figure 1. We did not use a latent variable
modeling (HLM 5; Bryk & Raudenbush, 1992). As suggested by approach due to sample size limitations and because two of the
Hofmann, Griffin, and Gavin (2000), we first tested a null model five variables (initial and final performance) were directly mea-
that had no predictors at either Level 1 (the individual level) or sured. Instead, we tested the model as a path analysis using
Level 2 (the team level). This null model is analogous to an structural equation modeling. The first model, based on the paths
analysis of variance in that it partitions variance in teamwork shown in Figure 1, fit the data only moderately well, ␹2(5, N ⫽ 50)
behavior into within-team (⌽2) and between-team (␶) components. ⫽ 11.51, p ⬍ .05 (root-mean-square error of approximation
The estimate of between-team variance is the intraclass correlation [RMSEA] ⫽ .16; goodness-of-fit index [GFI] ⫽ .92; comparative
coefficient and is computed as the proportion of between-teams fit index [CFI] ⫽ .92).
over the total variance (⌽2 ⫹ ␶). A test of the significance level of Because social cognitive theory states that performance and
the Level 2 residual variance of the intercept (␶00 ⫽ .24, p ⬍ .01) efficacy are closely related, we added an additional path from
was significant. Also, the ICC(1) was .28, indicating that 28% of initial performance to collective efficacy at Week 7. This second
the variance in individual teamwork behavior resided between model resulted in an improved degree of fit to the data, ␹2(4, N ⫽
teams, and 72% of the variance resided within teams. The signif- 50) ⫽ 4.26, p ⬍ .37 (RMSEA ⫽ .04; GFI ⫽ .97; CFI ⫽ .99), and
icant between-team variance provided justification for testing the it suggests that initial performance relates not only to initial
full model. collective efficacy, but also to collective efficacy that is assessed
five weeks later. Figure 2 shows that all of the paths in the model
Hypotheses 1a to 1c are statistically significant. In particular, the significant paths from
teamwork behavior to collective efficacy and from collective ef-
Hypotheses 1a to 1c predict that individual teamwork behavior ficacy to final team performance support Hypotheses 2 and 3.
is related to individual-level variables (self-efficacy and task-
relevant knowledge) and a team-level variable (collective effi-
Discussion
cacy). We estimated an HLM model in which the individual-level
antecedents were the Level 1 predictors, and then regressed the A growing body of research suggests that collective efficacy
intercept coefficient obtained from Level 1 on the Level 2 variable, does for teams what self-efficacy does for individuals. That is,
collective efficacy. As Table 2 shows, teamwork self-efficacy collective efficacy has properties that relate to teams through such
(␤ ⫽ .25, p ⬍ .05), task-relevant knowledge (␤ ⫽ .03, p ⬍ .01), motivational elements as choice of activities, effort expended, and
and collective efficacy (␥ ⫽ .11, p ⬍ .05) demonstrated a signif- degree of persistence when results fail to meet expectations (Kan-
icant relationship with individual teamwork behavior. Hence, Hy- fer, 1990). Thus, the question of how collective efficacy develops
potheses 1a to 1c were supported. over time has both theoretical and practical implications. Our study
We also calculated the total variance explained by the full model extends past research by relating teamwork behavior to the emer-
(R2total) using the following formula (Bryk & Raudenbush, 1992): gence of collective efficacy from a cross-level perspective.
2
Rwithin-team ⫻ [1 – ICC(1)] ⫹ Rbetween-team
2
⫻ ICC(1). The two
individual-level antecedents explained 22% of the within-team Theoretical Implications
variance in teamwork behavior, and collective efficacy explained
55% of the between-team variance. Thus, using the R2total formula, The first theoretical implication of the present study pertains to
the full model explained 31% of the total variance in teamwork the determinants of self-efficacy and collective-efficacy beliefs in
behavior. teams. Although self-efficacy and collective efficacy are each
related to performance, the antecedents of self-efficacy and
collective-efficacy perceptions can differ, depending on the level
Hypotheses 2 and 3
of analysis and the criterion for measuring performance. Self-
Hypothesis 2 states that aggregated teamwork behavior is sig- efficacy processes are conceptualized as occurring primarily at the
nificantly related to collective efficacy, and Hypothesis 3 states intraindividual level via social and cognitive processes (Bandura,
that collective efficacy is positively related to team performance. 1997). For example, individuals working within teams will assess

Table 2
Hierarchical Linear Modeling Results for Individual Teamwork Behavior

Variable Coefficient SE t

Intercept 4.79 .08 56.93**


Teamwork self-efficacy (Hypothesis 1a) 0.25 .10 2.40* (df ⫽ 171)
Task-relevant knowledge (Hypothesis 1b) 0.03 .01 3.56** (df ⫽ 171)
Collective efficacy (Week 2; Hypothesis 1c) 0.12 .06 2.01* (df ⫽ 48)

Note. Individuals n ⫽ 175, teams n ⫽ 50. Entries are estimations of the fixed effects with robust standard
errors.
*
p ⬍ .05. ** p ⬍ .01.
24 TASA, TAGGAR, AND SEIJTS

Team-level
Teamwork
Behavior

.28*

.39* .33* .58**


Collective Collective
Initial Team
Efficacy Efficacy
Performance Performance
T1 T2

.32*

Time Period Week 2 Week 7 Week 10

Figure 2. Results of the team-level structural equation model. All of the reported paths are standardized
coefficients. T ⫽ time. * p ⬍ .05. ** p ⬍ .01.

their self-efficacy for teamwork by considering their own capabil- visible teamwork behaviors within the team may create or continue
ities and limitations for performing whatever task-demands are spirals because they are assessed and incorporated into future
placed on them. Collective efficacy development, on the other perceptions of collective efficacy.
hand, occurs primarily at the interindividual level. Collective ef- Finally, our results support Marks et al. (2001), who claimed
ficacy evolves, in part, as the members of the team monitor that emergent states such as collective efficacy are not processes in
performance feedback and the extent to which the other members and of themselves because they do not describe how group mem-
of the team engage in teamwork behaviors. Our results suggest that bers interact while contributing to goal accomplishment. Rather,
to understand the evolution of collective efficacy, we must assess teamwork behaviors serve as inputs into the collective develop-
the factors that influence teamwork behavior. ment of perceptions of group capability.
A second contribution of this study is the finding that collective
efficacy appears to contextualize individual teamwork behavior. In Limitations and Future Research
this study, initial collective efficacy was significantly associated
with the frequency with which individuals in the team engaged in Several potential limitations of this study warrant consideration.
such activities as goal monitoring and developing action plans. First, as predicted, performance feedback influenced ratings of
Therefore, it appears that individuals are more likely to be moti- collective efficacy. This finding might raise concerns regarding the
vated to engage in teamwork behaviors when they find themselves extent to which individuals may have rated their group’s efficacy
in efficacious groups. in a biased manner. Indeed, research on the performance-cue bias
When performance feedback is directed to the team as a whole, (see Staw, 1975; Lord & Maher, 1989; Martell & Evans, 2005) has
as it was in this study, the contextual role of collective efficacy as found that evaluative feedback systematically biases individuals’
it relates to individual behavior is likely strongest. The underlying ratings of group process. It is important to recognize, however, that
assumption in treating team performance feedback as a team-level the performance feedback in our study was veridical, and thus
variable is that team members who receive common feedback are departs significantly from the methods applied by Staw and others,
exposed to similar evidence that their team can, or cannot, muster in which group performance is held constant and the feedback
whatever it takes to succeed. Thus, we expect that team members provided by the experimenter is deliberately false. Consequently,
are more likely to have a shared sense of collective efficacy when these two streams of research differ significantly in both their
performance feedback is team level than when performance feed- goals and their methodology. Bearing this in mind, there is little
back is individual level. reason to doubt that group performance levels affected perceptions
A third theoretical implication relates to linkages between per- of group efficacy, just as they do perceptions of individual self-
formance and collective efficacy. Our results support the notion of efficacy (Bandura, 1997).
performance and efficacy spirals proposed by Hackman (1990) Second, the nature of the simulation and the fact that the teams
and further developed by Lindsley et al. (1995). Consistent with were self-managing raises the possibility that the person with the
the view that groups high in collective efficacy are likely to use most HRM knowledge made all the team’s decisions. This would
high-quality group processes, perform well, and yield increasing be consistent with a disjunctive aggregation model in Steiner’s
collective efficacy and performance over time (Lindsley et al., (1972) task typology. On such tasks, only one person needs to
1995), we found that high collective efficacy appears to motivate perform well in order for the team to succeed. To test this possi-
greater performance-relevant team-member behavior in individual bility in our study, we examined the relationship between HRM
team members. Thus, collective efficacy motivates, or demoti- knowledge, as measured by the midterm exam result of the highest
vates, individual team member behavior. Our model suggests that scoring team member, and team performance. The correlation
DEVELOPMENT OF COLLECTIVE EFFICACY 25

between the midterm grade of the highest scoring group member Finally, because the independent variables were not manipu-
and team performance was not statistically significant (r ⫽ .13, lated, we cannot say with certainty that they “caused” changes in
p ⫽ .32). According to Steiner (1972), an additive model is one the dependent variables. In future research, we suggest using
whereby the mean level of a variable (in this case HRM knowl- interventions that manipulate teamwork behaviors to determine the
edge) may be the most appropriate aggregation model. An additive nature of their relationship with collective efficacy. In this way we
task is structured such that the contribution of each member on a will be able to increase our confidence that these variables cause
variable may be added to a collective pool of that variable, which change in collective efficacy.
can be drawn upon to contribute to team performance. The corre-
lation between average team score on the midterm and team Practical Implications
performance yielded a higher correlation coefficient (r ⫽ .17, p ⫽
.25) than the disjunctive model but was still not statistically sig- Our finding that teamwork behavior was significantly related to
nificant. collective efficacy, even after controlling for initial collective
These results suggest that expertise, as measured by task- efficacy, speaks to the practical importance of teamwork behavior
relevant knowledge, does not appear to drive team performance. and interventions that encourage such behavior. Most complex
Although research has shown that the relationship between collec- political and organizational decisions are made in groups because
tive efficacy and performance is strongest under conditions of high they frequently outperform individuals in complex problem-
task interdependence (Gully et al., 2002), we know relatively little solving tasks (Henry, 1995; Laughlin et al., 2002; Kerr & Tindale,
about how different task configurations relate to collective efficacy 2004). Our results suggest that to capitalize on the potential ad-
and team outcomes. Thus, future research on the relationship vantages offered by groups, organizations would be wise to ac-
between collective efficacy emergence and task type is clearly tively develop good teamwork behavior rather than focus solely on
warranted. Similarly, we also suggest that future research should group outcome issues (e.g., improved efficiency or reduced error
explore whether teamwork behaviors relate to collective efficacy rates). Researchers have developed typologies to delineate the
in teams that have fewer opportunities for interaction (e.g., virtual diverse types of behaviors found in different teams, with an aim
teams). toward determining what leads to team effectiveness (e.g., Stevens
& Campion, 1994; Hyatt & Ruddy, 1997; Taggar & Brown, 2001).
Third, the study participants were not HRM experts, and a
We note that there are multiple ways to encourage the display of
common criticism of studies using student samples is that the
these behaviors. Training programs that teach appropriate team-
results may fail to generalize to business settings. In discussing
work behavior are one way to increase teamwork skills. For
generalizability with respect to this study, it is essential to consider
example, Chen, Donahue, and Klimoski (2004) found that a uni-
the fact that collective efficacy is a perception that explicitly
versity course geared towards teaching individual teamwork KSAs
incorporates the group’s context and constraints. Though teams in
had a significant impact on individual learning of the KSAs.
business settings are likely to face different, and perhaps more,
Managers who are creating self-managing teams can also work
constraints than the student teams in our study (e.g., budgets, time,
to increase initial collective efficacy by selecting individuals who
policies, turnover), we note that such constraints are included in
have high self-efficacy for teamwork and high task-relevant
collective efficacy perceptions. We also note that past studies of
knowledge. In other words, designing effective teams should be
collective efficacy have found it to be a robust predictor of team given the attention it deserves; too often, teams are loosely formed
performance in a wide range of field and laboratory settings (Gully without consideration of the goals the team is setting out to attain
et al., 2002). Nevertheless, the generalizability of our results must or the KSAs that are needed. The result might be that the team
be assumed to be limited until empirical research explores other starts off at a performance disadvantage. Other interventions to
variables that may relate to teamwork behavior and collective encourage team behaviors could include behavioral modeling by
efficacy. the team leader (e.g., Lester, Meglino, & Korsgaard, 2002) and the
Fourth, our results support the assertion that collective efficacy fostering of a strong team culture, where peer pressure encourages
plays a contextual role in shaping individual teamwork behavior. It individuals to demonstrate particular team behaviors.
remains to be seen whether higher level organizational variables, Finally, we highlight the practical implications of creating per-
such as organizational culture and perceived managerial support, ceptions of early success within the team. As our results showed,
relate to the emergence of collective efficacy. A recent study by teams that had better initial performance had higher initial collec-
Chen and Bliese (2002) found that leadership climate exerted a tive efficacy. Initial collective efficacy, in turn, crossed levels and
top-down influence on collective efficacy. Additional research had a significant impact on individual teamwork behavior. For
examining top-down processes at a higher level than the team optimal performance from team members, managers should there-
would not only enhance the generalizability of the current litera- fore encourage perceptions of “early wins.” It appears that a high
ture, but also respond to Johns’s (2006) call to increase the level of level of collective efficacy brings out the best in individuals in
contextual richness in organizational research. terms of their teamwork behavior. Accordingly, it is in the orga-
Fifth, there may be partial same-source bias in our results nization’s best interests to positively shape the initial level of
because the same team members that assessed collective efficacy collective efficacy within the team.
also completed the peer ratings of teamwork behavior. We are In conclusion, our findings are consistent with the theory of
cautiously optimistic, however, regarding the validity of the indi- collective structure, which states that “it is not the collective
vidual teamwork behavior measure because it was significantly construct, per se, that determines the behavior of individuals—
correlated with measures that were not rated by the same source, rather, it is the individuals (or collective) who determine the
such as self-efficacy for teamwork and task-relevant knowledge. collective construct, and, through their actions, influence the be-
26 TASA, TAGGAR, AND SEIJTS

havior of others in the collective” (Morgeson & Hofmann, 1999, p. meta-analysis of team-efficacy, potency, and performance: Interdepen-
251). Our results also suggest that initial levels of collective dence and level of analysis as moderators of observed relationships.
efficacy are likely to have a long-lasting effect on how well teams Journal of Applied Psychology, 87, 819 – 832.
function and perform over time. Thus, understanding sources of Hackman, J. R. (Ed.). (1990). Groups that work (and those that don’t):
collective efficacy is critically important for predicting the long- Creating conditions for effective teamwork. San Francisco: Jossey-Bass.
term prospects of the success of teams. Furthermore, collective Henry, R. A. (1995). Improving group judgment accuracy: Information
sharing and determining the best member. Organizational Behavior and
efficacy will change as a result of both performance feedback and
Human Decision Processes, 62, 190 –197.
the behavior of team members. Even a highly efficacious team can
Hofmann, D. A., Griffin, M. A., & Gavin, M. B. (2000). The application
experience declining collective efficacy if its members do not of hierarchical linear modeling to organizational research. In K. J. Klein
continue to actively participate. Poor teamwork behavior is likely & S. W. Kozlowski (Eds.). Multilevel theory, research, and methods in
to result in a pattern of decreasing efficacy and performance over organizations: Foundations, extensions, and new directions (pp. 467–
time, in what Lindsley et al. (1995) termed “efficacy-performance 511). San Francisco: Jossey-Bass.
spirals.” On the other hand, such spirals can be positive and Hyatt, D. E., & Ruddy, T. M. (1997). An examination of the relationship
characterized by increasing levels of collective efficacy and per- between work group characteristics and performance: Once more into
formance. the breech. Personnel Psychology, 50, 553–585.
James, L. R., Demaree, R. G., & Wolf, G. (1984). Estimating within-group
interrater reliability with and without response bias. Journal of Applied
References Psychology, 69, 85–98.
Bandura, A. (1986). Social foundations of thought and action: A social Johns, G. (2006). The essential impact of context on organizational behav-
cognitive theory. Englewood Cliffs, NJ: Prentice-Hall. ior. Academy of Management Review, 31, 386 – 408.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Kanfer, R. (1990). Motivation theory and industrial/organizational psy-
W. H. Freeman. chology. In M. D. Dunnette (Ed.), Handbook of industrial and organi-
Bandura, A., & Locke, E. A. (2003). Negative self-efficacy and goal effects zational psychology (Vol. 1, pp. 75–170). Palo Alto, CA: Consulting
revisited. Journal of Applied Psychology, 88, 87–99. Psychologists Press.
Bass, B. M., & Avolio, B. J. (1994). Improving organizational effective- Kerr, N. L., & Tindale, R. S. (2004). Group performance and decision
ness through transformational leadership. Thousand Oaks, CA: Sage. making. Annual Review of Psychology, 55, 623– 655.
Bliese, P. D. (2000). Within-group agreement, non-independence, and Klein, K. J., & Kozlowski, S. W. J. (Eds.). (2000). Multilevel theory,
reliability: Implications for data aggregation and analysis. In K. J. Klein research, and methods in organizations: Foundations, extensions, and
& S. W. Kozlowski (Eds.), Multilevel theory, research, and methods in new directions. San Francisco: Jossey-Bass.
organizations: Foundations, extensions, and new directions (pp. 349 – Laughlin, P. R., Bonner, B. L., & Miner, A. G. (2002). Groups perform
381). San Francisco: Jossey-Bass. better than the best individuals on Letters-to-Numbers problems. Orga-
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: nizational Behavior and Human Decision Processes, 88, 605– 620.
Applications and data analysis methods. Newbury Park, CA: Sage. LePine, J. A., Hollenbeck, J. R., Ilgen, D. R., & Hedlund, J. (1997). Effects
Cannon-Bowers, J. A., Tannenbaum, S. I., Salas, E., & Volpe, C. E. (1995). of individual differences on the performance of hierarchical decision-
Defining competencies and establishing team training requirements. In making teams: Much more than g. Journal of Applied Psychology, 82,
R. A. Guzzo & E. Salas (Eds.), Team effectiveness and decision making 803– 811.
in organizations (pp. 333–380). San Francisco: Jossey-Bass. Lester, S. W., Meglino, B. M., & Korsgaard, M. A. (2002). The anteced-
Castro, S. L. (2002). Data analytic methods for the analysis of multilevel ents and consequences of group potency: A longitudinal investigation of
questions: A comparison of intraclass correlation coefficients, rwg(j), newly formed work groups. Academy of Management Journal, 45,
hierarchical linear modeling, within- and between-analysis, and random 352–368.
group resampling. Leadership Quarterly, 13, 69 –93. Lindsley, D. H., Brass, D. J., & Thomas, J. B. (1995). Efficacy–perfor-
Chan, D. (1998). Functional relations among constructs in the same content
mance spirals: A multilevel perspective. Academy of Management Re-
domain at different levels of analysis: A typology of composition mod-
view, 20, 645– 678.
els. Journal of Applied Psychology, 83, 234 –246.
Locke, E. A. (1986). Generalizing from laboratory to field settings. Lex-
Chen, G. L., & Bliese, P. D. (2002). The role of different levels of
ington, MA: Lexington Books.
leadership in predicting self- and collective efficacy: Evidence for dis-
Locke, E. A. (1991). The motivation sequence, the motivation hub, and the
continuity. Journal of Applied Psychology, 87, 549 –556.
motivation core. Organizational Behavior and Human Decision Pro-
Chen, G. L., Donahue, L. M., & Klimoski, R. J. (2004). Training under-
cesses, 50, 288 –299.
graduates to work in organizational teams. Academy of Management
Learning and Education, 3, 27– 40. Lord, R. G., & Maher, K. J. (1989). Cognitive processes in industrial and
Cohen, S. G., & Bailey, D. E. (1997). What makes teams work: Group organizational psychology. In C. L. Cooper & I. Robertson (Eds.),
effectiveness research from the shop floor to the executive suite. Journal International review of industrial and organizational psychology (pp.
of Management, 23, 239 –290. 49 –91). New York: Wiley.
Gibson, C. B. (1999). Do they do what they believe they can? Group Marks, M. A., Mathieu, J. E., & Zaccaro, S. J. (2001). A temporally based
efficacy and group effectiveness across tasks and cultures. Academy of framework and taxonomy of team processes. Academy of Management
Management Journal, 42, 138 –152. Review, 26, 356 –376.
Gibson, C. B., Randel, A. E., & Earley, P. C. (2000). Understanding group Martell, R. F., & Evans, D. P. (2005). Source-monitoring training: Toward
efficacy: An empirical test of multiple assessment methods. Group & reducing rater expectancy effects in behavioral measurement. Journal of
Organization Management, 25, 67–97. Applied Psychology, 90, 956 –963.
Gist, M. E., & Mitchell, T. R. (1992). Self-efficacy: A theoretical analysis Meyer, D. E., & Schvaneveldt, R. W. (1971). Facilitation in recognizing
of its determinants and malleability. Academy of Management Review, pairs of words: Evidence of dependence between retrieval operations.
17, 183–211. Journal of Experimental Psychology, 90, 227–234.
Gully, S. M., Incalcaterra, K. A., Joshi, A., & Beaubien, J. M. (2002). A Morgeson, F. P., & Hofmann, D. A. (1999). The structure and function of
DEVELOPMENT OF COLLECTIVE EFFICACY 27

collective constructs: Implications for multilevel research and theory Steiner, I. D. (1972). Group process and productivity. New York: Aca-
development. Academy of Management Review, 24, 249 –250. demic Press.
Mowday, R. T., & Sutton, R. I. (1993). Organizational behavior: Linking Stevens, M. J., & Campion, M. A. (1994). The knowledge, skill, and ability
individuals and groups to organizational contexts. Annual Review of requirements for teamwork: Implications for human resource manage-
Psychology, 44, 195–229. ment. Journal of Management, 20, 503–530.
Neuman, G. A., & Wright, J. (1999). Team effectiveness: Beyond skills Taggar, S., & Brown, T. C. (2001). Problem-solving team behaviors:
and cognitive ability. Journal of Applied Psychology, 84, 376 –389. Development and validation of BOS and a hierarchical factor structure.
Parsons, T. (1951). The social system. Glencoe, IL: Free Press. Small Group Research, 32, 698 –726.
Rousseau, D. M. (1985). Issues of level in organizational research: Mul- Taggar, S., & Seijts, G. H. (2003). Leader and staff role-efficacy as
tilevel and cross-level perspectives. In L. L. Cummings & B. M. Staw
antecedents of collective efficacy and team performance. Human Per-
(Eds.), Research in organizational behavior: Vol. 7 (pp. 1–37). Green-
formance, 16, 131–156.
wich, CT: JAI Press.
Tasa, K., & Whyte, G. (2005). Collective efficacy and vigilant problem
Seijts, G. H., Latham, G. P., & Whyte, G. (2000). Effect of self- and group
solving in group decision making: A non-linear model. Organizational
efficacy on group performance in a mixed-motive situation. Human
Behavior and Human Decision Processes, 96, 119 –129.
Performance, 13, 279 –298.
Smith, J. R., & Golden, P. A. (2001). Human resource management Torrance, E. P. (1988). The nature of creativity as manifest in its testing.
simulation: Player’s manual. Upper Saddle River, NJ: Prentice Hall. In R. J. Sternberg (Ed.), The nature of creativity: Contemporary psy-
Sosik, J. J., Kahai, S. S., & Avolio, B. J. (1998). Transformational chological perspectives (pp. 43–75). New York: Cambridge University
leadership and dimensions of creativity: Motivating idea generation Press.
in computer-mediated groups. Creativity Research Journal, 11, 111– Wood, R. E. (1986). Task complexity: Definition of the construct. Orga-
121. nizational Behavior and Human Decision Processes, 37, 60 – 82.
Staw, B. M. (1975). Attribution of the “causes” of performance: A general Wood, R. E., & Bandura, A. (1989). Impact of conceptions of ability on
alternative interpretation of cross-sectional research on organizations. self-regulatory mechanisms and complex decision making. Journal of
Organizational Behavior and Human Processes, 13, 414 – 432. Personality and Social Psychology, 56, 407– 415.

Appendix

Self-Efficacy for Teamwork and Teamwork Behavior Questionnaire Items

Reflecting on your past experience in all of the teams that you 8. Ignore or discourage off-topic conversations
have been on, please answer the following questions regarding the
extent to which you personally feel capable of exhibiting the below 9. Steer team members towards on-topic conversations
behaviors. (1 ⫽ not at all; 7 ⫽ very much so)
10. Address conflict immediately by raising it for discus-
1. Set time deadlines for achieving tasks sion with other team members

2. Take steps to ensure everyone participates in group 11. Try to calm down team members that are in conflict
discussions
12. Assume leadership
3. Take the group’s ideas and develop specific plans of action
Note. For the peer-rated teamwork behavior measure, each
4. Make correct judgments about connections in complex participant was asked to rate their fellow team members according
situations to the frequency with which they displayed each of the 12 behav-
iors (1 ⫽ almost never, 7 ⫽ almost always).
5. Participate in developing strategies to achieve team goals

6. Remind other team members of the team’s goal


Received October 28, 2005
7. Draw team members into discussions that are relevant to Revision received March 17, 2006
achieving the goal Accepted March 27, 2006 䡲

View publication stats

You might also like