group-level relative performance information 1
TRANSCRIPT
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 1
The Effects of Group-Level Relative Performance Information on Group Task
Performance
Robert Marleya
Uday S. Murthyb
aGeorgia Southern University School of Accountancy
P.O. Box 8141 Statesboro, GA 30460-8141
(912) 478-8008 [email protected]
bUniversity of South Florida
School of Accountancy 4202 East Fowler Avenue, BSN 3403
Tampa, FL 33620-7800 (813) 974-6516
March 2013
Under review at Organizational Behavior and Human Decision Processes.
Please do not distribute or cite without the authors’ permission.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 2
Abstract
This study investigates whether providing group-level relative performance information (RPI) to individuals in an effort sensitive, non-interdependent task yields incremental performance value beyond that associated with group identity. We develop theory to predict the social comparison process will foster inter-group comparisons, leading group members to exert greater effort to improve their group’s relative ranking, even when individual group members’ effort is not separately identifiable. We design an experiment where relative performance information and group identity are manipulated. Consistent with our predictions, we find that providing group-level RPI yields positive group performance improvement, particularly among groups of average ability. Further, we find that providing group-level RPI yields incremental group performance improvement beyond that attributable to group identity among groups of average ability.
Keywords: Effort sensitive task; non-interdependent task; group identity; relative performance
information; social comparison theory
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 3
The Effects of Group-Level Relative Performance Information on Group Task Performance
Introduction
Organizations frequently assign individuals to groups for work-related tasks (IMA, 1999).
Firms use groups to obtain a variety of benefits including better task performance, greater
employee participation, increased process attention, and higher employee satisfaction (Wellins,
Byham, & Dixon, 1994). Consequently, studies that investigate the efficacy of mechanisms
aimed at improving the performance of individuals within a group context should be of interest to
organizations.1 In this study, we investigate providing group-level relative performance
information (RPI) improves task performance in a group setting. We also investigate whether
providing group-level RPI yields performance improvements when employed in conjunction with
an often-used non-monetary mechanism for enhancing group performance—fostering a sense of
group identity.
Relative performance information (RPI) is feedback that ranks the recipient’s
performance in comparison to others. Prior research finds that providing RPI provided at the
individual level affects task performance (Hannan, McPhee, Newman, & Tafkov, 2012; Newman
& Tavkov, 2012; Murthy & Schafer, 2011; Young & Lindquist,1993). Group-level RPI is
feedback that communicates how the performance of the individual’s group compares with that
of other groups. Drawing on social comparison theory (Festinger, 1954), we examine whether
task performance of groups who receive such RPI is superior to that of groups who do not
1 In this study, the term “group” refers to several individuals working towards a common goal. This definition is similar to Shaw (1976, p. 8), who, after reviewing 80 group characterizations, defined a group as “two or more persons who are interacting with one another in such a manner that each person influences and is influenced by others.”
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 4
receive RPI.2 Per social comparison theory (SCT), providing group-level RPI should foster inter-
group social comparisons, leading group members to exert greater effort to improve the
performance of their group and in turn improve the group’s relative ranking.
Although prior research finds positive performance effects associated with providing
individuals with RPI, providing group-level RPI may not yield positive performance effects
when individual members’ effort is not separately identifiable within the group. Thus,
information communicating only how a group’s performance ranks relative to other groups may
not motivate individual members of the group to improve their task performance, since each
individual’s task-specific performance is not identifiable to other individuals.
A means commonly used by organizations to improve group productivity is to foster a
sense of group identity. Organizations expend considerable resources on “team building”
exercises designed to heighten group identity. American firms spent over $20 billion dollars in
2005 alone on team building exercises (Business Magazine, 2006). Research has documented that
group’s with a high sense of group identity outperform those with a low sense of identity (Towry,
2003; Worchel, Rothgerber, Day, Hart,& Butemeyer, 1998; James & Greenberg, 1989). In this
study, we examine the joint and separate effects of group identity and RPI in an effort-based
production task. That is, we investigate whether providing group-level RPI has incremental
performance effects beyond those attributable to group identity. Since RPI becomes increasingly
useful for social comparison purposes as the relationship between ability and task performance
2 To isolate the effects of group-level RPI, we do not also provide individual-level RPI. Providing both individual- level and group-level RPI would make it difficult to disentangle the two effects, since both kinds of RPI could affect performance.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 5
becomes stronger we also examine task performance effects of group-level RPI for different
group ability levels (Tafkov, 2009; Martin, 2000).
Investigating the performance effects of providing group-level RPI in a non-
interdependent task where an individual’s performance is not directly observable is important
because of its significant academic and organizational implications. From an academic
perspective, literature in economics and psychology suggest that providing individuals with
group-level RPI should not have a performance effect due to the lack of identifiability (Karau &
Williams, 1993; Williams, Nida, Baca, & Latane, 1989; Williams, Harkins, & Latane,1981;
Latane, Williams, & Harkins, 1979; Alchian & Demsetz, 1972).3 The arguments from this
literature are based on the free riding hypothesis, which posits that individual members can loaf
without penalty when individual performance is not directly observable. We argue that the social
comparison and group identity tendencies will outweigh the social loafing tendency, thus causing
group-level RPI and social identity interventions to foster performance improvements.
From an organizational perspective, although it is common for organizations to place
individuals who perform interdependent tasks into groups to improve task performance, research
demonstrates that placing individuals into groups also improves task performance when they are
assigned to perform non- interdependent, effort-sensitive tasks (Jouini, Dallery, & Nait-
Abdallah, 2008). In such a setting, whether providing group- level RPI to individuals will
improve task performance is an empirical question that is the subject of the current investigation.
We further investigate whether providing group-level RPI offers incremental value above and
3 This study defines “identifiability” as the ability for individuals to discern the task performance of each other in a production-related task setting
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 6
beyond that provided by heightening the sense of group identity, a treatment commonly used by
organizations to improve task performance.
The context of the study is a multi-person, multi-period repetitive letter decoding task in a
computerized environment with undergraduate business students as proxies for employees in a 2
x 2 experiment where group-level relative performance information and group identity are
manipulated. The first manipulated factor is group-level RPI, which is either present
(electronically displayed to groups at the end of every decoding session) or absent. The second
manipulated factor is group identity, which is either high or low. In the high-identity condition,
participants are randomly assigned to groups where each group member introduces him/herself to
other group members, sits in close proximity to other group members, and adorns the group’s
color card. In the low-identity condition, participants are randomly assigned to groups via a
textual, on-screen announcement made on the participant’s laptop.4
The results indicate that providing group-level RPI to individuals who perform a non-
interdependent task in a group setting yields positive group performance improvement,
particularly among groups of average task ability. Consistent with existing literature, fostering a
sense of group identity increased the task performance effects of groups of all task ability levels.
Further, providing group-level RPI was not found to yield positive group performance
improvement effects beyond those attributable to group identity among groups of high- and low-
task ability. While members of high-ability groups may have been bound by ceiling effects, we
speculate that members of low-ability groups were comforted in the knowledge that they are not
alone at the bottom and hence felt little incentive to improve performance. Importantly, we found 4 All forms of communication were prohibited in both high-and-low identity treatments to control for communication effects.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 7
that providing group-level RPI yields incremental group performance improvement beyond that
attributable to group identity among groups of average task ability. Since average-ability
employees typically comprise the largest segment of the workforce in most organizations, our
results indicating that RPI and group identify jointly foster the highest performance are
particularly relevant.
Our results have implications for organizations, suggesting that firms can improve a
group’s task performance by fostering a sense of group identity, even if the task performed by the
individuals in the group are not interdependent. Our results also suggest that organizations can
employ group-level RPI to spur performance improvement among average groups. Further, there
is no significant performance benefit to providing top- and bottom-task ability groups with
group-level RPI, although it does not hinder their performance. Thus, organizations may wish to
configure their management information systems to provide group-level RPI, which should
improve the performance of individuals in groups of average task ability, even when they already
have a high sense of group identity.
This study contributes to the academic literature by introducing social comparison theory
as the basis for expecting group-level RPI to have positive performance effects on individuals
assigned to groups and asked to perform a repetitive, non-interdependent task. Although the
workplace is one of the most popular settings for organizational research, there is a paucity of
research which draws upon social comparison processes as means of explaining or improving
human behavior within organizational contexts (Greenberg, Ashton-James, & Ashkanasy, 2007).
Additionally, this study extends the group identity literature by empirically testing whether
group-level RPI has incremental performance effects above those fostered through group
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 8
identity. Finally, this study answers the call from Goodman and Haisley (2007) for organizational
research on social comparison processes using the group as the unit of analysis, to extend
research that has focused primarily on social comparison processes at the individual level.
The remainder of this paper is organized as follows: in the next section we discuss the
relevant literature and present the study’s hypotheses. In the method section, we outline the
experimental design, variables of interest, and the nature of participants in the experiment. In the
results section, we present the study’s findings. Finally, in the concluding section, we summarize
the study’s contributions, acknowledge limitations, and note future research opportunities.
Background and Hypotheses Development
In this study, we investigate the task performance effects of providing information to
individuals that discloses how the individual’s group performance ranks relative to other groups.
We also consider how RPI affects performance when provided along with an intervention
designed to heighten group identity. Specifically, we investigate the effects of group-level
relative performance information (present or absent) and group identity (high or low) on
performance in a repetitive, non-interdependent, effort- driven group production task. Psychology
literature suggests that providing information to individuals that permits them to make
comparisons should foster competition, resulting in increased effort and thus better task
performance (Festinger, 1954). However, it is unclear whether providing information that enables
an individual to make performance comparisons at the group level, rather than at the individual
level, will lead to improved task performance. Additionally, it is unclear whether providing
group-level RPI will yield task performance effects above and beyond those engendered by group
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 9
identity. In this section, the existing literature on group identity and relative performance
information are discussed, drawing upon theory to propose hypotheses.
Group Identity and Individuals’ Task Performance
Tajfel’s (1972) social identity theory provides a theoretical basis for explaining why an
individual’s task performance can be improved by placing an individual into a group and making
the individual’s membership in that group salient vis-à-vis when the individual performs a task
alone. Social identity theory (SIT) rests on individuals’ inclination to make comparisons
between groups to which an individual perceives she is a member (“in-groups”), and groups the
individual does not perceive herself to be a member (“out-groups”). SIT posits that individuals
engage in these comparisons to construct a sense of who they are, or social identity (Hogg,
2000). Consequently, the formation of an individual’s social identity occurs in the presence of
perceived group differences because the individual contrasts in-groups and out-groups to
construct their sense of self (Hogg, 2000). Affiliation with a professional sports team is a classic
example of an individual’s in-group/out-group mentality. For example, fans of the same sports
team constitute an individual’s in-group, while fans of other teams constitute an individual’s out-
group.
The individual’s sense of social identity affects how the individual interprets information
and how decisions are made (Lembke & Wilson, 1998). As an individual’s sense of group
identity, or perceived membership in a group, becomes more developed, the individual’s actions
shift to achieving a better group outcome as opposed to the individual’s outcome (Brewer, 1979).
As an individual shifts to a group focus, the individual’s perception that his or her actions will
determine the group’s success or failure becomes stronger (Wech, Mossholder, Steel, & Bennett,
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 10
1998). Thus, an individual who possess a more developed sense of group identity both cares
more about the success of the group and believes more strongly that the group’s success rests on
her actions, relative to an individual who possess a less-developed sense of group identity.
Individuals form group identity in a myriad of ways. The “minimal group paradigm” (Brown,
Tajfel, & Turner, 1980) demonstrates that individuals can form a relatively developed sense of
group identity if even minor similarities among group members are made salient, even among
individuals placed into ad hoc and randomly formed groups (Brewer, 1979; Hogg & Sunderland,
1991; Towry, 2003). A large body of empirical evidence supports the notion that group identity
can be engendered by the use of seemingly trivial criteria such as placing group members in close
geographic proximity to each other, assigning each group a unique color, and framing materials
such that an individual’s membership to a specific group is made explicit (Tajfel & Turner, 1979;
Akerlof & Kranton, 2000).
Research consistently finds a direct relationship between group identity and task
performance in effort-sensitive, simple tasks (James & Greenberg, 1989; Worchel et al., 1998;
Towry, 2003). Consistent with prior research, we expect that in our setting groups with a sense of
identity (“high- identity”) will exhibit better task performance improvement than groups without
a sense of identity (“low identity”). Our first hypothesis is essentially a replication of prior
studies, but is necessary to confirm before exploring whether RPI yields an incremental
improvement beyond that engendered by heightened group identity:
H1: Task performance improvement will be greater for high-identity groups than
for low- identity groups.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 11
Social Comparison Theory and Relative Performance Information
Relative performance information (RPI) communicates how a recipient’s performance
ranks in comparison to others. While existing literature examines the performance effects of
providing RPI that communicates how an individual’s performance ranks in comparison to other
individuals, this study examines the performance effects of providing individuals with RPI that
communicates how the performance of the individual’s group ranks relative to other groups.
Festinger’s (1954) social comparison theory (SCT) provides a theoretical basis for predicting that
providing individuals with group-level RPI should improve the task performance of individuals’
groups. SCT holds that individuals constantly make comparisons among each other to determine
how they, as an individual, are doing.5 These comparisons form the basis of an individual’s
evaluation of success or failure. In a group context, the provision of RPI at the group level should
similarly foster social comparisons, spurring members of each group to undertake actions aimed
at making their group appear better relative to other groups.
Social Comparison Theory: Studies Examining the Effects of RPI on Individuals
Although SCT has been the subject of extensive research in social psychology (Suls &
Wheeler, 2000 provide a comprehensive review), the existing literature largely focuses on
individuals’ affective reactions rather than on performance effects following social comparisons
(Goodman & Haisley, 2007). Aside from research conducted on the self-improvement effects in
coping and adjustment domains (Buunk & Ybema, 1997), relatively few studies have
5 While social comparison theory predicts that individuals primarily engage in upward comparisons, Wills (1981) finds empirical evidence suggesting that individuals engage in downward comparisons when encountering self- esteem threats the individual perceives cannot be overcome with action. Within the context of this study, participants can control their effort, directly influencing their performance. Thus, upward comparisons are expected to prevail within this study since individuals can take action to improve their performance.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 12
investigated the effects of social comparison on performance in individual achievement or
organizational contexts. One exception is Blanton, Bunk, Gibbons, and Kuyper (1999), who find
that making upward comparisons in an educational setting is strongly associated with superior
academic performance. In contrast, this study focuses on the performance outcomes of social
comparison in an organizational setting.
Within an organizational context, Murthy and Schafer (2011) investigate whether
providing RPI that communicates how an individual’s performance ranks in comparison to other
individuals is associated with performance improvement and how the framing of feedback in
conjunction with RPI affects performance. They find main effects for feedback framing and RPI,
although they did not find that providing RPI had incremental performance improvement effects
beyond positively framed feedback. When individuals compete against each other to win a
financial incentive, Hannan, Krishnan, and Newman (2008) find the motivational power of RPI
to be limited to contexts where relatively precise relative performance information is not
available. Thus, when relatively precise performance information is available to individuals who
are compensated in a tournament incentive scheme, their task performance becomes worse than
non-tournament incentive scheme participants (Hannan et al., 2008), unless bottom performers
are penalized (Newman & Tavkov, 2012).
Social Comparison Theory and RPI in a Group Context
Though SCT is generally used to explain the behavior of individuals, we suggest that it is
applicable to a group context when individuals are provided with information that enables them
to compare the performance of their group relative to other groups. Consequently, providing
individuals with group-level RPI should enable social comparison, fostering inter-group
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 13
competition, ultimately leading to improved group task performance.
Young et al. (1993) provide empirical evidence that providing group-level RPI to
individuals affects group performance. Using an interdependent task, Young et al. manipulate the
feedback given to groups such that each group was told they were just ahead, just behind, always
ahead, or always behind the other groups. Their results suggest that that groups provided with
feedback indicating they are just behind or just ahead perform the best.6 Further, although
providing group-level RPI to individuals performing a task in a group setting enables social
comparison, generalizing the findings of the existing literature is not appropriate because
providing RPI in a group setting contains dynamics not found when an individual performs a task
alone.
A major difference between providing RPI that communicates an individual’s
performance in relation to other individuals and RPI that communicates the performance of the
individual’s group relative to other groups is the identifiability of the individual. Identifiability
refers to the ability of an individual to be associated with his or her output. An individual
working alone is perfectly aware of how her effort affects performance, but this relationship is
less pronounced when an individual works in a group. When performance information is
provided only at the group level, an individual’s contribution to the group’s overall success or
failure is less salient. Additionally, when the individual’s specific contribution to the group’s 6 An interdependent task is one where “…the efforts of more than one person in creating the final product or outcome…” are required (p. 467, Young et al. 1993). Relatedly, this study extends the work of Young et al. (1993), who provided all treatment conditions (i.e., groups) with artificial relative performance information to investigate whether performance feedback affected productivity in a interdependent task. Whether providing group-level RPI to individuals performing a task within a group setting improved the performance of groups vis-à-vis groups that did not receive RPI was a question unanswered by Young et al. (1993). This study provides empirical evidence to answer that unanswered question, in addition to examining whether there are any incremental performance improvement effects associated with providing group-level RPI to individuals above those associated with group identity.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 14
effort is not known by other group members, shirking (aka social loafing, free riding) can result.
Individuals may choose to shirk because they do not care about the group, may feel their efforts
are less important to the group’s overall success or failure, and/or may perceive shirking to be
more difficult to detect in a group setting (Albanese & Van Fleet, 1985).
Drawing upon social comparison theory and prior research, we hypothesize that
providing group- level RPI to individuals performing an effort-sensitive, non-interdependent task
in a group setting will result in significantly better performance improvement when compared to
groups who do not receive RPI:
H2: Task performance improvement will be greater when group-level RPI is
provided than when group-level RPI is not provided.
RPI and Group Task Ability
Although providing individuals with group-level RPI is expected to significantly improve
a group’s task performance, it seems logical to expect that RPI’s performance to be affected by a
group’s ability to perform the task. Individuals in groups of high task ability are inherently well-
positioned to perform the task better vis-à-vis average and bottom performing groups. When
information enabling an individual to compare the performance of her group to others is not
available, an individual within a high task ability group has no basis for recognizing her group’s
inherent ability to perform the task well.
Applying social comparison theory, inter-group competition can be fostered by providing
individuals with group-level RPI that ranks the performance of each group in comparison with
the others. Since top task ability groups have the greatest task ability, we predict that providing
group-level RPI to top task ability groups will lead to a significant improvement in task
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 15
performance. However, considering that there may inherent production limits within the
underlying task (i.e., a ceiling effect), RPI’s performance effects among the highest ability groups
may be limited.
H2a: The performance improvement of high task ability groups will be greater when group-level RPI is provided than when group-level RPI is not provided.
As with groups of high task ability, providing groups of average task ability with group-
level RPI should foster inter-group competition. However, in contrast to the potential ceiling
effects faced by high task ability groups, those with average task ability have ample room for
improvement. Thus, we predict that providing group-level RPI to average task ability groups will
lead to significant improvements to task performance:
H2b: The performance improvement of average task ability groups will be greater
when group- level RPI is provided than when group-level RPI is not provided. Whether providing RPI to groups of low task ability will improve task performance is an
empirical question. The tenets of social comparison theory suggest that RPI should improve the
performance of bottom performers the most since bottom performers will have the greatest
number of upward comparisons. Further, psychology literature provides a theoretical basis for
suggesting bottom performers have the greatest motivation to take action to receive information
which improves their self- esteem (Aspinwall & Taylor, 1997). Consistent with these theoretical
propositions, Murthy (2011) finds that providing individuals with individual-level RPI most
improves the performance of bottom performing individuals. Extending the finding from Murthy
& Schafer (2011) to a group setting, groups of low task ability should be the most motivated to
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 16
improve their performance in response to RPI indicating that their group ranks low relative to
other groups.
Group dynamics make it unclear, however, whether these propositions can be generalized
to low task ability groups. A bottom performing individual has nobody but himself to blame for
poor performance in a setting where the individual competes directly against other individual.
Thus, in such a setting the individual must either take action to improve his performance or suffer
the self-esteem consequences of his performance. Within a group setting, however, the individual
is not solely responsible for the performance outcome. Information communicating that the
performance of an individual’s group is poor may be less threatening to the individual’s self-
esteem because he can disassociate himself from the group. In other words, an individual in a low
performing group can disassociate himself from the group, blaming others for the group’s poor
performance to avoid suffering a self-esteem threat.7 Further, an individual may take comfort in
knowing that while the performance of his group is poor, he is not alone in the proverbial
“performance cellar.” Thus, it is unclear whether providing group-level RPI to bottom task
ability groups will improve or hinder their task performance, leading to the following non-
directional hypothesis:
H2c: The performance improvement of bottom task ability groups will be
affected when group-level RPI is provided than when group-level RPI is not provided.
7Mussweiler, Gabriel, and Bodenhausen (2000) find disassociation to be an esteem-preservation strategy that can be employed when an individual can mentally remove himself from membership in a specific group. Disassociation nullifies the negative feedback provided by a social comparison because the individual no longer considers it relevant to his identity.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 17
Group Identity and RPI: Incremental Task Performance Effects
“Ceiling effects” are an inherent aspect of effort-sensitive tasks. Ceiling effects dictate
the production boundaries of an effort-sensitive task such that an individual’s physical limitations
form the task’s production limits, setting production boundaries that an individual cannot
transcend. In the context of this study, group identity is predicted to improve an individual’s task
performance. Relative to settings where individuals perform a task within low-identity groups,
individuals who perform a task within high-identity groups are found to “care” more for their
group, empirically demonstrated through a greater focus on group rather than individual
outcomes (Wit & Wilke, 1992; De Cremer & van Vugt, 1998), enhanced task coordination
strategies (Brewer, 1979), greater effort levels (Towry, 2003), and more effective communication
strategies (King, 2002). Thus, it seems logical to predict that providing RPI to high-identity
groups should improve task performance because enabling social comparison should foster
competition between groups, spurring individuals within groups to perform better. However, as
the task employed in this study is effort-sensitive, it is unclear whether providing group-level RPI
to high-identity groups will significantly improve task performance because such groups may
already be motivated to improve performance from a social identity perspective. Thus, while
providing high-identity groups with group-level RPI should theoretically improve task
performance, it is unclear whether sufficient performance slack exists, given that the social
identity effect is also operating. Applying social comparison theory, we predict that providing
group-level RPI to high-identity groups should improve task performance improvement because
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 18
the social comparisons fostered by RPI should be additionally meaningful to individuals within
high-identity groups, spurring those individuals to the limits of their individual task ability:
H3: Providing group-level RPI to high-identity groups will yield incremental
performance effects beyond those attributable to high-identity alone. Recognizing the potential ceiling effects inherent in this study’s task, we predict that
finding significant incremental task performance effects of providing group-level RPI to high-
identity groups may hinge upon on a group’s ability to perform the task. Since high-identity
groups with high task ability have the greatest inherent advantage in performing the task, an
argument predicting that these groups should benefit most from receiving group-level RPI can be
made. However, since the performance of high ability groups is already elevated relative to
average and bottom ability groups, it is worth considering the fact that the task’s ceiling effects
may inhibit incremental task performance effects, because high-identity, high ability groups are
more likely to already be performing near the edge of the task’s ceiling even before receiving
group-level RPI. From a pragmatic perspective, high-identity groups of average and low task
ability have greater room for task improvement than groups of high ability because their
performance is less likely to be near the task’s production ceiling. Accordingly, we suggest that
significant positive incremental task performance effects associated with providing group-level
RPI to high-identity groups are likely to be found among average and low task ability groups,
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 19
although the application of social comparison theory suggests positive incremental task effects
should be found among even groups of high task ability:
H4a: Providing group-level RPI to high-identity groups of high task ability will
yield incremental performance improvement beyond those attributable to high-identity alone.
H4b: Providing group-level RPI to high-identity groups of average task ability will
yield incremental performance improvement beyond those attributable to high-identity alone.
H4c: Providing group-level RPI to high-identity groups of low task ability will yield
incremental performance improvement beyond those attributable to high-identity alone.
Method
Participants and Research Design
Participants were 267 undergraduate students enrolled in introductory business courses
from a large university in the Southeastern United States, who were randomly assigned to three
person groups, constituting a total of 89 group observations.8 The task involved multiple sessions
of decoding numbers to letters, with experimental manipulations taking place after task training
was complete. Each participant was awarded a small amount of course credit for performing the
experimental task. In compliance with Institutional Review Board (IRB) requirements, all
participants were verbally informed that full credit would be given to anyone who completed the
experiment, regardless of task performance.
8 Because RPI was provided to individuals at the group level, individuals’ performance responses are attributable to the group-level information. Thus, consistent with the call for organizational social comparison research from Goodman and Haisley (2007), the group serves as both the focus point for the individual and as the unit of analysis in this study
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 20
The experiment uses a 2 x 2 factorial design, crossing group identity (high or low) and
relative performance information (present or absent). Figure 1 illustrates the experimental
design. The experiment consisted of several letter decoding sessions in the same computer
laboratory, with each experimental session representing one of the four treatment conditions.
Each experimental session lasted for approximately 45 minutes, and participants completed the
entire experiment in one sitting using a web-based application specifically designed for the
experiment.
<<INSERT FIGURE 1 HERE>>
Task
The experiment task is adopted from Chow (1983) and involves six, four minute periods
which individuals in three-person groups decoded numbers to letters. Figure 2 shows a
screenshot of the main window of the web-based application; the task is similar to that used in
several prior studies (Murthy & Schafer 2011; Fisher, Frederickson, & Peffer 2000, 2002; Fisher,
Maines, Peffer, & Sprinkle, 2002; Chow 1983). Participants used two visual displays to perform
the task. The top display showed the decoding key, matching numbers 1 through 42 with random
alphabet characters. The top display also displayed the time remaining in decoding session and
gave participants a number to decode.
Participants used the bottom display to input (i.e., enter) the letter corresponding to the
number they were asked to decode. To control for experience effects, the decoding key was
randomly generated by the web-based application at the start of each decoding session so that no
two decoding sessions contained the exact same decoding key. Participants received immediate
feedback after entering each letter which indicated whether the entry was correct or incorrect.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 21
Additionally, the bottom screen showed the participant’s running total of letters correctly and
incorrectly decoded. The researcher orally instructed all participants to decode as many letters as
possible before each decoding session started.
<<INSERT FIGURE 2 HERE>>
Experiment Procedure and Manipulations
As participants arrived at the computer laboratory, they were randomly assigned to a
closed-lid laptop and individually told that all forms of communication were prohibited.
Participants were also told not to open the top of their laptop until instructed. At the appointed
start time, the laboratory doors were closed and no additional participants were admitted. The
researcher then directed participants to lift the lid of their laptop and asked to read the experiment
instructions. After reading the instructions, participants decided whether they wished to
participant in the experiment, affirming their consent by signing an IRB consent form.
After reading and signing the informed consent form, participants accessed the web-based
application, which explained the experimental task. Next, participants completed a two-minute
training session to become familiar with the task. At the conclusion of the training session,
participants were given the opportunity to ask questions. After any questions were answered,
participants next completed three four-minute practice decoding sessions. The purpose of these
practice sessions was to familiarize participants with the task and to provide a basis for assessing
their individual task performance ability. At the end of each practice session, the participant’s
laptop provided feedback regarding how many letters the participant correctly (and incorrectly)
decoded.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 22
After participants completed three, four-minute practice sessions, the researcher
introduced the experimental manipulations. Participants then performed three main decoding
sessions, each lasting four minutes. To discourage participant end-game behavior, participants
were not told how many main decoding sessions they would experience. Consistent with the
practice decoding sessions, a running total of the number of letters the individual correctly
decoded during each main session was displayed. After all the main sessions were over, each
participant completed an electronic post-experimental questionnaire (PEQ). To encourage
participants to answer the PEQ conscientiously, participants were orally informed that they
would not be dismissed from the lab until all individuals had completed the PEQ.
Group Identity Manipulation
Group identity was manipulated at two levels (low-identity and high-identity) after the
last practice decoding session ended and before the first main decoding session began.
Participants in the low-identity condition were randomly assigned to numbered groups (e.g.,
“Group 3”) via a display on the individual’s laptop screen. Although participants in the low-
identity condition were assigned to a group, they did not know which individuals belonged to
their group. In the high-identity condition, the researcher randomly assigned participants colored
groups (e.g., “Group Blue”). As the researcher assigned individuals to a group, he asked each
group member to come forward, asking each member to introduce him/herself to the other group
members, and required all group members to wear a colored placard which corresponded to the
color of the group to which each individual belonged.9
9 While it may seem unlikely that such ad hoc manipulations can facilitate group identity, the empirical evidence supporting this “minimal group paradigm” is robust. Turner (1987) provides a comprehensive review of the research that supports use of the minimal group paradigm in fostering group identity
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 23
RPI Manipulation
RPI was manipulated at two levels: present or absent. In experimental sessions where
RPI was present, the participants within each of the experimental session’s groups were provided
with a table showing the rank-ordered performance of all groups and the total number of letters
decoded by each group at the end of each decoding session. The table also presented each
participant with the number of letters they personally decoded, although participants did not
know how many letters the other members in their group decoded. The rank-ordered information
presented to participants was determined by summing the total number of letters that each group
member decoded during the main session. The RPI communicated each group’s ordinal rank, the
total number of letters correctly decoded by their group, and the total number of letters correctly
decoded by the other competing groups. Figure 3 provides an example of the screen presented to
participants who received RPI at the end of each main decoding session.10 In experimental
sessions where RPI was absent, participants only received information communicating the total
number of letters decoded by their group. That is, a participant who did not receive group-level
RPI received only a table presenting the number of letters decoded by the participant at the end
of each decoding session, thus receiving no information communicating the ordinal rank of their
group’s performance vis-à-vis the other groups.
<<INSERT FIGURE 3 HERE>>
10 Although all participants within this study received feedback that communicated how many letters the participant correctly decoded at the end of each decoding session, social comparisons exist only at the group-level and only within experimental sessions where RPI was provided because the feedback provided in RPI conditions did not permit participants to compare their task performance against that of other participants, it only permitted participants to compare the task performance of their group vis-à-vis other groups.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 24
Measures
To test this study’s hypotheses, task performance is measured in terms of the number of
additional letters decoded by each group after treatment manipulations occur.11 A change score is
used as this study’s dependent measure because any task performance effect should manifest
itself as an incremental change of performance from the session where experimental
manipulation was not present compared to the session where experimental manipulation was
present. After obtaining the appropriate change score, statistical tests utilizing between-subjects
analysis were performed. Additionally, to test H3, predicting that task performance effects will
vary based upon group task ability, groups are categorized as high (top third), average (middle
third), and low (bottom third) performers based on summing the total number of letters decoded
by each member of a group during the last trial decoding session.
The purpose of H1 is to examine whether group identity has performance effects beyond
those attributable to merely assigning individuals to work in groups. H1 is tested by comparing
the change in the average number of letters decoded by groups between the last training session
and the first main decoding session. Conducting statistical analysis in this manner examines if
manipulating high-identity has performance effects beyond the mere assignment of individuals to
a group (low-identity).
The purpose of H2 is to examine whether providing group-level RPI has task
performance effects, and if so, whether performance effects varied based upon a group’s task 11 The focus of this study is on the increase in a group’s task performance. Consequently, we use the change in the total number of letters decoded as the dependent measure. Although using a percentage change measure presents an alternative measure to test this study’s hypotheses, our approach is more rigorous because it biases against finding results. For example, a six-letter increase in decoding performance is larger for a group decoding 60 letters (10 percent change) relative to a group who decodes 100 letters (6 percent change). Thus, using the raw change in the number of letters correctly decoded is a more conservative measure. Further, our results were robust to a sensitivity analysis that employed a percentage change measure as the dependent variable.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 25
ability. The first time group’s received RPI was after main decoding session one. Thus, any
performance effect engendered by providing group-level RPI would manifest itself in main
decoding session two, as participants made effort adjustments based upon the feedback
provided to them at the end of main decoding session one.12
Consequently, H2 is tested by comparing the change in the average number of letters
decoded by groups between the second main decoding session and the first main decoding
session. The purpose of H3 is to examine whether providing RPI to high-identity groups has any
incremental performance effects beyond those attributable to group identity, and if so, whether
performance effects varied based upon group task ability. Thus, only high-identity groups were
included in the statistical analysis used to test H3. Specifically, H3 is tested by comparing the
change in the average number of letters decoded by high-identity groups between the second
main decoding session and the first main decoding session, since any performance effect of RPI
should manifest itself during main decoding session two.
Results
Group Identity Manipulation Check and Group Task Ability Categorization
A reliability analysis of the ten-item group-identity scale indicates good internal
consistency (Cronbach’s alpha = 0.89). Participants within high-identity sessions reported
stronger feelings of group identification than participants within low identity sessions (Q1,
12 By designing the experiment such that any performance effects of group identity occur in main decoding session one, while performance effects of group-level relative performance information occur in main session two, empirical evidence is obtained which speaks to whether providing group-level RPI has incremental performance effects beyond those attributable to group identity. As a result, this paper provides empirical evidence on whether group identity has performance effects beyond those attributable to merely assigning individuals to groups, and on whether providing group-level RPI has performance effects beyond those attributable to both the mere assignment of individuals to group and group identity
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 26
μHigh-Identity=3.51, μLow-Identity=2.92, p=0.004). Thus, participants within high-identity
treatments more strongly identified with their groups than participants within low-identity
treatments, indicating that the manipulation of group identity was successful. Table 1 presents
the ten-item group identity scale with participant responses by treatment.
Group task ability was calculated by summing the total number of letters decoded by each
group member during the last trial decoding session, then classifying each group into one of three
categories: high (top third), average (middle third), and low (bottom third). Since group task
ability is a measured variable, the number of groups in each cell could not be fully controlled.
Consequently, the number of observations available for statistical analysis regarding group task
ability is not equal.
<<INSERT TABLE 1 HERE>>
Hypotheses Testing
Hypothesis 1 (H1) predicts that a group's task performance improvement will be
significantly greater among high-identity groups as compared to low-identity groups. Panel A of
Table 2 presents descriptive statistics regarding the change in the number of letters decoded
between the last training session and the first main decoding session (µHigh-Identity=15.00,
µLow-Identity=5.04). To test H1, an ANCOVA was created with group identity as the factor and
change in the average number of letters decoded by groups between the last training session and
the first main decoding session as the dependent variable. The ANCOVA includes group task
ability and RPI as covariates.13 As shown in Panel B of Table 2, results indicate that group
13 Group task ability is included as a covariate to rule out the explanation that groups of higher task ability fell disproportionately into high-identity conditions. RPI is included as a covariate to rule out it's role as a potential explanation for task improvement effects
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 27
identity has a significant effect on a group's task performance improvement (F=20.943, p=
0.000). Thus, H1 is supported.
<<INSERT TABLE 2 HERE>>
Hypothesis 2 (H2) predicts that groups receiving RPI will experience more positive task
performance improvement relative to groups who do not receive RPI. Panel A of Table 3
presents descriptive statistics regarding the change in the number of letters decoded between the
first main decoding session and the second main decoding session (µRPI=12.05, µNo-RPI=7.22).
To test H2, an ANCOVA was created with RPI and group identity as the factors and the change
in the average number of letters decoded by groups between the first main decoding session and
the second main decoding session as the dependent variable, with group task ability included as a
covariate. As shown in Panel B of Table 3, results indicate that providing RPI has a significant
effect on a group's task performance improvement (F=5.588, p= 0.010, one-tailed). Thus, H2 is
supported.
<<INSERT TABLE 3 HERE>>
Hypotheses 2a, 2b, and 2c explore whether the performance effects of RPI vary based
upon a group's ability level. Specifically, H2a predicts that providing RPI to groups of top task
ability (i.e., those groups whose task ability ranks within the top one-third of the groups within
the experimental session) will result in greater improvements in task performance than when RPI
is not provided to groups of top task ability. Stratifying this study's group observations into an
examination of only groups of top task ability, descriptive statistics indicate that the average
change in the number of letters decoded between the first main decoding session and the second
main decoding session for top task ability groups was different (µRPI=15.69, µNo-RPI=10.60).
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 28
To test H2a, an ANCOVA was created with RPI and group identity as the factors and the change
in the average number of letters decoded by top task ability groups between the first main
decoding session and the second main decoding session as the dependent variable. Although the
treatment means were in the predicted direction as discussed above, no significant difference in
task performance improvement was found when RPI was provided to top task ability groups vis-
a-vis when RPI was not provided to top task ability groups (F=0.489, p= 0.246, one-tailed).
Thus, H2a is not supported.
H2b predicts that providing RPI to groups of average task ability (i.e., those groups
whose task ability ranks below the top third and above the bottom third of groups within the
experimental session) will result in greater improvements in task performance than when RPI is
not provided to groups of average task ability. Stratifying this study's group observations into an
examination of only groups of average task ability, Panel A of Table 4 presents descriptive
statistics regarding the change in the number of letters decoded between the first main decoding
session and the second main decoding session for average task ability groups (µRPI=11.31, µNo-
RPI=2.00). To test H2b, an ANCOVA was created with RPI and group identity as the factors
and the change in the average number of letters decoded by average task ability groups between
the first main decoding session and the second main decoding session as the dependent variable.
As shown in Panel B of Table 4, providing RPI to groups of average task ability was found to
improve task performance improvement vis-a-vis when RPI was not provided to groups of
average task ability. (F=9.933, p= 0.002, one-tailed). Thus, H2b is supported.
<<INSERT TABLE 4 HERE>>
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 29
H2c predicts that providing RPI to groups of bottom task ability (i.e., those groups whose
task ability ranks in the bottom one-third of the groups within the experimental session) will
affect task performance. Stratifying this study's group observations into an examination of only
groups of bottom task ability, descriptive statistics indicate that the average change in the number
of letters decoded between the first main decoding session and the second main decoding session
for bottom task ability groups was different (µRPI=8.80, µNo-RPI=10.15). To test H2c, an
ANCOVA was created with RPI and group identity as the factors and the change in the average
number of letters decoded by bottom task ability groups between the first main decoding session
and the second main decoding session as the dependent variable. Test results indicate there was
no significant difference in task performance improvement between groups of bottom task ability
provided with RPI and those not provided with RPI (F=0.062, p= 0.806, two-tailed). Thus, H2c
is not supported.
Hypothesis 3 (H3) explores whether providing RPI to high-identity groups has any
incremental performance effects beyond those fostered through high-group identity. H3 predicts
that providing RPI can provide incremental task performance value above the task performance
improvement fostered by high group identity. Descriptive statistics indicate that the average
change in the number of letters decoded between the first main decoding session and the second
main decoding session for high-identity groups was different based upon whether RPI was
provided (µHigh-identity with RPI=8.00, µHigh-identity, without RPI=4.40). To test H3, we
stratify the data into high-identity group observations only. Next, an independent samples t-test
was performed with the RPI's presence (absence) as the independent variable and the change in
the average number of letters decoded between main session one and main session two serving as
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 30
the dependent variable. Test results indicate there a significant difference in task performance
improvement between high-identity groups who received RPI and high-identity groups that did
not receive RPI (t=1.357, p=0.092, one-tailed). Thus, there is evidence that providing RPI has
incremental value above and beyond that provided by group identity. Consequently, H3 is
supported.
Hypotheses 3a, 3b, and 3c examine whether RPI's incremental performance effects vary
based upon group task ability. This examination is important because even though H3 predicts
positive task performance effects among high-identity groups that receive RPI, it is important to
examine whether incremental performance effects are uniform, especially in light of the potential
ceiling effects that may be inherent in this study's task.
H3a predicts that providing group-level RPI to high-identity groups of top task ability
(i.e., those groups whose task ability ranks on the top third of their experimental session) will
yield incremental task performance improvement beyond that attributable to high-identity
alone. Descriptive statistics suggest that providing RPI to high-identity groups of top task
ability does not yield incremental performance effects beyond those attributable to high-
identity, so no further hypothesis testing procedures are performed (µHigh-identity top-task
ability groups receiving RPI=7.20, µHigh-identity top-task ability groups not receiving
RPI=8.00). Thus, H3a is not supported.
H3b predicts that providing group-level RPI to high-identity groups of average task
ability (i.e., those groups whose task ability ranks below the top third and above the bottom third
of groups in their experimental session) will yield incremental task performance improvement
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 31
beyond that attributable to high-identity. Descriptive statistics suggest that providing RPI to high-
identity groups of average task ability may yield incremental performance effects beyond those
attributable to high-identity (µHigh-identity average task ability groups receiving RPI=10.59,
µHigh-identity average task ability groups not receiving RPI=0.45). To test H3b, we stratify the
data into high-identity, average task ability group observations alone. Next, an independent
samples t-test was performed with the RPI's presence (absence) as the independent variable and
the change in the average number of letters decoded between main session one and main session
two serving as the dependent variable. Test results indicate there a significant difference in task
performance improvement between high-identity groups of average task ability who received RPI
and high-identity groups of average task ability that did not receive RPI (t=3.104, p=0.004, one-
tailed). Thus, H3b is supported.
H3c predicts that providing group-level RPI to high-identity groups of bottom task ability
(i.e., those groups whose task ability ranks in the bottom third of their experimental session) will
yield incremental task performance improvement beyond that attributable to high-identity alone.
Descriptive statistics suggest that providing RPI to high-identity groups of bottom task ability
does not yield incremental performance effects beyond those attributable to high-identity, so no
further hypothesis testing procedures are performed (µHigh-identity bottom-task ability groups
receiving RPI=8.80, µHigh-identity bottom-task ability groups not receiving RPI=10.15). Thus,
H3c is not supported.
In summary, this study's results indicate that that group-identity and group-level RPI are
two factors that positively affect groups' task performance improvement. However, a deeper
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 32
investigation reveals that providing group-level RPI only significantly improves the task
performance of average task ability groups. The results also indicate that RPI provides
incremental task performance improvement above that fostered by group identity. A closer
investigation of the task performance effects of providing group-level RPI to high-identity groups
finds incremental task performance effects to be significant among groups of average task ability.
Table 5 presents this study's hypotheses and identifies whether each was supported.
<<INSERT TABLE 5 HERE>>
Summary and Conclusion
This study examines whether providing individuals with group-level RPI improves
groups’ task performance improvement, and whether the performance effects associated with
providing group-level RPI are incremental to those attributable to group identity. Using a non-
interdependent group task where individuals decoded numbers to letters, we measure the change
in group performance over multiple decoding sessions. Group identity (high or low) and the
provision of relative performance information (present or absent) were varied between subjects.
Groups’ task ability level served as a measured independent variable, with groups classified as
high, average, or bottom based on the task performance of the individuals who comprised the
group during the last of three trial decoding sessions. Our results reveal that providing group-
level RPI yields positive group task performance improvement effects.14 An examination of
whether RPI’s performance effects vary by a group’s ability to perform the task reveals that
RPI’s motivational effects lie predominantly within groups of average task ability. Further, our
14 Consistent with prior literature, our results also indicate that group identity improves groups’ task performance improvement
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 33
findings reveal that providing group-level RPI can foster incremental performance improvement
effects beyond those attributable to group identity, particularly among groups of average task
ability.
Our results have a number of implications for the academic body of knowledge on RPI
and SCT and for organizations. While the performance effects of providing RPI to individuals
has been examined by the literature, this study is one of the few using the group as the unit of
analysis to examine the performance effects of providing RPI to groups, a setting with different
dynamics than a setting where an individual performs a task alone. Differences in settings may
explain why, in contrast to Murthy (2011) who finds the most pronounced task performance
improvement effects to be among bottom performing individuals that receive RPI, we find the
most pronounced task performance improvement to be among groups of average task ability and
no performance effect among bottom task ability groups. Thus, our results highlight the care that
needs to be taken in generalizing the findings of the existing RPI literature to a group setting. Our
study also contributes to the academic literature by empirically testing whether group-level RPI
has incremental performance effects above those fostered through group identity, extending both
the RPI and group identity literature.
From an organizational perspective, our results suggest that firms can improve group task
performance by encouraging groups to form a sense of group identity even in a non-
interdependent task. Organizations may also find it beneficial to configure their management
information systems to provide group-level RPI, particularly to groups of average task ability
because doing so was found to yield incremental performance improvement beyond that
attributable to group identity. Since our findings did not indicate RPI provided an incremental
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 34
task performance effect of beyond group-identity for high- and bottom- task ability groups, our
results suggest organizations may experience marginal returns to providing them with both
group-level RPI and fostering a sense of group identity.
Although one of the strengths of this study is that we measure groups’ actual task
performance improvement in response to treatment manipulations, we acknowledge that group
task ability is a measured, rather than a manipulated, variable. Since categorizing groups into
high, average, and bottom task ability occurred after individuals had been assigned to groups,
uneven cell sizes resulted. Thus, the analyses regarding whether group-level RPI has differential
motivational effects based upon a group’s task ability is based on relatively uneven cell sizes,
constituting a limitation of this study.
Future research could build upon this study by assigning participants to groups based on
the participant’s task ability ex ante. That is, participants could be required to remotely complete
a task “qualifier” before the researcher scheduled any experimental sessions. Once the researcher
had empirical data regarding participants’ task ability, she could explicitly assign participants to
groups of evenly- matched participant task ability.
Because treatment means directionally suggested that providing group-level RPI to
groups of high task ability may improve task performance improvement, future research could
continue to examine this issue. Considering that the task used in this study contained an inherent
production limit, (i.e., there is a limit to the number of letters an individual can physically decode
in a limited time period) we suggest that ceiling effects inherent in the decoding task used in this
experiment explain why providing group- level RPI to high task ability groups did not have a
significant effect on task improvement. That is, the high-identity, high ability groups within our
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 35
setting may have been operating near the task’s production ceiling before receiving our
experimental manipulations. Future research could examine this issue using a task with less
inherent ceiling effects, such as the creativity-based task used by Kachelmeier, Reichert, and
Williamson (2008). To conclude, the main findings of this study are that providing group-level
RPI yields positive group task performance improvement effects, particularly within groups of
average task ability, where RPI provides performance improvements beyond those attributable to
group identity.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 36
References
Akerlof, G. A. & Kranton, R. E. (2000). Economics and identity. The Quarterly Journal of
Economics, 115(3): 715-753.
Albanese, R. & VanFleet, D. (1985). Rational behavior in groups: The free-riding tendency. The
Academy of Management Review, 10(2): 244-255.
Alchian, A. & Demsetz, H. (1972) Production, information costs and economic organization. The
American Economic Review, 62(5): 777-795.
Aspinwall, L. G. & Taylor, S. E. (1997). A stitch in time: Self-regulation and proactive coping.
Psychological Bulletin, 121(3): 417.
Blanton, H., Buunk, B. P., Gibbons, F. X., & Kuyper, H. (1999). When better-than-others
compare upward: Choice of comparison and comparative evaluation as independent
predictors of academic performance. Journal of Personality and Social Psychology,
76(3): 420-430.
Brewer, M. (1979). In-group bias in the minimal intergroup situation: A cognitive-motivational
analysis. Psychological Bulletin, 86(2): 307-324.
Brown, R., Tajfel, H. , & Turner, J. (1980). Minimal group situations and intergroup
discrimination: Comments on the paper by Aschenbrenner and Schaefer. European
Journal of Social Psychology, (10): 399-414.
Buunk, B. P. & Ybema, J. F. (1997). Social comparisons and occupational stress: The
identification contrast model. In B. P. Buunk & F. X. Gibbons (Eds.), Health, coping and
well being: Perspectives from social comparison theory (pp. 359–388). Hillsdale, NJ:
Erlbaum.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 37
Business magazine (2006). Smells like team spirit, (4). Available at
http://www.busmanagement.com/article/Smells-like-team-spirit/. Accessed: February 13,
2013.
Chow, C. W. (1983). The effects of job standard tightness and compensation scheme on
performance: An exploration of linkages. The Accounting Review, (58): 667-685.
De Cremer, D. & van Vugt, M. (1998). Collective identity and cooperation in a public goods
dilemma: A matter of trust or self-efficacy? Current Research in Social Psychology 3(1).
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7: 117-140.
Fisher, J.G., Frederickson, J. R., & Peffer, S. A. (2002). The effect of information asymmetry on
negotiated budgets: An empirical investigation. Accounting, Organizations and Society,
27(1): 27-43.
Fisher, J.G., Maines, L. A., Peffer, S. A., & Sprinkle, G. B. (2002). Using budgets for
performance evaluation: Effects of resource allocation and horizontal information
asymmetry on budget proposals, budget slack, and performance. The Accounting Review,
77(4): 847-865.
Goodman, P. S. & Haisley, E. (2007). Social comparison processes in an organizational context:
New directions. Organizational Behavior and Human Decision Processes, (102): 109-
125.
Greenberg, J., Ashton-James, C. E., & Ashkanasy, N. M. (2007). Social comparison processes in
organizations. Organizational Behavior and Human Decision Processes, (102): 22-41.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 38
Hannan, R. L., McPhee, G. P., Newman, A. H., & Tafkov, I. D. (2012). The effect of relative
performance information on effort allocation and performance in a multi-task
environment. The Accounting Review, forthcoming.
Hogg, M. A. (2000). Social identity and social comparison. In J. Suls, J. & L. Wheeler (Eds.),
Handbook of social comparison: Theory and research (pp. 401-421). New York: Kluwer
Academic/Plenum Publishers.
Hogg, M. A. & Sunderland, J. (1991). Self-esteem and intergroup discrimination in the minimal
group paradigm. British Journal of Social Psychology, (30):51-62.
James, K. & Greenberg, J. (1989). In-group salience, intergroup comparison, and individual
performance and self-esteem. Personality and Social Psychology Bulletin, 15(4): 606-
616.
Jouini, O., Dallery, Y. D., & Nait-Abdallah, R. (2008). Analysis on the impact of team-based
organizations in call center management. Management Science, 54(2): 400-414.
Kachelmeier, S. J., Reichert, B. E., & Williamson, M. G. (2008). Measuring and motivating
quantity, creativity, or both. Journal of Accounting Research, 46(2): 341-373.
Karau, S. & Williams, K. (1993). Social loafing: A meta-analytical review and theoretical
integration. Journal of Personality and Social Psychology, 65(4): 681-706.
King, R. (2002). An experimental investigation of self-serving biases in an auditing trust game:
The effect of group affiliation. The Accounting Review, 77(2): 265-284.
Latane, B., Williams, K., & Harkins, S. (1979). Many hands make light the work: The causes
and consequences of social loafing. Journal of Personality and Social Psychology, 37(6):
822-832.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 39
Lembke, S. & Wilson, M. (1998). Putting the “team” into teamwork: Alternative theoretical
contributions for contemporary management practice. Human Relations, 51(7): 927-944.
Martin, R. (2000). Can I do X? Using the proxy comparison model to predict performance. In
Suls, J. & Wheeler, L. (Eds.), Handbook of social comparison: Theory and research (pp.
67-80). New York: Kiuwer Academic/Plenum Publishers.
Murthy, U. S. & Schafer, B. A. (2011). The effects of relative performance information and
framed information systems feedback on performance in a production task. Journal of
Information Systems, 25(1): 159-184.
Mussweiler, T., Gabriel, S., & Bodenhausen, G. V. (2000). Shifting social identities as a strategy
for deflecting threatening social comparisons. Journal of Personality and Social
Psychology, 79(3): 398.
Shaw, M. E. (1976). In Group dynamics: The Psychology of Small Group Behavior. McGraw-
Hill series in psychology. New York: McGraw-Hill.
Suls, J. M. & Wheeler, T. A. (2000). Handbook of social comparison: Theory and research.
New York: Kluwer Academic/Plenum Publishers.
Tajfel, H. (1972). Social categorization. In Suls, J. M. & Wheeler, T. A. Handbook of Social
Comparison: Theory and Research (pp. Please insert page numbers here). New York:
Kluwer Academic/Plenum Publishers.
Tajfel, H. & Turner, J. C. (1979). An integrative theory of intergroup conflict. The Social
Psychology of Intergroup Relations, 33(1): 47.
Tavkov, I. D. (2009). Private and public relative performance information under different
incentive systems. Working paper.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 40
Towry, K. (2003). Control in a teamwork environment—The impact of social ties on the
effectiveness of mutual monitoring contracts. The Accounting Review, 78(4): 1069-1095.
Turner, J. C. (1987). A self-categorization theory. In Turnber, J. C. (Ed.)’s Rediscovering the
social group: A self-categorization theory (pp. Please insert page numbers here). Oxford,
U.K.: Basil Blackwell Ltd.
Wech, B. A., Mossholder, K. W., Steel, R. P., & Bennett, N. (1998). Does work group
cohesiveness affect individuals’ performance and organizational commitment? A cross-
level examination. Small Group Research, 29(4): 472-494.
Wellins, R. S., Byham, W. C., & Dixon, G. R. (1994). Inside teams: How 20 world-class
organizations are winning through teamwork. San Francisco, CA: Jossey-Bass.
Williams, K., S. Nida, L. Baca, B. Latane (1989). Social loafing and swimming: effects of
identifiability on individual and relay performance of intercollegiate swimmers. Basic
and Applied Social Psychology 10(1): 73-81.
Williams, K., Nida, S., Baca, L. , & Latane, B. (1989). Social loafing and swimming: Effects of
identifiability on individual and relay performance of intercollegiate swimmers. Basic
and Applied Social Psychology, 10(1): 73-81.
Williams, K., Harkins, S., & Latane, S. B. (1981). Identifiability as a deterrent to social loafing:
Two cheering experiments. Journal of Personality and Social Psychology, 40(2): 303-
311.
Wills, T. A. (1981). Downward comparison principles in social psychology. Psychological
Bulletin (90): 245-271.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 41
Wit, A. P. & Wilke, H. A. M. (1992). The effect of social categorization on cooperation in three
types of social dilemmas. Journal of Economic Psychology, 13(1): 135-151.
Worchel, S., Rothgerber, H., Day, E., Hart, D., & Butemeyer, J. (1998). Social identity and
individual productivity within groups. British Journal of Social Psychology, 37(1): 389-
413.
Young, M., Fisher, J., & Lindquist, T. (1993). The effects of intergroup competition and
intragroup cooperation on slack and output in a manufacturing setting. The Accounting
Review, 68(3): 466-481.
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 42
Figure Captions Figure 1. Research design Figure 2. Decoding task screen viewed by all participants during the main decoding sessions Figure 3. Group-level relative performance information provided to each participant after each main decoding session was complete
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 43
Figure 1. Research design
Group Identity
Feedback Condition
High-Identity Low-Identity
Group-Level Relative Performance
Information Provided Condition 1 Condition 2
No Group-Level Relative Performance Information Provided
Condition 3 Condition 4
GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION 44
Figure 2. Decoding task screen viewed by all participants during the main decoding sessions
Top Display
Bottom Display Figure 3. Group-level relative performance information provided to each participant after each main decoding session was complete
Running head: GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION
Running head: GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION
Table 1. Individuals perceptions of group identity on a seven-point Likert scale
Question (1 = Strongly disagree, 4 = neither agree nor disagree, 7 = strongly agree)
High Identity (n=120)
Low Identity (n=147)
Overall (n=267)
I identify with other members of the team
μ=3.53 (1.567)
μ=2.92 (1.718)
μ=3.19 (1.677)
I am like other members of my team
3.28 (1.42)
3.08 (1.542)
3.17 (1.489)
My team is an important reflection of who I am
2.82 (1.676)
2.73 (1.819)
2.77 (1.754)
I think my team has much to be proud of
4.08 (1.754)
3.72 (2.010)
3.88 (1.905)
I feel good about my team 4.36 (1.555)
4.01 (1.876)
4.17 (1.745)
I have little respect for my team 2.07 (1.561)
2.65 (1.770)
2.39 (1.702)
I would rather not tell that I belong to this team
2.07 (1.419)
2.41 (1.583)
2.26 (1.519)
I would like to consider working with my team
4.22 (1.397)
4.13 (1.745)
4.17 (1.596)
I dislike being a member of my team
2.19 (1.469)
2.53 (1.689)
2.38 (1.600)
I would rather belong to another team
2.42 (1.623)
2.85 (1.910)
2.66 (1.797)
Overall Average 3.51 (1.767)
2.92 (1.86)
3.10 (1.834)
Running head: GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION
Table 2. H1: the effect of group identity on groups’ task performance improvement Panel A: Change in the Average Number of Letters Decoded by Group Group Identity Treatment
N Change in Number of Letters Decoded
Standard Deviation
High-Identity 40 + 15.00 10.460 Low-Identity 49 + 5.04 9.584
Dependent Variable: Average number of letters decoded by group during main decoding session one minus the average number of letters decoded by group during the last training decoding session. Panel B: ANCOVA Results Sum of
Squares df Mean
Square F Significance
Model
2564.901 4 641.225 6.372 0.000
Intercept
533.177 1 533.177 5.298 0.024
Group Identity (0 = Low, 1 = High)
2107.633 1 2107.633 20.943 0.000*
Covariate: Group Task Ability
207.869
1
207.869
2.066
0.154
Covariate: RPI (0 = Not provided, 1 = Provided)
145.513
1
145.513
1.446
0.233
Identity * RPI
25.430 1 25.430 0.253 0.617
Dependent Variable: Average number of letters decoded by group during main decoding session one minus the average number of letters decoded by group during the last training decoding session. *Significant at p<0.01, two tailed.
Running head: GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION
Table 3. H2: the effect of group-level relative performance information (RPI) on groups’ task performance improvement
Panel A: Change in the Average Number of Letters Decoded (Standard Deviation) by Group: Group-Level RPI
Provided Group-Level RPI Not
Provided
Mean High-Identity Groups + 8.00
(7.95) n = 20
+ 4.40 (8.80) n = 20
+ 6.20 (8.48) n = 40
Low-Identity Groups
+ 15.42 (7.95) n = 24
+ 9.48 (8.80) n = 25
+ 12.30 (10.17) n = 49
Mean
+ 12.05*
(7.95) n = 44
+ 7.22* (9.34) n = 45
+ 9.61 (10.17) n = 89
Dependent Variable: Average number of letters decoded by group during main decoding session two minus the average number of letters decoded by group during main decoding session one. *Significant at p<0.01, one-tailed. Panel B: ANCOVA Results Sum of
Squares df Mean
Square F Significance
Model
1602.847 4 400.712 4.803 0.002
Intercept
34.706 1 34.706 0.416 0.521
RPI (0 = Not provided, 1 = Provided)
466.232 1 466.232 5.588 0.010*
Group Identity
809.677 1 809.677 9.704 0.015*
Identity * RPI
20.092
1
20.092
0.241
0.625
Covariate: Group Task Ability
198.485
1
198.485
2.379
0.127
Dependent Variable: Average number of letters decoded by group during main decoding session two minus the average number of letters decoded by group during main decoding session one. *Significant at p<0.01, two tailed.
Running head: GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION
Table 4. H2b: the effect of group-level relative performance information (RPI) on task improvement among average task ability groups Panel A: Change in the Average Number of Letters Decoded (Standard Deviation) by Average
Ability Groups Group-Level RPI
Provided Group-Level RPI Not
Provided
Mean High-Identity Groups + 10.29
(5.35) n = 7
+ 0.45 (8.09) n = 11
+ 4.28 (8.54) n = 18
Low-Identity Groups
+ 12.45 (11.85) n = 6
+ 4.83 (6.62) n = 6
+ 6.03 (9.99) n = 12
Mean
+ 11.31*
(8.61) n = 13
+ 2.00* (7.70) n = 17
+ 6.03 (9.24) n = 30
Dependent Variable: Average number of letters decoded by average task ability groups during main decoding session two minus the average number of letters decoded by average task ability groups during main decoding session one. *Significant at p<0.01, one-tailed. Panel B: ANCOVA Results Sum of
Squares df Mean
Square F Significance
Model
912.057 4 228.014 3.647 0.018
Intercept
156.004 1 156.004 2.495 0.127
RPI (0 = Not provided, 1 = Provided)
620.986 1 620.986 9.933 0.002*
Group Identity
46.578 1 46.578 0.745 0.396
Identity * RPI 2.737 1 2.737 0.044 0.836 Covariate: Group Task Ability
183.580
1
183.580
2.937
0.099
Dependent Variable: Average number of letters decoded by average task ability groups during main decoding session two minus the average number of letters decoded by average task ability groups during main decoding session one. *Significant at p<0.01, one-tailed.
Running head: GROUP-LEVEL RELATIVE PERFORMANCE INFORMATION
Table 5. Summary of this study’s hypotheses and results
Hypothesis Supported?
H1 A group’s task performance improvement will be greater among high-identity groups than low-identity groups. Yes.
H2 A group’s task performance improvement will be greater when group-level RPI is provided than when group-level RPI is not provided.
Yes.
H2a The performance improvement of top task ability groups will be greater when group-level RPI is provided as compared to when group-level RPI is not provided.
No.
H2b The performance improvement of average task ability groups will be greater when group-level RPI is provided as compared to when group-level RPI is not provided.
Yes.
H2c
The performance improvement of bottom task ability groups will be affected when group-level RPI is provided as compared to when group-level RPI is not provided.
No.
H3 Providing group-level RPI to high-identity groups will yield incremental performance effects beyond those attributable to high-identity alone.
Yes.
H4a Providing group-level RPI to high-identity groups of top task ability will yield incremental performance improvement beyond those attributable to high-identity alone.
No.
H4b Providing group-level RPI to high-identity groups of average task ability will yield incremental performance improvement beyond those attributable to high-identity alone.
Yes.
H4c Providing group-level RPI to high-identity groups of low task ability will yield incremental performance improvement beyond those attributable to high-identity alone.
No.