THE ROLE OF CONTEXT IN WORK TEAM DIVERSITYRESEARCH: A META-ANALYTIC REVIEW

APARNA JOSHIHYUNTAK ROH

University of Illinois at Urbana-Champaign

Integrating macro and micro theoretical perspectives, we conducted a meta-analysisexamining the role of contextual factors in team diversity research. Using data from8,757 teams in 39 studies conducted in organizational settings, we examined whethercontextual factors at multiple levels, including industry, occupation, and team, influ-enced the performance outcomes of relations-oriented and task-oriented diversity. Thedirect effects were very small yet significant, and after we accounted for industry,occupation, and team-level contextual moderators, they doubled or tripled in size.Further, occupation- and industry-level moderators explained significant variance ineffect sizes across studies.

Research in the area of work team diversity hasgrown exponentially in the last four decades. How-ever, several comprehensive reviews have notedthat the findings in this area do not provide a clearconsensus regarding the performance effects ofwork team diversity (Harrison & Klein, 2007; Jack-son, Joshi, & Erhardt, 2003; Milliken & Martins,1996; Van Knippenberg & Schippers, 2007; Wil-liams & O’Reilly, 1998). In some studies, research-ers have reported that team diversity is positivelyassociated with performance (e.g., Ely, 2004; Vander Vegt, Van de Vliert, & Huang, 2005). In anotherset of studies, team diversity has been found tonegatively predict performance (e.g., Jehn, North-craft, & Neale, 1999; Leonard, Levine, & Joshi,2004). A majority of these studies, however, havereported a nonsignificant, direct relationship be-tween team diversity and performance. Further-more, even within studies, the effects of gender,race, age, and tenure diversity on team performancehave varied (e.g., Kirkman, Tesluk, & Rosen, 2004;Kochan et al., 2003).

Current theoretical perspectives framing diver-sity research, such as social identity theory, socialcategorization theory, and the attraction-selection-attrition framework, appear to be insufficient forresolving these mixed findings. In general, current

applications of these theoretical perspectives haveoffered broad generalizations for why differenceswithin work groups may manifest in specific atti-tudinal outcomes, such as conflict or cohesion, orin behavioral outcomes, such as turnover or absen-teeism (see Jackson et al., 2003). In addition toasking why diversity manifests in specific out-comes, a careful examination of situational settingswould also ask when, where, and how diversity dy-namics unfold in workplaces; these contextual con-siderations, not often captured (see Johns, 2006), arepertinent for reconciling the mixed findings from pastresearch. Researchers have attempted to reconcilethese findings within prevalent theoretical traditionsby considering the influence of organization-leveland team-level factors on diversity outcomes (e.g.,Kirkman et al., 2004; Kochan et al., 2003). However,these explanations are often offered post hoc; contex-tual factors have less often been incorporated in hy-pothesis development or in study design. We proposethat a unified and comprehensive contextual frame-work for diversity research has the potential to re-solve these inconsistent findings and can contributeto further theoretical and empirical developments inthe field.

This article highlights contextual issues in teamdiversity research. Our study takes a substantivelydifferent approach from past meta-analytic reviewson this topic (Bowers, Pharmer, & Salas, 2000; Hor-witz & Horwitz, 2007; Webber & Donahue, 2001).Rather than test whether diversity attributes have apositive or a negative effect on team performance,we accounted for aspects of context at multiplelevels and examined whether these contextual fac-tors shaped the team diversity-performance rela-tionship. In proposing and testing the contextual

A version of this paper was presented at the 2008 AnnualMeeting of the Academy of Management and won the Do-rothy Harlow Distinguished Paper Award. We are indebtedto Joseph Martocchio for advice and feedback on drafts ofthis work. We are also extremely grateful to Jason Colquittand three anonymous reviewers for invaluable feedbackthroughout the review process. We thank Niti Pandey andErik Young for research assistance.

� Academy of Management Journal2009, Vol. 52, No. 3, 599–627.

599

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download or email articles for individual use only.

framework presented in this article, we attempt tomake several contributions to work team diversityresearch. First, we develop and test a theoreticallydriven framework for work team diversity contextacross multiple levels of analysis. Several research-ers have acknowledged that contextual consider-ations are critical in diversity research (e.g., Jack-son et al., 2003; Martins, Milliken, Wiesenfeld, &Salgado, 2003). However, a theoretically drivenmultilevel framework explicating contextual deter-minants of work team diversity outcomes has notbeen forthcoming. Our study addresses this gap.Some scholars have also drawn attention to struc-tural and institutional factors that give meaning todemographic differences in organizations (DiTo-maso, Post, & Parks-Yancy, 2007; Ragins & Sund-storm, 1990). Yet the prevalent theoretical perspec-tives we referred to in our introduction do notaccount for these institutional and structural fac-tors. Therefore, a second contribution we attempt isan integration of work team diversity research withmacro theoretical perspectives that account forthese factors and have received relatively scant at-tention in the past. Furthermore, team task charac-teristics can also shape the salience of diversityattributes within teams (Van Knippenberg & Schip-pers, 2007); we also consider whether these charac-teristics are a relevant team-level context influenc-ing the diversity-performance link. Third, weintegrate 15 years of field research on the perfor-mance outcomes of work team diversity. Our re-view distinguishes between the effects of task-ori-ented (e.g., function, education, and tenure) andrelations-oriented (e.g., gender, race/ethnicity, andage) aspects of diversity in relation to performance.Finally, this study answers a call in the broaderdomain of management research for a more contextbased understanding of workplace phenomena(Bamberger, 2008; Johns, 2006; Rousseau & Fried,2001). Context theorizing in management researchcan enhance its “market orientation,” which notonly makes study findings more accessible to man-agers but is also important for future theory build-ing (Bamberger, 2008; Dubin, 1976).

In the following section, we discuss in detail howwe conceptualize diversity context and outline var-ious aspects of this context. Next, we develop hy-potheses that delineate the moderating effects ofthese contextual factors on team diversity in rela-tion to team performance. Finally, we present thefindings from a meta-analytic review and considerthe implications of these contextual considerationsfor future theoretical and empirical developmentsin diversity research.

KEY CONCEPTS ANDTHEORETICAL BACKGROUND

We define diversity as an aggregate team-levelconstruct that represents differences among mem-bers of an interdependent work group with respectto a specific personal attribute (Jackson et al.,2003). In line with past research, we distinguishbetween task-oriented and relations-oriented as-pects of diversity (Jackson, May, & Whitney, 1995).Relations-oriented diversity attributes such as gen-der, race/ethnicity, and age are cognitively accessi-ble, pervasive, and immutable and are associatedwith social categorization processes (Fiske, 1998;Van Knippenberg, De Dreu, & Homan, 2004). Thesesocial categorization–based processes, which man-ifest in intergroup bias and negative attitudestoward dissimilar others in a group may have neg-ative performance consequences. In contrast, task-oriented diversity attributes, such as education,function, and tenure, are associated with skill-basedand informational differences among work groupmembers (Jackson et al., 1995). These aspects of di-versity are assumed to constitute a team’s cognitiveresource base and are associated with elaboration-based processes, defined as the exchange of in-formation and perspectives among group mem-bers, individual-level information processing, gainingfeedback, and integrating information and per-spectives. These elaboration-based processes ex-plain the positive performance outcomes of workgroup diversity (see Van Knippenberg et al., 2004).In the subsequent sections, we discuss how va-rious aspects of diversity context can influence thecategorization-based processes associated with re-lations-oriented diversity and the elaboration-based processes associated with task-oriented di-versity, and implications for team performance.Although more sophisticated conceptualizations ofteam diversity have been developed (e.g., Harrison& Klein, 2007; Lau & Murnighan, 1998), we draw ona more simplified yet established typology of diver-sity attributes to develop our key propositions re-garding the role of contextual factors in diversityresearch.1

Team performance is defined as the extent to

1 Some researchers have also conceptualized team di-versity by differentiating between surface- and deep-level diversity (e.g., Harrison, Price, & Bell, 1998). Otherresearchers have proposed an approach based on “fault-lines” (e.g., Lau & Murnighan, 1998). Recently, Harrisonand Klein (2007) posed “separation,” “variety,” and “dis-parity” as diversity dimensions to consider. We discussthese alternative approaches later, in the Discussionsection.

600 JuneAcademy of Management Journal

which a team accomplishes its goal or mission (De-vine & Phillips, 2001). We focus on performance asan outcome of diversity because this outcome hasreceived the most research attention and representsan area in which the mixed findings have been themost prevalent (Jackson et al., 2003). Results pertain-ing to performance outcomes of diversity may needthe most integration, and understanding the contex-tual factors shaping performance outcomes may havethe most value for resolving past findings.

Context has been defined as the situational set-ting in which workplace phenomena occur (Cap-pelli & Sherer, 1991). In recent theoretical advance-ments, it has been recognized that various aspectsof context may serve as “situational opportunitiesfor and countervailing constraints against organiza-tional behavior [and] be represented as a tensionsystem or force field comprising such opportunitiesand constraints” (Johns, 2006: 387). Drawing onthis perspective, we propose that context can setspecific constraints and opportunities that eitherenhance or minimize the direct effects of workteam diversity on performance. Opening the dis-course on team diversity to contextual influences ischallenging because it requires one to specify team,organizational, and extraorganizational factors thatmay comprise a “tension system” shaping diversityeffects. Therefore, we undertook an extensive re-view of team diversity research conducted in or-ganizational settings from 1992 through 2008 toidentify aspects of diversity context that past re-search has explicitly acknowledged as either mod-erator or as control variables. Table 1 represents thefindings of this review.

Our review indicated that approximately 60 per-cent of the direct effects reported in past researchwere nonsignificant for various diversity attributes.Among the remainder, 20 percent of the effectsreported were significantly positive, and 20 percentwere significantly negative. Researchers have con-sidered contextual variables primarily at the teamlevel to explicate these mixed effects of team diver-sity on performance. Among the studies we reviewed,over 70 percent accounted for team-level contextualfactors such as task interdependence, complexity, cli-mate, and other team-level perceptual variables (e.g.,Jehn et al., 1999; Pelled, Eisenhardt, & Xin, 1999;Schippers, Den Hartog, Koopman, & Wienk, 2003;Van der Vegt & Bunderson, 2005). Approximately 20percent of the studies represented in Table 1 exam-ined contextual moderators at the organizationallevel, including organizational demography, diver-sity training participation, and organizational culture(e.g., Ely, 2004; Jackson & Joshi, 2004; Jehn &Bezrukova, 2004). Less than 10 percent of the studiesreviewed examined extraorganizational factors.

Among the studies that did incorporate these vari-ables, the focus was on national culture, customerbase demography, market competition, and rate oftechnological change (e.g., Ancona & Caldwell, 1992;Lovelace, Shapiro, & Weingart, 2001; Leonard et al.,2004; Reagans & Zuckerman, 2001; Van der Vegt etal., 2005). On the basis of this review, we aimed totake stock of contextual effects that are conceptuallydistinct and have been considered in past researchand also to incorporate additional contextual vari-ables that have received less attention in the past.

Theoretical and practical considerations gov-erned our efforts to identify key contextual factors.Sociopsychological theoretical perspectives sug-gest that the demography of the job or occupationin which diverse work groups are embedded canshape categorization-based processes in thesegroups (DiTomaso, Post, & Parks-Yancy, 2007; Lar-key, 1996; Reskin, McBrier, & Kmec, 1999).2 Unlikein a balanced setting, in an occupational contextdominated by a single demographic group, negativestereotypes about underrepresented groups can in-fluence categorization-based outcomes withinwork groups (Hilton & Von Hippel, 1996; Larkey,1996). We extend these perspectives to focus onoccupational demography as a contextual factorthat can enhance or minimize the influence ofteam diversity on performance outcomes.

We also consider industry as an embedding con-text for diversity-based outcomes. Strategic man-agement research suggests that interindustry varia-tion in levels of technological change, regulatorypressure, customer demands, and market competi-tion are greater than intraindustry variations (Bour-geois, 1985; Porter, 1980). Furthermore, these in-dustry-level contingencies can serve as situationalenhancers or minimizers of diversity effects on per-formance (Hambrick, Cho, & Chen, 1996; Richard,Murthi, & Ismail, 2007). Within the strategic man-agement domain, considerable research on topmanagement teams and research on the firm diver-sity–performance link has incorporated industry-level context (e.g., service versus manufacturing) asa key moderator (e.g., Haleblian & Finkelstein,1993; Hambrick et al., 1996; Keck, 1997; Richard etal., 2007). We extend these perspectives to considerindustry setting as a relevant aspect of a team’s

2 We acknowledge that the proximal work context rep-resented by organization-level demography, culture, andclimate has received some attention in past research andthat these are also important contextual variables to con-sider. However, of the studies included in this review, onlythree provided information regarding organization-level de-mography, culture, or climate, and therefore these variablescould not be included in the meta-analysis.

2009 601Joshi and Roh

TA

BL

E1

Su

mm

ary

ofT

eam

Div

ersi

tyR

esea

rch

,19

92–2

008a

Stu

die

sD

iver

sity

Att

ribu

tes

Con

sid

ered

Sam

ple

and

Fin

din

gs

Con

text

Var

iabl

esC

onsi

der

edat

Dif

fere

nt

Lev

els

Tea

mO

rgan

izat

ion

Ind

ust

ry/O

ccu

pat

ion

Oth

er

Ain

oya

(200

4)b

Rac

e,te

nu

re59

wor

kte

ams

from

vari

ous

ind

ust

ries

.N

od

irec

tre

lati

onsh

ips,

and

lim

ited

sup

por

tfo

rm

oder

atin

gef

fect

s.

Tas

kin

terd

epen

den

ce,

con

flic

tm

anag

emen

tp

ract

ice

An

con

a&

Cal

dw

ell

(199

2)F

un

ctio

n,

ten

ure

45n

ewp

rod

uct

dev

elop

men

tte

ams

inh

igh

-tec

hco

mp

anie

s.P

arti

alsu

pp

ort

for

dir

ect

rela

tion

ship

s.

Res

ourc

eav

aila

bili

tyd

Com

pan

yca

pab

ilit

y(p

rod

uct

dev

elop

men

tex

per

ien

ce)d

Mar

ket

com

pet

itio

nd

Bal

kun

di,

Kil

du

ff,

Bar

snes

s,&

Mic

hae

l(2

007)

Eth

nic

ity,

gen

der

,ag

e19

pro

du

ctio

nan

dm

ain

ten

ance

team

sat

aw

ood

pro

du

cts

com

pan

y.N

od

irec

tre

lati

onsh

ips.

Bau

gh&

Gra

en(1

997)

Gen

der

,ra

ce31

pro

fess

ion

alp

roje

ctte

ams

ina

gove

rnm

ent

agen

cy.

No

sign

ific

ant

mai

nef

fect

s.

Bu

nd

erso

n(2

003)

Sp

ecif

icst

atu

scu

es(c

omp

any

and

ind

ust

ryte

nu

re,

edu

cati

on),

dif

fuse

stat

us

cues

(gen

der

,et

hn

icit

y)c

35p

rod

uct

ion

team

sin

ah

igh

-tec

hco

mp

any.

No

dir

ect

rela

tion

ship

sbu

tsu

pp

ort

for

mod

erat

ing

effe

cts.

Tea

mp

roce

ss(i

nfl

uen

cece

ntr

alit

y),

team

ten

ure

,le

ader

role

d

Cad

y&

Val

enti

ne

(199

9)R

ace,

gen

der

,ag

e,fu

nct

ion

50p

robl

emso

lvin

gte

ams

ina

hig

hte

chco

mp

any.

Par

tial

sup

por

tfo

rd

irec

tre

lati

onsh

ips.

Cam

pio

n,

Med

sker

,&

Hig

gs(1

993)

Exp

erie

nce

80ad

min

istr

ativ

ew

ork

team

sin

afi

nan

cial

serv

ice

com

pan

y.L

imit

edsu

pp

ort

for

mai

nef

fect

s.

Tra

inin

g,m

anag

eria

lsu

pp

ort,

com

mu

nic

atio

nan

dco

oper

atio

nbe

twee

ngr

oup

sd

Cam

pio

n,

Pap

per

,&

Med

sker

(199

6)E

xper

ien

ce60

pro

fess

ion

alte

ams

ina

fin

anci

alse

rvic

eco

mp

any.

Lim

ited

sup

por

tfo

rm

ain

effe

cts.

Tra

inin

g,m

anag

eria

lsu

pp

ort,

com

mu

nic

atio

nan

dco

oper

atio

nbe

twee

ngr

oup

sd

Ch

atm

an&

Fly

nn

(200

1)R

ace,

gen

der

,ci

tize

nsh

ip16

1m

anag

ers

from

10bu

sin

ess

un

its

ofa

fin

anci

alse

rvic

efi

rm.

Par

tial

sup

por

tfo

rd

irec

tre

lati

onsh

ips.

Ch

oi(2

007)

Age

,ge

nd

er,

ten

ure

,fu

nct

ion

188

wor

kte

ams

ina

larg

eel

ectr

onic

sco

mp

any

inK

orea

.P

arti

alsu

pp

ort

for

dir

ect

rela

tion

ship

s.

Ch

oi,

Pri

ce,

&V

inok

ur

(200

3)A

ge,

gen

der

,ra

ce,

edu

cati

on16

9tr

ain

ing

grou

ps

from

3or

gan

iza-

tion

s.P

arti

alsu

pp

ort

for

mai

nef

fect

s.

Col

quit

t,N

oe,

&Ja

ckso

n(2

002)

Eth

nic

ity,

gen

der

,ag

e88

pro

du

ctio

nte

ams.

No

dir

ect

rela

tion

ship

and

lim

ited

sup

por

tfo

rm

oder

atin

gef

fect

.

Pro

ced

ura

lju

stic

ecl

imat

e

Dra

ch-Z

ahav

y&

Som

ech

(200

2)T

ask-

rela

ted

(fu

nct

ion

,ed

uca

tion

,te

amte

nu

re),

rela

tion

s-or

ien

ted

(gen

der

,ag

e)

48ad

min

istr

ativ

ete

ams

inm

idd

le/

hig

hsc

hoo

lsin

Isra

el.

Par

tial

sup

por

tsfo

rd

irec

tre

lati

onsh

ips.

Tea

mp

roce

ss(f

requ

ency

ofm

eeti

ngs

)d

Ely

(200

4)A

ge,

gen

der

,ra

ce,

ten

ure

486

reta

ilba

nk

bran

ches

.P

arti

alsu

pp

ort

for

mai

nan

dm

oder

atin

gef

fect

s.

Tea

mp

roce

ss(c

oop

erat

ive

team

pro

cess

)

Div

ersi

tytr

ain

ing

pro

gram

Her

ron

(199

3)b

Gen

der

,et

hn

icit

y93

wor

kte

ams

ina

larg

eae

rosp

ace

com

pan

y.M

ixed

sup

por

tfo

rm

ain

effe

ctre

lati

onsh

ips.

Jack

son

&Jo

shi

(200

4)G

end

er,

eth

nic

ity,

team

ten

ure

365

sale

ste

ams

ina

larg

eF

ortu

ne

500

com

pan

y.N

od

irec

tre

lati

onsh

ips

but

par

tial

sup

por

tfo

rm

oder

atin

gef

fect

s.

Mu

ltid

imen

sion

alte

amd

iver

sity

(in

tera

ctio

ns

amon

gat

trib

ute

s)

Wor

ku

nit

dem

ogra

ph

y,m

anag

erd

emog

rap

hy

Jarr

ell

(200

2)b

Fu

nct

ion

,ed

uca

tion

leve

l11

0in

terd

isci

pli

nar

yh

ealt

hca

rete

ams.

Tas

kin

terd

epen

den

ce,

task

com

ple

xity

Sig

nif

ican

tsu

pp

ort

for

mai

nan

dm

oder

atin

gef

fect

s.

Jeh

n&

Bez

ruko

va(2

004)

Gen

der

,ra

ce,

age,

team

ten

ure

,fu

nct

ion

,ed

uca

tion

alle

vel

1,52

8te

ams

ina

larg

ein

form

atio

np

roce

ssin

gfi

rm.

Par

tial

sup

por

tsfo

rm

ain

and

mod

erat

ing

effe

cts.

Org

aniz

atio

ncu

ltu

re,

busi

nes

sst

rate

gy,

HR

pra

ctic

eor

ien

tati

on

Jeh

n,

Nor

thcr

aft,

&N

eale

(199

9)S

ocia

lca

tego

ry(a

ge,

gen

der

),in

form

atio

n(e

du

cati

on,

fun

ctio

n)

92w

ork

team

sin

the

hou

seh

old

good

sm

ovin

gin

du

stry

.S

up

por

tsfo

rm

ain

and

mod

erat

ing

effe

cts.

Tas

kin

terd

epen

den

ce,

task

typ

e(c

omp

lexi

ty)

Josh

i(2

002)

bA

ge,

race

,ge

nd

er,

ten

ure

,ed

uca

tion

772

serv

ice

team

sin

am

anu

fact

uri

ng

com

pan

y.O

rgan

izat

ion

ald

emog

rap

hy

Par

tial

sup

por

tfo

rm

ain

and

mod

erat

ing

rela

tion

ship

s.

Kea

rney

&G

eber

t(2

009)

Age

,ed

uca

tion

,n

atio

nal

ity

62R

&D

team

sin

aG

erm

anp

har

mac

euti

cal

com

pan

y.N

od

irec

tre

lati

onsh

ipbu

tsi

gnif

ican

tm

oder

atin

gef

fect

s.

Tra

nsf

orm

atio

nal

lead

ersh

ip,

team

lon

gevi

ty,d

task

inte

rdep

end

ence

d

Kea

rney

,G

eber

t,&

Voe

lpel

(200

9)A

ge,

edu

cati

on,

gen

der

,c

nat

ion

alit

y,c

ten

ure

c83

team

sfr

omG

erm

anor

gan

iza-

tion

s.S

up

por

tfo

rsi

gnif

ican

tm

oder

atin

gef

fect

s.

Tea

mm

embe

rs’

nee

dfo

rco

gnit

ion

,ta

skin

terd

epen

den

ce,d

task

com

ple

xity

,d

team

lon

gevi

tyd

Kel

ler

(200

1)F

un

ctio

n93

R&

Dte

ams.

Sig

nif

ican

tm

ain

effe

cts.

Kir

kman

,T

eslu

k,&

Ros

en(2

004)

Age

,ge

nd

er,

race

,co

mp

any

ten

ure

,te

amte

nu

re11

1w

ork

team

sfr

omfo

ur

dif

fere

nt

orga

niz

atio

ns.

Few

mai

nef

fect

san

dp

arti

alsu

pp

ort

for

mod

erat

ing

effe

cts.

Lea

der

-mem

ber

dem

ogra

ph

icm

atch

Org

aniz

atio

nal

wor

ken

viro

nm

entd

Con

tin

ued

onfo

llow

ing

pag

e

TA

BL

E1

Con

tin

ued

Stu

die

sD

iver

sity

Att

ribu

tes

Con

sid

ered

Sam

ple

and

Fin

din

gs

Con

text

Var

iabl

esC

onsi

der

edat

Dif

fere

nt

Lev

els

Tea

mO

rgan

izat

ion

Ind

ust

ry/O

ccu

pat

ion

Oth

er

Leo

nar

d,

Lev

ine,

&Jo

shi

(200

4)R

ace/

eth

nic

ity,

gen

der

,ag

e70

0re

tail

stor

es.

Lim

ited

sup

por

tfo

rm

ain

effe

cts

and

no

sup

por

tfo

rm

oder

atin

gef

fect

s.

Com

mu

nit

yd

emog

rap

hy

(cu

stom

erd

emog

rap

hic

mat

ch)

Lov

elac

e,S

hap

iro,

&W

ein

gart

(200

1)F

un

ctio

n43

new

pro

du

ctd

evel

opm

ent

team

sin

ah

igh

tech

ind

ust

ry.

No

dir

ect

rela

tion

ship

but

sign

ific

ant

mod

erat

ing

effe

cts.

Tea

mp

roce

ss(c

omm

un

icat

ion

styl

e),

lead

eref

fect

iven

ess

Tec

hn

olog

ical

chan

ged

Mar

stel

ler

(200

3)b

Age

,ge

nd

er,

edu

cati

on,

ten

ure

40h

ealt

hca

rete

ams.

Par

tial

sup

por

tfo

rm

ain

effe

cts.

Org

aniz

atio

nty

ped

May

o,P

asto

r,&

Mei

nd

l(1

996)

Gen

der

,ra

ce,

age,

ten

ure

68w

ork

team

sfr

omva

riou

sin

du

stri

es.

Lea

der

dem

ogra

ph

yd

Sig

nif

ican

tm

ain

effe

cts.

Pel

led

(199

6)R

ace,

gen

der

,co

mp

any

ten

ure

42p

rod

uct

ion

team

sin

man

ufa

ctu

rin

gfa

cili

ties

.N

od

irec

tre

lati

onsh

ips.

Pel

led

,E

isen

har

dt,

&X

in(1

999)

Gen

der

,ra

ce,

age,

com

pan

yte

nu

re,

fun

ctio

n45

team

sin

elec

tron

icd

ivis

ion

s.N

od

irec

tre

lati

onsh

ips

but

sign

ific

ant

mod

erat

ing

effe

cts.

Tas

kro

uti

nen

ess,

grou

plo

nge

vity

Org

aniz

atio

nal

wor

ken

viro

nm

entd

Por

tero

-Bro

wn

(199

9)b

Age

,ge

nd

er,

eth

nic

ity,

fun

ctio

n,

com

pan

yte

nu

re37

R&

Dte

ams.

Par

tial

sup

por

tfo

rm

ain

and

mod

erat

ing

effe

cts.

Tea

mid

enti

fica

tion

,co

llec

tive

nor

m,

team

pro

cess

Pu

ck,

Ryg

l,&

Kit

tler

(200

6)E

thn

icit

y20

pro

cess

imp

rove

men

tte

ams

ina

Ger

man

spor

tsw

ear

com

pan

y.L

imit

edsu

pp

ort

for

the

dir

ect

rela

tion

ship

.

Rea

gan

s&

Zu

cker

man

(200

1)C

omp

any

ten

ure

c22

4co

rpor

ate

R&

Dte

ams.

No

dir

ect

rela

tion

ship

.T

eam

task

typ

edM

arke

tco

mp

etit

ion

d

Rea

gan

s,Z

uck

erm

an,

&M

cEvi

ly(2

004)

Ed

uca

tion

,ge

nd

er,

ten

ure

1,51

8p

roje

ctte

ams

ina

pro

fess

ion

alR

&D

firm

.N

od

irec

tre

lati

onsh

ips.

Ric

har

d(2

000)

Rac

e63

ban

ks.

No

dir

ect

rela

tion

ship

but

sign

ific

ant

mod

erat

ing

effe

cts.

Bu

sin

ess

stra

tegy

(gro

wth

vs.

dow

nsi

zin

g)

Ric

har

d,

Bar

net

t,D

wye

r,&

Ch

adw

ick

(200

4)

Rac

e,ge

nd

er15

3m

anag

emen

tte

ams

ina

ban

kin

gin

du

stry

.N

od

irec

tre

lati

onsh

ips

but

sign

ific

ant

mod

erat

ing

effe

cts.

Bu

sin

ess

stra

tegy

(in

nov

atio

n/r

isk

taki

ng)

Sac

co&

Sch

mit

t(2

005)

Rac

e,ge

nd

er,

age

2,37

3qu

ick

serv

ice

rest

aura

nts

.P

arti

alsu

pp

orts

for

mai

nef

fect

sbu

tn

om

oder

atin

gre

lati

onsh

ip.

Com

mu

nit

yd

emog

rap

hic

mat

ch

Sch

ipp

ers,

Den

Har

tog,

Koo

pm

an,

&W

ien

k(2

003)

Age

,ge

nd

er,

edu

cati

onle

vel,

team

ten

ure

54w

ork

team

sin

the

Net

her

lan

ds.

Few

mai

nef

fect

sbu

tsi

gnif

ican

tm

oder

atin

gef

fect

s.

Ou

tcom

ein

terd

epen

den

ce,

grou

plo

nge

vity

Som

ech

(200

6)F

un

ctio

n13

6p

rim

ary

hea

lth

care

team

sin

Isra

el.

Lea

der

ship

styl

e

No

dir

ect

rela

tion

ship

san

dp

arti

alsu

pp

ort

for

mod

erat

ing

effe

cts.

Tyr

an&

Gib

son

(200

8)G

end

er,

eth

nic

ity,

ten

ure

,co

llec

tivi

sm57

ban

kbr

anch

team

s.P

arti

alsu

pp

ort

for

mai

nef

fect

rela

tion

ship

s.

Tea

mou

tcom

ety

pe

Van

Der

Veg

t&

Bu

nd

erso

n(2

005)

Exp

erti

se(f

un

ctio

n)

57R

&D

team

sin

oil

and

gas

ind

ust

ryin

the

Net

her

lan

ds.

No

dir

ect

rela

tion

ship

but

sign

ific

ant

mod

erat

ing

effe

cts.

Col

lect

ive

team

iden

tifi

cati

on,

team

pro

cess

(tas

kan

dre

lati

onsh

ipco

nfl

ict,

com

mu

nic

atio

n)d

Van

Der

Veg

t,V

anD

eV

lier

t,&

Hu

ang

(200

5)

Age

,ge

nd

er,

ten

ure

,fu

nct

ion

270

man

ufa

ctu

rin

gu

nit

sof

am

ult

inat

ion

alco

mp

any.

Par

tial

sup

por

tfo

rm

ain

rela

tion

ship

san

dsi

gnif

ican

tsu

pp

ort

for

mod

erat

ing

effe

cts.

Job

typ

e,jo

bch

alle

nge

dN

atio

nal

pow

erd

ista

nce

(cu

ltu

re)

Yeh

&C

hou

(200

5)F

un

ctio

n,

pos

itio

n88

cros

s-fu

nct

ion

alte

ams.

Su

pp

ort

for

mai

nef

fect

sbu

tn

osu

pp

ort

for

the

mod

erat

ing

effe

ct.

Tea

mle

arn

ing

beh

avio

rs

aN

�43

,co

mp

risi

ng

fiel

dst

ud

ies

con

du

cted

du

rin

gth

e19

92–2

008

per

iod

.b

Un

pu

blis

hed

.c

Div

ersi

tyat

trib

ute

sco

nsi

der

edas

con

trol

vari

able

sin

the

rese

arch

des

ign

.d

Con

text

ual

fact

ors

incl

ud

edas

con

trol

vari

able

s.

diversity context. On a practical note, informationregarding occupational and industry setting wasavailable from most study descriptions. Addition-ally, publicly accessible databases provide a wealthof objective information regarding the occupationaland industry contexts in which teams are embed-ded. Therefore, we focus on industry setting andoccupational demography as contextual factors thatprovide a comprehensive understanding of a team’smacrolevel context and could potentially accountfor the mixed findings reported.

The nature of a team’s task can have a significantinfluence on the extent to which team members areinterdependent in terms of goals and task outcomes(see Ilgen, Hollenbeck, Johnson, & Jundt, 2005;LePine, Hanson, Borman, & Motowidlo, 2000). Asthe review above suggests, team-level contextualvariables have received the most attention in pastresearch. Social categorization theory would sug-gest that aspects of a team’s task can minimize thesalience of diversity attributes by reinforcing acommon group identity or by placing demands onthe team’s diverse cognitive resource base (Gaert-ner & Dovidio, 2000; Jehn et al., 1999). Informationprocessing theory would suggest that the nature oftasks would place requirements on a team’s cogni-tive resource base with implications for the sa-lience of diversity attributes within the team (Jehnet al., 1999; Williams & O’Reilly, 1998). Our reviewalso indicated that information regarding aspects ofteams’ tasks, such as team interdependence andteam type (i.e., short-term versus long-term), wasmore consistently available across studies. There-fore, we chose to focus on task interdependenceand team type as aspects of team-level context.Ultimately, we sought to examine the effects ofconceptually distinct contextual variables that to-gether provide a comprehensive picture of contextat the team level. In the subsequent sections, wediscuss in greater detail how these aspects of occu-pational, industry, and team-level context mayshape the performance outcomes of both task-ori-ented and relations-oriented team diversity.

A CONTEXTUAL FRAMEWORK FOR TEAMDIVERSITY RESEARCH

Occupational Demography

In the United States, workers are classified intooccupational categories based on work performed,skills, education, training, and credentials. Someoccupations are found in just one or two industries(e.g., post office clerks); however, many occupa-tions are found in a large number of industries (e.g.,auditors, accountants, software developers, net-

work analysts). Thus, the term “occupation” refersto this collective description of a number of indi-vidual jobs performed, with minor variations, inmany establishments (U. S. Bureau of Labor Statis-tics, 2007). A significant body of sociological re-search shows that occupation-level demographiccomposition can have important implications forgender-based and ethnicity- or race-based diversityin organizations (for a review, see Reskin et al.[1999]). We propose below that occupational de-mography serves as a situational setting that canenhance the effects of relations-oriented diversityand minimize the effects of task-oriented diversityon team performance.

Considerable research on stereotype formationenables us to specify the psychological processesby which an occupational demography can influ-ence diversity dynamics within teams. This re-search suggests that stereotypes, which emerge inresponse to environmental factors—such as differ-ences in social roles (Eagly, 1995) and in power(Fiske, 1993)—and as a means to justify the statusquo (Jost & Banaji, 1994; Sidanius, 1993), result incategorization-based responses toward targetedgroups (Allport, 1954). When a single demographicgroup dominates an occupation, negative stereo-type-based categorization processes against under-represented groups are likely (Fiske, 1993; Reskinet al., 1999). Once categorization-based processesare invoked, additional information regarding tar-geted group members is filtered out, and individu-ating processes are blocked (Allport, 1954; Brewer,1988). In diverse teams, when the environmentprimes negative stereotypes against specific demo-graphic groups, these categorization-based pro-cesses are likely to influence interactions. In workenvironments, where stereotypes are less salient,individuating information regarding demographi-cally dissimilar individuals is likely to be cogni-tively acceptable and result in more positive inter-actions (see Larkey, 1996). Integrating theseperspectives, we propose that in occupational set-tings dominated by a single demographic group, itis likely that stereotypic reactions against under-represented groups will be triggered (Fiske, 1993;Reskin et al., 1999). These reactions result in cate-gorization-based processes that hinder effectivegroup interactions, with negative performance con-sequences (Larkey, 1996). These categorization-based processes are less likely in demographicallybalanced occupational settings.

A rich body of research on the age, race, andgender typing of jobs corroborates our argumentthat the demographic attributes of a job category oroccupation are associated with job stereotypes thatmay trigger social categorization processes based

606 JuneAcademy of Management Journal

on these demographic attributes (Eagly & Steffen,1984; Perry, 1997). For example, consider the male-dominated occupational category of production en-gineers. On the basis of the research reviewedabove, we would surmise that female engineers (anunderrepresented group) are likely to be targets ofnegative stereotypes (e.g., “women are less techni-cally competent”) enhancing the salience of genderas a basis for categorization when men and womenwork together in a production team (Eagly & Stef-fen, 1984; Ely, 1994, 1995). We propose that genderdiversity is likely to have negative performanceeffects in this setting. In a more gender balancedoccupational setting, such as postsecondary sci-ence teaching, on the other hand, this researchsuggests that gender may not emerge as a salientbasis for categorization within a team of teachersworking on organizing a school science fair. In thissetting, gender diversity is less likely to have anegative influence on team performance. A similarargument can be made with regard to the effects ofoccupational race/ethnic composition on the per-formance outcomes of race/ethnicity-based diver-sity in teams (Reskin et al., 1999).

Regarding the effects of occupational age compo-sition, a significant body of research suggests thatnegative stereotypes against older workers arefairly prevalent and can have detrimental conse-quences for these workers (Fiske, Cuddy, Glick, &Xu, 2002; Rosen & Jerdee, 1977; see Shore andGoldberg [2005] for a recent review). Althougholder workers may experience more unfavorablejob outcomes overall, research also indicates thatthe age composition of a job context can influenceperformance ratings and advancement potential forolder employees. This research suggests that olderworkers may face more unfavorable outcomes inoccupations with fewer older workers (Cleveland,Festa, & Montgomery, 1988; Cleveland, Montgom-ery, & Festa, 1984). In view of this research, wewould propose that among, for example, softwareprogrammers (an occupation dominated byyounger workers), older programmers may be tar-gets of negative stereotypes (e.g., viewed as lessskilled or motivated); the implications of age diver-sity–based outcomes would thus be negative whenolder and younger team members collaborate on ajoint software development project. In work set-tings with a higher proportion of older workers (forinstance, among welfare service workers or postoffice clerks), these workers are less likely to facestereotypes and discrimination (Shore & Goldberg,2005). In these settings, categorization-based effectsbased on age are less likely to disrupt groups’functioning.

To date, diversity research has not addressed the

role of occupational demography in shaping perfor-mance outcomes within diverse teams. At the firmlevel, Frink et al. (2003) found that an invertedU-shaped relationship between gender composi-tion and performance was only observed in gender-balanced occupational settings and not in male-dominated settings. The authors noted that thesefindings may reflect the inability of organizationsin male-dominated contexts to capitalize on thebenefits of gender diversity (Frink et al., 2003).Although the researchers had not accounted forthis contextual factor in their theoretical frame-work, hypotheses, or study design, they discussedthe importance of occupational demography in ex-plaining their findings.

Cumulatively, the research perspectives dis-cussed above suggest that in occupational settingsdominated by a single demographic group, diverseteams may face performance losses primarily fortwo reasons. First, these teams may perform subop-timally because the work context enhances stereo-typing and bias against underrepresented demo-graphic groups that triggers social categorizationbased on these attributes within the teams (Di-Tomaso, Post, & Parks-Yancy, 2007; Skaggs & Di-Tomaso, 2004; Van Knippenberg et al., 2004). Sec-ond, in these settings, teams with higherproportions of underrepresented group members(e.g., women or ethnic minorities) may be valuedless and receive poorer performance ratings or ac-cess to resources, which is likely to impact subjec-tive or objective performance outcomes (Baugh &Graen, 1997; Hultin & Szulkin, 1999; Joshi, Liao, &Jackson, 2006). In view of these considerations,we proposed that occupational demography willmoderate the relationship between relations-ori-ented diversity and team performance and testedspecific aspects of that general relationship meta-analytically with the following:

Hypothesis 1a. Occupational gender composi-tion moderates the negative effect of team gen-der diversity on performance. The negative ef-fect of gender diversity on performance isweaker in gender-balanced settings.

Hypothesis 1b. Occupational race/ethnic com-position moderates the negative effect of teamrace/ethnic diversity on performance. The neg-ative effect of race/ethnic diversity on perfor-mance is weaker in racially/ethnically bal-anced settings.

Hypothesis 1c. Occupational age compositionmoderates the negative effect of team age di-versity on performance. The negative effect of

2009 607Joshi and Roh

age diversity on performance is weaker in rel-atively age-balanced settings.

So far we have argued that occupational demog-raphy creates a context that may enhance or mini-mize categorization-based processes in workgroups. Our arguments raise the possibility thatoccupational demography will moderate the rela-tionship between task-oriented diversity and teamperformance as well. We surmise that when a sin-gle demographic group dominates an occupation,relations-oriented diversity may be correlated withtask-oriented diversity. For example, white menmay be more tenured and have a specific educa-tional background. In these settings, individuals’task-relevant contributions may be confoundedwith their demographic attributes (Berger, Fisek,Norman, & Zelditch, 1977) so that any positiveoutcomes of task-oriented diversity on performancemay also be mitigated. Researchers have also notedthat when the negative effects of categorization-based diversity are salient, information processingin groups is disrupted (Van Knippenberg et al.,2004). Some studies have shown that when rela-tions-oriented diversity is associated with greaterconflict, task-oriented diversity is also less likely tohave positive outcomes (Homan & Van Knippen-berg, 2003; Jehn et al., 1999). In view of these per-spectives, to account for the possibility that occu-pational demography may have implications forthe outcomes of task-oriented diversity, we proposethe following broad hypothesis:

Hypothesis 1d. Occupational demographymoderates the positive effect of task-orienteddiversity on team performance. The effects oftask-oriented diversity are stronger in morebalanced settings.

Industry Setting

Industry settings refer to fairly specific businessenvironments in which teams are nested that mayhave important implications for team diversity dy-namics that go over and above the occupationaleffects discussed above (Batt, 2002; Datta, Guthrie,& Wright, 2005). Porter’s (1980) analysis identifiesfactors such as customers, suppliers, and regulatorygroups that vary by industry and pose varying con-tingencies for firms. Contingency theory suggeststhat these industry factors provide firms with op-portunities as well as challenges for utilizing keyorganizational resources to enhance performance(Burns & Stalker, 1961; Drazin & Van de Ven, 1985;Lawrence & Lorsch, 1967; Porter, 1980). Drawingon this perspective, a significant body of strategicmanagement research has identified industry con-

text as a key contingency that influences the rela-tionship between organizational processes/prac-tices and performance outcomes (see Combs, Lieu,Hall, and Ketchen [2006] for a meta-analysis). In-dustry-level context has received some, albeit lim-ited, attention in past diversity research (e.g., An-cona & Caldwell, 1992; Lovelace et al., 2001;Reagans & Zuckerman, 2001). On the basis of stra-tegic management perspectives and in line with ourconceptualization of context, we consideredwhether contingencies associated with three spe-cific contexts—service, manufacturing, and hightechnology—may serve as situational enhancers orminimizers of diversity effects on performance out-comes. Together, these three industrial settings in-corporate a bulk of the research settings consideredin past diversity research.

Service industries, defined as customer-orientedindustries that require front-line customer interac-tion and engagement, include retail trade, hospital-ity, and education (U.S. Census Bureau, 2002).Relative to manufacturing industries, service in-dustries are characterized by more frequent andcloser interactions with customers and a greateremphasis on discretionary behavior that can man-ifest directly in performance outcomes such assales, customer satisfaction, and customer reten-tion (Datta et al., 2005). Researchers have arguedthat increasing demographic attribute–based diver-sity can enhance a firm’s “market competence,”which is a form of competitive advantage in theservice industry (Richard et al., 2007). Consider asan example that a retail store with a diverse groupof store employees is more likely to attract diversecustomers and thereby also more likely to havestrong store sales. On the other hand, a competingstore that is unable to enhance employee diversityto attract customers and increase market share islikely to perform poorly in comparison. On thebasis of these considerations, we would expect thataspects of service settings such as direct customercontact and higher levels of discretionary behaviorswould serve as situational enhancers of relations-oriented diversity effects on performance. In theservice industry, since these aspects of diversitycan be considered as a form of market competence,we propose that any negative effects of gender,race, and age diversity are likely to be reversed.

Manufacturing industries are based on the fabri-cation, processing, or preparation of products fromraw materials and commodities. These industriesare typically highly capital-intensive and includeautomobile manufacturing, chemical manufactur-ing, and paper and wood manufacturing (U. S. Cen-sus Bureau, 2002). In contrast to service industryfirms, manufacturing firms rely more on plant and

608 JuneAcademy of Management Journal

equipment, technology, and raw materials toachieve business goals (Quinn, Anderson, & Finkel-stein, 1996). Because manufacturing relies more onphysical capital and equipment and less on directcustomer-based interactions, we would expect di-versity attributes to be less likely to directly impactperformance in this setting (Richard et al., 2007).Further, Combs and colleagues (Combs, Hall, &Ketchen, 2006) noted that manufacturing indus-tries are more likely than service industries to im-plement total quality management techniques, reg-ular training, and formalized human resource (HR)practices because of the need to monitor employeesand to develop their knowledge, skill, and ability toutilize expensive and sometimes dangerous ma-chinery. Consider, as an example, the automobilemanufacturing industry, which has implemented anumber of total quality management techniques in-volving team-based interventions and undertakenregular training for employees. These attributes ofthe automobile industry may serve to buffer anydirect effects of team diversity. Thus, we surmisethat aspects of manufacturing settings, such as re-liance on machinery and HR practices that involvegreater monitoring of employee behavior, may actas situational minimizers of diversity effects onperformance outcomes.

High-technology industries follow invention andinnovation in business strategy and compete inglobal and short-cycle product-markets (cf. Milkov-ich, 1987). In contrast to service and manufacturingindustries, these industries rely more heavily onintellectual capital and invest significantly more inresearch and development (Organisation for Eco-nomic Cooperation and Development, 2006). In-dustries such as information technology, biomedi-cal technology, telecommunication, and dataservices are included in this category. Researchershave begun to examine this particular setting as adistinct context framing firm practices and out-comes (Collins & Smith, 2006). This research hassuggested that employers in this setting are alsolikely to adopt commitment-based practices aimedat recruiting and retaining highly skilled employ-ees, which can create a climate fostering knowledgeexchange and combination (Collins & Smith, 2006).These situational contingencies may enhance task-oriented diversity effects. Relative to service andmanufacturing settings, given contingencies suchas rapidly changing technology, reliance on intel-lectual capital, and a need for creativity and inno-vation in a dynamic environment, we would expecttask-oriented attributes that form a team’s cognitiveresource base to significantly impact performanceoutcomes in this setting.

Although a significant body of research on top

management teams has examined industrial envi-ronment as a relevant influence on the outcomes ofdiversity (e.g., Hambrick et al., 1996; Hambrick &Finkelstein, 1987; Keck, 1997), research on lower-level teams has rarely taken this contextual factorinto account. In general, top management team re-search suggests that industry attributes, such as rateof technological change, munificence, and environ-mental uncertainty, can enhance the salience ofdiversity in top management teams in relation tofirm performance (see Carpenter, Geletkanycz, andSanders [2004], and see Hambrick and Finkelstein[1987] for a review). Richard and colleagues (2007)examined whether industry (service versus manu-facturing) moderated the relationship between ra-cial diversity and firm performance in a sample ofover 800 large U.S. companies; in support of thepropositions, results indicated that the relationshipbetween racial diversity and firm performance wasstronger in service than in manufacturing firms(Richard et al., 2007).

On the basis of the theoretical and empiricalperspective detailed above, we would expect thethree distinct industrial settings to impose varyingcontingencies on diversity outcomes. In service set-tings, customer-based contingencies may enhancepositive performance outcomes of relations-orienteddiversity. In manufacturing settings, a reliance onphysical equipment and standardized practices tomonitor employee behavior may minimize any directperformance outcomes of diversity; and in high-technology settings, technological and intellectualcapital–based contingencies may enhance positiveeffects of task-oriented diversity. Thus, we proposethat industry setting moderates the relationship be-tween relations- and task-oriented diversity andteam performance. Specifically:

Hypothesis 2a. Relations-oriented diversity islikely to have a positive effect on performance inservice industries. In manufacturing and high-technology settings, relations-oriented diversityis less likely to have a significant effect onperformance.

Hypothesis 2b. The positive effect of task-orienteddiversity on performance is stronger in high-tech-nology industries than in manufacturing and ser-vice settings.

Team-Level Diversity Context

We propose below that specific team character-istics—team interdependence and team type—con-stitute a team-level diversity context that canenhance or minimize the direct effects of relations-

2009 609Joshi and Roh

oriented and task-oriented diversity on teamperformance.

Team interdependence. Teams display varyinglevels of interdependence based on their tasks,goals, and outcomes. Task interdependence is de-fined as the extent to which team members rely oneach other to complete their task (Shea & Guzzo,1987). In highly task interdependent teams, teammembers engage in both sequential and reciprocalexchanges to accomplish the team tasks; in less taskinterdependent teams, team members’ independentcontributions are aggregated to accomplish theteam tasks (Saavedra, Earley, & Van Dyne, 1993;Thompson, 1967; Van de Ven, Delbecq, & Koenig,1976). Teams may also vary on goal and outcomeinterdependence. Goal interdependence refers tothe extent to which a team as a whole has a collec-tive goal. Outcome interdependence refers to theextent to which team members are interdependentin terms of rewards and feedback. These aspects ofinterdependence tend to be highly correlated andtherefore researchers have suggested combiningthem into an overall team interdependence con-struct (Gully, Incalcaterra, Joshi, & Beaubien, 2002).

Social categorization theory would predict thathigher outcome and goal interdependence is likelyto unite team members to work toward a commongoal and motivate them to cast aside differences(Gaertner & Dovidio, 2000). Task interdependencemay facilitate intergroup contact conducive to re-ducing categorization-based processes in teams(Pettigrew, 1998). Team interdependence has re-ceived considerable attention as a moderating in-fluence in team diversity research. For example,Schippers and colleagues (2003) found that out-come interdependence reinforces common groupgoals that can counteract the negative effects ofdiversity, so that highly outcome-interdependentteams with high levels of diversity display moretask-related discussions and communication thanhighly diverse teams with low outcome interdepen-dence. Jehn and colleagues also reported that de-mographic diversity was positively associated withsatisfaction and commitment when task interde-pendence was high (Jehn et al., 1999). Other re-search has also supported this pattern of findings(e.g., Van der Vegt & Janssen, 2003). We wouldexpect that high interdependence creates a contextfor elaboration-based processes within a team andthus that the outcomes of task- and relations-oriented diversity may be more positive or lessnegative when the level of interdependence is high.Hence, we propose:

Hypothesis 3. Team interdependence moder-ates the relationship between task- and rela-

tions-oriented diversity and team performance.The positive effect of task-oriented diversity onteam performance is stronger among highlyinterdependent teams. The negative effect ofrelations-oriented diversity on team perfor-mance is weaker among highly interdependentteams.

Team type. The durability of a team’s member-ship—that is, whether the team has been assembledto accomplish a short-term goal or whether it isinstead a stable and permanent unit in an organi-zation—is likely to have consequences for interper-sonal interactions among diverse team members.Task-related contingencies are likely to differ inshort- and long-term teams. In short-term teams,greater urgency may surround goals and missions.On the other hand, in long-term teams, task require-ments may be more stable, and distribution of tasksand roles may also be more clearly defined (DeDreu & Weingart, 2003). Another dimension alongwhich these teams are likely to differ is the longev-ity of team membership. The members of short-term teams likely have shorter tenure, than themembers of long-term teams.3 Thus, temporal dy-namics are also likely to vary in these two types ofteams.

Some studies that have examined temporal influ-ences on team diversity outcomes have shown thatthe length of time team members spend togethermay diminish the salience of visible aspects of di-versity and enhance the salience of attitudinal orvalue-based aspects of diversity (Harrison et al.,1998; Harrison, Price, Gavin, & Florey, 2002).Schippers and colleagues (2003) found that over alonger duration, highly diverse teams were lesslikely to display elaboration-based processes. In theshort run, however, diverse teams engaged in moretask-relevant debates and discussions that had pos-itive consequences for team performance. The au-thors noted that in long-standing diverse teams,team members may attribute conflicts to relationaldifferences, and the motivation and willingness toresolve differences through greater communicationmay erode over time. In the short term, on the otherhand, the members of highly diverse teams aremore likely to communicate across differences toaccomplish the teams’ tasks (Schippers et al.,2003). Corroborating this finding, Watson, Johnson,and Merrit (1998) also found that demographic di-

3 Although team tenure or longevity could also be im-portant variables to consider, our review indicated thatthese variables were sparsely reported in extant researchand we were, therefore, unable to include these in thepresent study.

610 JuneAcademy of Management Journal

versity was negatively associated with outcomesover time. We propose, on the basis of this empir-ical research, that in short-term teams, urgency sur-rounding the accomplishment of the tasks may re-quire team members to overlook differences andaim at utilizing diversity based on task-relevantdimensions. Since short-term teams are also asso-ciated with shorter team tenure, team membersmay be less likely to attribute task-based differ-ences to deeper attitudinal or personality-based dif-ferences (Harrison et al., 1998; Schippers et al.,2003). In long-term teams, divisions based on di-versity attributes may become more entrenchedand self-reinforcing, so that conflicts and differ-ences based on relations-oriented attributes have amore debilitating impact on team performance(Schippers et al., 2003; Watson et al., 1998). In viewof these considerations, we meta-analytically testedthe following:

Hypothesis 4. Team type moderates the rela-tionship between task- and relations-orienteddiversity and team performance. The positiveeffect of task-oriented diversity is stronger inshort-term teams than in long-term teams. Thenegative effect of relations-oriented diversity isstronger in long-term teams than in short-termteams.

METHODS

Literature Search

We employed multiple search techniques toidentify prior empirical research that examined therelationship between work team diversity and per-formance. First, we searched the computerized da-tabases PsycINFO, ABI/Inform, and SocIndex usingkeywords such as “team/group diversity,” “team/group composition,” “team/group performance,”and “team/group effectiveness” as well as searchterms associated with specific team diversity at-tributes (e.g., “gender,” “race/ethnicity,” “age,”“tenure,” “education,” and “functional backgrounddiversity”). Second, the electronic search was sup-plemented by a manual search of 19 major journals,including Academy of Management Journal, Ad-ministrative Science Quarterly, Journal of AppliedPsychology, Personnel Psychology, OrganizationScience, and Journal of Organizational Behavior,and others considered the most highly cited jour-nals in the field of management (see Gomez-Mejia &Balkin, 1992). Third, we also consulted the refer-ence lists from previous reviews on this topic, in-cluding both narrative (Harrison & Klein, 2007;Jackson et al., 2003; Milliken & Martins, 1996; VanKnippenberg & Schippers, 2007; Williams &

O’Reilly, 1998) and quantitative (Bowers et al.,2000; Horwitz & Horwitz, 2007; Webber &Donahue, 2001) reviews. Finally, in an effort toidentify relevant unpublished studies, we searchedProQuest Digital Dissertations and conference pro-ceedings for the annual meetings of the Academy ofManagement and the Society of Industrial and Or-ganizational Psychology for the previous five years.Researchers in related areas were also contacted toobtain current and unpublished studies that mightfit our criteria for inclusion.

Since this study concerned detecting the moder-ating effects of several contextual variables embed-ded at multiple levels, we focused only on studiesthat had been conducted in field settings wherethese variables were likely to influence diversityoutcomes. We did not include studies that relied onstudent samples, were conducted in laboratory set-tings, or involved simulated tasks or tasks in artifi-cial environments. Although these types of studiesare of great value in theory development, for thepurposes of this research we were interested inidentifying contextual moderators influencing di-versity outcomes in naturally occurring, intactwork teams in business settings. Research suggeststhat these types of teams differ substantively fromlaboratory teams (e.g., McGrath, 1984). From aninitial set of 95 field studies conducted in organi-zational settings, we applied the following addi-tional criteria to select articles for our meta-analy-sis. First, although diversity research has oftenbeen conducted at multiple levels of analysis, werestricted the present meta-analysis to the teamlevel. For a study to be included in the meta-anal-ysis, both diversity and performance had to be mea-sured at the team level. Second, team diversityvariables had to include either relations-orienteddiversity attributes (race/ethnicity, gender, or age)or task-oriented diversity attributes (education,functional background, or organizational tenure)pertinent to our theoretical arguments. Third, wealso excluded studies that examined top manage-ment teams (n � 47) because outcomes for suchteams were often measured at the firm level (e.g., asorganizational financial performance), and topmanagement teams were generally considered asoperating under different dynamics than lower-level general work teams in organizations (e.g.,Hambrick & Mason, 1984; Webber & Donahue,2001). Furthermore, team effectiveness models alsosuggest that performance is a proximal outcome ofteam composition among general work teams; thisassumption may not hold among top managementteams (Hackman, 1987). Finally, a study had toreport sample sizes and an appropriate statistic(e.g., mean and standard deviation, chi-square, t, F)

2009 611Joshi and Roh

that allowed the computation of a correlation coef-ficient with the formula provided by Hunter andSchmidt (1990: 272). Based on these criteria, thefinal data set included 8,757 teams from 39 studiesconducted between 1992 and 2009, yielding a totalof 117 effect sizes. The data included in the finalanalyses represent approximately twice the num-ber of effect sizes and three times the number ofteams included in past published meta-analyses onthis topic (e.g., Webber & Donahue, 2001).

Coding of Studies

Our initial coding scheme was based on opera-tionalization of team characteristics (team interde-pendence and team type) and research setting(occupational and industry setting). Using thisscheme, each author and two additional raters in-dependently coded a random selection of five arti-cles; initial interrater agreement ranged between 75and 94 percent. To resolve disagreements, we wentback to the studies and reached consensus throughdiscussion. From this discussion, we developed adetailed set of decision rules and used them to codean additional seven articles. Interrater agreementwas almost 100 percent for these articles. The au-thors then coded the remaining articles using thedecision rules we had developed.

Measures

Team diversity. Drawing on past research (e.g.,Jackson et al., 1995; Webber & Donahue, 2001), weclassified diversity attributes into two categories:relations-oriented (race/ethnicity, gender, and age)and task-oriented (organizational tenure, educa-tion, and functional background). In the studiesincluded in the analyses, categorical diversity at-tributes were measured using Blau’s (1977) indexor Teachman’s (1980) entropy measure; Allison’s(1978) coefficient of variation was used for contin-uous variables such as age and tenure. Becausemeasures based on “faultlines” (hypothetical linessplitting a group into attribute-based subgroups[Lau & Murnighan, 1998: 328; 2005: 645]) wereused less frequently and in some cases combineddemographic and task-oriented attributes, thesewere not included in the present analysis. In gen-eral, the operationalization of relations- and task-oriented diversity in this study is theoreticallydriven and consistent with past research. Our ini-tial review also indicated that the diversity at-tributes included in this meta-analysis were themost commonly studied variables in past research.However, since the measurement of diversity at-tributes in studies involved in the sample predates

recent reconceptualizations (e.g., Harrison & Klein,2007), we were unable to incorporate these newerdirections in this study.

Team performance. Measures of team perfor-mance included financial and operational mea-sures (e.g., sales, productivity), product quality/quantity, team innovation, supervisor ratings ofteam performance/effectiveness, and self-ratingsby team members. Most studies provided objectiveteam performance measures or ratings by supervi-sors. For subjective performance measures, reli-abilities of measurement instruments were re-corded whenever reported. In those cases in whichno reliabilities were reported, we took the averagereliability of the same variable from all other stud-ies. The average reliabilities were .82 for subjectiveperformance measures. To maintain the statisticalindependence of the data set, when multiple mea-sures for team performance were available, we in-cluded only the most objective external measure;when no other source was available, we relied onteam members’ own assessments of their perfor-mance. In a study measuring multiple dimensionsof team performance (e.g., its quality and quantity),we calculated a composite effect by averaging eachcorrelation coefficient.

Occupational demography. We relied on the ar-chival data from the Labor Force Statistics of theCurrent Population Survey (U.S. Bureau of LaborStatistics, 2006) to obtain definitions and statisticsof occupational categories. Initially we coded theresearch settings (based on sample descriptions) ofstudies included in the meta-analysis using Bureauof Labor Statistics (BLS) definitions; then we ob-tained the occupational gender, race/ethnicity, andage composition data from the BLS data set andassigned this demographic composition to the oc-cupations included in the studies. For example, if astudy’s respondents were described as members ofproduction teams from multiple electronics manu-facturing firms, we obtained data on electrical,electronics, and electro-mechanical assemblers’gender, ethnicity, and age from the BLS websiteand used them to measure occupational demogra-phy (n � 16). For teams that included multipleoccupations (e.g., cross-functional product devel-opment teams consisting of engineers and produc-tion supervisors), we found demographic data foreach occupational category involved in a team andthen calculated a composite value by averaging thedemographic information on all participating occu-pations from the BLS data set (n � 8). In aggregatingdata over multiple occupational categories, wewere careful not to include studies that incorpo-rated occupations that were differently skewed de-mographically. For example, when a research sam-

612 JuneAcademy of Management Journal

ple involved multiple occupations in singleteams—combining, say, the white-male-domi-nated occupation of network analyst and the rel-atively balanced occupation of insurance under-writer or HR professional (Campion, Papper, &Medsker, 1996)—we performed no aggregationand excluded the study from this set of analyses.Moreover, when studies were unclear about thesamples used, had non-U.S. samples for whichdata were not available, or had samples withmultiple types of teams, we also excluded themfrom the analyses.

Among the studies included in these analyses,the occupational percent female ranged from 20.6to 73.4 percent, and the occupational percent eth-nic minority (i.e., nonwhite) ranged from 15.7 to37.6 percent. If either of these percentages wasbelow the overall mean for all U.S. occupations(i.e., 46.3 percent for female composition and 29percent for minority composition), the occupa-tional setting was characterized as “majority male”or “majority white” for the purposes of categoricalanalyses. Occupational settings with above-averagepercents female or minority were categorized as“balanced” settings. In our sample, we did not en-counter enough study settings that could be cate-gorized as majority female or majority ethnic mi-nority; therefore, we were unable to test the effectsof diversity in these settings. Occupational agecomposition data were based on the informationprovided by the BLS, which offers data on four agecategories by occupation. We considered an occu-pation as “majority younger worker” if the propor-tion of older (over 55 years) workers in that occu-pation was lower than the average acrossoccupations (i.e., 18%) and as “balanced” if theproportion of older workers was above the averageacross occupations.

Industry setting. Industry was also coded when-ever it was reported and whenever the researchsetting or sample was described. Following strate-gic management research (e.g., Collins & Smith,2006; Datta et al., 2005) and drawing from the de-tailed industry descriptions based on the NorthAmerican Industry Classification System (NAICS;U. S. Census Bureau, 2002), we classified indus-tries examined in previous research into threebroad categories: service, manufacturing, andhigh technology. Service industries in the sampleincluded wholesale/retail trade, finance/insur-ance, health care, educational service, moving/transportation, and government service; manu-facturing industries included automobilemanufacturing, paper and wood manufacturing,textile manufacturing, oil and gas, chemicalproduct manufacturing, and general equipment

manufacturing. High technology included semi-conductor/electronics, information processing,telecommunication, and professional R&D ser-vices. Studies that included samples from multi-ple industry settings were excluded from ouranalyses (n � 5).

Team interdependence. Drawing on conceptualdefinitions of interdependence presented in previ-ous research (e.g., Campion, Medsker, & Higgs,1993; Saavedra et al., 1993; Shea & Guzzo, 1987),we determined team interdependence using threeseparate ratings for task, goal, and outcome inter-dependence. Each dimension was rated on a scaleranging from 1 to 3 (low, moderate, and high) andaveraged to provide a composite score of overallteam interdependence. Low-interdependence teams in-cluded, for example, production or sales teamswith sequentially related activities or individual-based goals and rewards; cross-functional R&Dteams whose members frequently exchanged ideasand shared common goals were deemed highlyinterdependent.

Team type. We also coded team type based onthe length of time a team was expected to exist(Schippers et al., 2003). Two broad categories,short-term and long-term, were used. Cross-func-tional project teams existing for a limited periodwere considered short-term teams, for example.Permanent work teams and general work teams ex-isting for longer than two years were generally con-sidered long-term.

Meta-Analytic Techniques

We used Hedge and Olkin’s (1985) meta-analyticprocedures to analyze the data. Zero-order correla-tions between work team diversity and perfor-mance were taken or calculated from each studyand corrected for measurement error. FollowingHunter and Schmidt’s (1990) formula, we also cor-rected the correlations for unreliability using theartifact distributions for subjective team perfor-mance measures. We calculated weighted meancorrelations by adopting the inverse varianceweights and applying Fisher’s Z transformationprocedures (Hedge & Olkin, 1985; Lipsey & Wilson,2001). Ninety-five percent confidence intervalswere calculated around the sample-weighted corre-lation as a measure of accuracy of the effect size(Whitener, 1990). The failsafe k was also calculatedto identify the number of “file drawer” (unknown)studies of the same relationship with a true corre-lation of zero needed to widen the reported confi-dence interval enough to include zero (Orwin,1983; Rosenthal, 1979).

2009 613Joshi and Roh

Heterogeneity of effect sizes. To determinewhether effect sizes were consistent over the stud-ies reviewed, we tested the homogeneous distri-bution of the effect sizes by calculating the Q-statistic (Hedge & Olkin, 1985). A significant Qindicates the likelihood of moderators that ex-plain variability in correlations over studies (Lip-sey & Wilson, 2001). We also examined the po-tential impact of outliers by calculating thesample-adjusted meta-analytic deviancy statisticsuggested by Huffcut and Arthur (1995). Threeoutliers were identified by use of this statistic,but a detailed review of the potential outliersrevealed no problematic correlations, so thesewere not eliminated from the analyses.

Moderator analysis. We conducted detailedmoderator analyses to determine whether multi-level contextual variables were related to the het-erogeneity of effect sizes (Hedge & Olkin, 1985;Lipsey & Wilson, 2001). The logic of the categoricalmodel moderator test is analogous to analysis ofvariance (ANOVA). Calculating the categoricalmodels results in (1) the between-group goodness-of-fit statistic QB, which has an approximate chi-square distribution with p � 1 degrees of freedom,where p is the number of groups, and (2) the with-in-group goodness-of-fit statistic Qw, which has anapproximate chi-square distribution with m � 1degrees of freedom, where m is the number of effectsizes in the group. That is, QB is analogous to amain effect in an ANOVA, and Qw indicates homo-geneity within each group in an ANOVA. In thepresent analysis, as recommended in previous re-search on this topic (e.g., Webber & Donahue,2001), we used a QB statistic to test whether thecategorical moderator model was statistically sig-nificant and then examined each subgroup withinthe sample by testing the confidence intervals forstatistical significance and by comparing the effectsizes across subgroups whenever possible. For thecontinuous moderators (i.e., occupational percentfemale and occupational percent minority),weighted least squares (WLS) regression was alsoused, as suggested by Hedge & Olkin (1985). Thisprocedure involves weighting each observed effectsize by the inverse of its variance, as with formulaefor weighted means and confidence intervals (Lip-sey & Wilson, 2001). Using such an approachavoids the artificial categorization of continuousmoderating variables. Two indexes assessing theoverall fit of the weighted regression model canbe calculated: a Q attributable to the regressionand a Q error or residual (denoted as QR and QE,respectively, and both distributed as a chi-square). QR is analogous to an F for a regressionmodel and, if significant, indicates that the re-

gression model explains significant variability inthe correlations of the relationship between teamdiversity variables and team performance (Lipsey& Wilson, 2001).4

RESULTS

Main Effects: Work Team Diversityand Performance

Using the meta-analytic techniques describedabove, we tested the main effects of work teamdiversity on performance outcomes as well as themoderating effects of contextual factors embeddedat multiple levels. Table 2 presents the main effectresults. We first examined the correlations betweenall types of diversity and performance and obtaineda near-zero, nonsignificant result (r � �.01, k �117, 95% CI � �.02 to .00). This initial resultcorroborated past meta-analytic findings (e.g., Web-ber & Donahue, 2001). We then conducted separateanalyses for relations- and task-oriented diversityand found a different pattern of results for eachtype of diversity. For relations-oriented diversity,we found a very weak negative but significant re-lationship with performance (r � �.03, k � 69,95% CI � �.05 to �.02). The relationship betweentask-oriented diversity and performance was alsovery weak but positive and significant (r � .04, k �48, 95% CI � .02 to .06). The failsafe k’s in Table 2suggest that, although the effect sizes are verysmall, at least 118 (for relations-oriented diversity)and 73 (for task-oriented diversity) file-drawer nulleffects would need to be reported before those tworelationships would lose statistical support. Wealso conducted additional analyses for each diver-sity attribute and found that functional backgrounddiversity was most positively related to perfor-mance (r � .13, k � 20, 95% CI � .09 to .17) andthat age diversity showed the most negative perfor-

4 We undertook additional OLS regressions whereby weregressed the effect of diversity-performance on the keyindependent variables. Results indicated that the pattern offindings for OLS regressions mirrors the pattern of findingsreported here. We also undertook a stepwise regressionprocedure and found that adding each contextual variablewas associated with significant incremental variance in thediversity-performance effect. Collectively, in these analy-ses, contextual variables accounted for 56 percent of thevariance in relations-oriented effect sizes across studies.Further, the variance inflation factors in all cases were wellbelow 10, which is the rule-of-thumb cutoff for high mul-ticollinearity. The full results are available from the authorsupon request. We thank our action editor, Jason Colquitt,for this suggestion.

614 JuneAcademy of Management Journal

mance effect (r � �.06, k � 21, 95% CI � �.09 to�.04).

Table 2 also shows that considerable heterogene-ity among effect sizes exists, as indicated by theQ-statistic. Both Q values for relations- and task-oriented diversity are highly significant (p’s � .01),indicating that correlations vary across studies andthat potential moderators might exist that can ex-plain these correlations.

Occupational Demography

Hypotheses 1a and 1b predicted that the negativeeffects of gender and race/ethnicity diversity wouldbe weaker in more gender-balanced and ethnicallybalanced settings, respectively. We conducted cat-egorical moderator analyses and contrasted the dif-ference between majority male or majority whiteoccupations and relatively gender-balanced andethnically balanced occupations. Table 3 summa-rizes the results. The categorical model testing themoderating effect of occupational percent femalewas highly significant (QB[1] � 39.19, p � .01). Ashypothesized, gender diversity had a significant,negative effect on team performance in majoritymale occupational settings (r � �.09, k � 12, 95%CI � �.12 to �.05). The effect of gender diversitywas significantly positive in relatively gender bal-anced settings (r � .11, k � 7, 95% CI � .06 to .15).The moderating effect of occupational percentminority was also significant (QB[1] � 48.65, p �.01). The average correlation was significantlynegative in majority white occupations (r � �.07,k � 10, 95% CI � �.10 to �.04) and positive inrelatively balanced occupations (r � .11, k � 6,95% CI � .07 to .14). In addition, because theoccupational gender and race/ethnicity demogra-

phy variables were measured as continuous vari-ables (i.e., occupational percent female and occu-pational percent minority), to further understandthe patterns of these moderating effects, we alsoperformed WLS regressions. In support of bothhypotheses, we found that occupational percentfemale and minority accounted for significantamounts of variance in the correlations betweengender and race/ethnicity diversity and team per-formance (� � .32, p � .01, R2 � .10, for genderdiversity; � � .37, p � .01, R2 � .14, for race/ethnicity diversity). Significant values of QR forgender diversity (QR[1] � 16.78, p � .01) andrace/ethnicity diversity (QR[1] � 22.71, p � .01)also indicate that both regression models werestatistically significant.

Hypothesis 1c predicted that the negative effectsof age diversity would be strengthened in occupa-tions composed of relatively younger workers andweakened in relatively age-balanced occupations.We conducted a categorical analysis and did notfind a strong moderating relationship (QB[1] �0.85, p � .10) (see Table 3). Although occupationscomposed of younger workers displayed slightlymore negative effects than occupations that weremore balanced in terms of age, considering thenonsignificant QB and significantly overlappingconfidence intervals, the two occupational groupswere not statistically different.

Hypothesis 1d proposed that the positive perfor-mance outcome of task-oriented diversity wouldbe stronger in more balanced occupational set-tings. Contrary to this proposition, task-orienteddiversity showed more positive performance ef-fects in majority male and white settings. How-ever, considering the nonsignificant subgroup re-sults for balanced occupational settings across all

TABLE 2Main Effects: The Relationship between Team Diversity and Performancea

Diversity TypeEffect Sizes

(k)Total Teams

(N)WeightedMean r

95%Confidence

IntervalFailsafe

k Q

All diversity 117 29,608 �.01 �.02, .00 635.16**Relations-oriented diversity 69 19,779 �.03 �.05, �.02 118 479.54**

Gender 26 5,473 �.02 �.04, .01Race/ethnicity 22 7,089 �.01 �.04, .01Age 21 7,217 �.06 �.09, �.04 48

Task-oriented diversity 48 9,829 .04 .02, .06 73 155.63**Function 20 3,085 .13 .09, .17 65Education 9 2,863 �.02 �.06, .01Tenure 19 3,881 .03 �.01, .06

a N is the total number of teams counted by effect sizes; failsafe k indicates the number of unpublished studies reporting null resultsneeded to reduce the cumulative effect across studies to the point of nonsignificance (p � .05) and is only reported for statisticallysignificant results (p � .05); Q is the effect size heterogeneity statistic indicating the possibility of moderators.

** p � .01

2009 615Joshi and Roh

diversity attributes and the overlapping confi-dence intervals in general, we were unable inter-pret this finding meaningfully.

Industry Setting

Hypotheses 2a and 2b proposed that industrysetting moderated the relationship between teamdiversity and performance outcomes. Table 4 pre-sents the results. With regard to relations-orienteddiversity, the overall categorical model testing in-dustry impact was highly significant (QB[2] �209.89, p � .01). As predicted in Hypothesis 2a,relations-oriented diversity had a positive effect onperformance in service industries (r � .07, k � 21,95% CI � .05 to .09). Inconsistently with Hypoth-esis 2a, however, in the manufacturing industrysetting, the effect of relations-oriented diversitywas negative (r � �.04, k � 16, 95% CI � �.07 to

�.01) and interestingly, relations-oriented diver-sity displayed the strongest negative performanceeffect in high-technology industry settings (r ��.18, k � 21, 95% CI � �.20 to �.15). This findingis also resistant to unpublished null effects with afailsafe k of 152.

Regarding Hypothesis 2b, we found weak sup-port for the moderating effects of industry settingon the performance outcome of task-oriented diver-sity. The overall categorical model was modestlysignificant (QB[2] � 7.28, p � .05). Task-orienteddiversity was positively related to performance inthe high-technology industry setting, as hypothe-sized (r � .06, k � 23, 95% CI � .04 to .09), but thiseffect was only slightly larger than the overall maineffect of task-oriented diversity (r � .04). We didnot find any significant support for industry mod-erating effects on task-oriented diversity effects inthe manufacturing and service settings.

TABLE 3Contextual Influences: Occupational Demographya, b

Team Diversity � OccupationalDemographyc

Effect Sizes(k)

Total Teams(N)

WeightedMean r

95%Confidence

IntervalFailsafe

k QB

Gender diversity (Hypothesis 1a) 39.19**Majority male settings 12 2,952 �.09 �.12, �.05 20Balanced settings 7 1,832 .11 .06, .15 12

Race/ethnicity diversity (Hypothesis 1b) 48.65**Majority white settings 10 4,071 �.07 �.10, �.04 17Balanced settings 6 2,584 .11 .07, .14 13

Age diversity (Hypothesis 1c) 0.85Majority younger worker settings 6 4,758 �.08 �.10, �.05 13Balanced settings 9 1,869 �.05 �.10, �.00 3

Task-oriented diversity (Hypothesis 1d)Occupational gender demography 9.67**

Majority male settings 22 6,866 .06 .03, .09 39Balanced settings 14 1,823 �.03 �.08, .02

Occupational race/ethnicity demography 0.60Majority white settings 32 8,497 .04 .02, .06 39Balanced settings 4 192 �.02 �.17, .13

Occupational age demography 5.36*Majority younger worker settings 9 5,199 .06 .03, .09 13Balanced settings 27 3,490 .01 �.03, .04

a N is the total number of teams counted by effect sizes; failsafe k indicates the number of unpublished studies reporting null resultsneeded to reduce the cumulative effect across studies to the point of nonsignificance (p � .05) and is only reported for statisticallysignificant results (p � .05); QB is the between-group heterogeneity statistic indicating the statistical significance of the categoricalmoderator model.

b For continuous occupational demography variables (i.e., occupational percent female and occupational percent minority), we alsoconducted the WLS regression analyses and obtained the same pattern of findings as reported in this table of results (in relation toHypothesis 1a and 1b).

c Occupational settings in which the percent female or minority was below the overall mean level (i.e., 46.3 percent for femalecomposition and 29 percent for minority composition) were considered as majority male or majority whites, respectively, in the analyses;we considered an occupation as a majority-younger-workers setting if the proportion of older workers (over 55 years old) in that occupationwas less than the overall average (i.e., 18%).

* p � .05** p � .01

616 JuneAcademy of Management Journal

Team Interdependence and Team Type

Table 5 presents the results regarding the effectsof team-level moderators on the relationship be-tween team diversity and performance. Hypothesis3 proposed that effects of task-oriented diversitywill be stronger and the effects of relations-orienteddiversity will be weaker among highly interdepen-dent teams. The findings with regard to relations-oriented diversity were contrary to this hypothesis.Although the categorical model for relations-ori-ented diversity was significant (QB[2] � 174.21,p � .01), among teams with low interdependencerelations-oriented diversity was positively relatedto performance (r � .08), and among teams withmoderate and high interdependence, relations-ori-ented diversity was negatively related to perfor-mance (r � �.12 and r � �.04, respectively). Thecategorical analysis for task-oriented diversity sup-ported the hypothesized pattern, although the over-all moderating model only showed limited statisti-cal support (QB[2] � 7.14, p � .05). The positiveperformance effect of task-oriented diversity in-creased as team tasks, goals, and outcomes becamemore interdependent; however, these results alsoneed to be interpreted cautiously because the 95percent confidence interval of the low-interdepen-dence subgroup included zero and overlapped withthose for the other two subgroups.

Our final hypothesis addressed whether teamtype—the length of time a team was expected toexist (i.e., long-term versus short-term)—affected

the relationship between team diversity and perfor-mance. Hypothesis 4 proposed that the negativeeffects of relations-oriented diversity would bestrengthened in long-term teams. We found strongsupport for relations-oriented diversity: the cate-gorical moderator model was highly significant(QB[1] � 222.91, p � .01). The performance effect ofrelations-oriented diversity was positive in rela-tively short-term teams (r � .09, k � 23, 95% CI �.07 to .12) but became negative in more stable orlong-term teams (r � �.14; k � 43, 95% CI � �.16to �.12). Both findings are resistant to unpublishednull effects, with failsafe k’s of 89 and 305, respec-tively. For task-oriented diversity, we did not find astatistical support for the hypothesis (QB[1] � 0.63,p � .10). Although the performance effect of task-oriented diversity was more positive in short-termteams (r � .08) than in long-term teams (r � .04),these subgroup results were not statistically differ-ent from each other.

DISCUSSION

This meta-analytic review took stock of past re-search on the diversity-performance relationshipconducted in organizational settings over the past15 years and examined the sensitivity of this rela-tionship to contextual variables at multiple levels.Our findings revealed that when one considers therelationship between all types of diversity (i.e., col-lapsing relations- and task-oriented distinction)

TABLE 4Contextual Influences: Industry Settinga

Team Diversity � Industry SettingbEffect Sizes

(k)Total Teams

(N)WeightedMean r

95%Confidence

IntervalFailsafe

k QB

Relations-oriented diversity (Hypothesis 2a) 209.89**High-technology industry 21 6,068 �.18 �.20, �.15 152Service industry 21 9,139 .07 .05, .09 58Manufacturing industry 16 3,687 �.04 �.07, �.01 8

Task-oriented diversity (Hypothesis 2b) 7.28*High-technology industry 23 6,475 .06 .04, .09 45Service industry 16 1,702 �.00 �.05, .05Manufacturing industry 3 1,194 .01 �.05, .06

a N is the total number of teams counted by effect sizes; failsafe k indicates the number of unpublished studies reporting null resultsneeded to reduce the cumulative effect across studies to the point of nonsignificance (p � .05) and is only reported for statisticallysignificant results (p � .05); QB is the between-group heterogeneity statistic indicating the statistical significance of the categoricalmoderator model.

b Three industry categories were analyzed: (1) high-technology included electronics/semiconductors, information processing, telecom-munication, and professional R&D service; (2) service included retail trade, finance/insurance, health care, education service, moving/transportation, and government service; (3) manufacturing industries included automobile manufacturing, paper and wood manufacturing,textile manufacturing, oil and gas extraction, and general and chemical product manufacturing.

* p � .05** p � .01

2009 617Joshi and Roh

and performance, the direct effect of diversity onperformance is essentially zero. Relations-orienteddiversity attributes such as gender, race/ethnicity,and age had very small, though significantly nega-tive, effects on team performance. Although func-tional diversity had a more substantial positive ef-fect, other forms of task-oriented diversity (i.e.,education and tenure) also had very small effectson team performance. These findings make it ap-pear that diversity does not really matter for teamperformance. However, we contend that these weakdirect relationships may be obscuring the specificconditions under which diversity can have benefi-cial or detrimental effects on performance out-comes. Our study developed and tested a frame-work identifying boundary conditions under whichthe diversity-performance relationship is likely tobe significant. Specifically, we found that after weaccounted for moderating variables at multiple lev-els, diversity effects doubled or tripled in size. Fur-ther, industry and occupational moderators, whichhave received relatively scant attention in past re-search, explained significant variance in relations-oriented diversity effects across studies. Below we

discuss the theoretical, empirical, and practical im-plications of our findings.

The Role of Occupational Demography andIndustry Setting as Context

We found that in occupations dominated by maleor by white employees, gender and ethnic diversityhad more negative effects on performance out-comes. These findings draw attention to the impor-tance of extraorganizational context in shaping di-versity outcomes, and we call for a more detailedand comprehensive acknowledgement of macro-level context in future research. Apart from stereo-types associated with underrepresented groups in aparticular occupational context, implicit status dif-ferences between demographic groups may also bea mechanism by which contextual factors such asoccupational demography filter into team-level in-teractions (see Ridgeway, 1991, 1997). The domi-nance of a particular demographic group within aparticular occupational setting can signal greateraccess to resources and privilege for this group. Theprivilege accrued by a demographic group in a par-

TABLE 5Contextual Influences: Team Interdependence and Team Typea

Team Diversity � Team ContextEffect Sizes

(k)Total Teams

(N)WeightedMean r

95%Confidence

IntervalFailsafe

k QB

Team interdependence (Hypothesis 3)b

Relations-oriented diversity 174.21**Low interdependence 17 8,051 .08 .06, .10 56Moderate interdependence 38 10,770 �.12 �.14, �.10 238High interdependence 14 958 �.04 �.11, .03

Task-oriented diversity 7.14*Low interdependence 5 557 �.03 �.11, .06Moderate interdependence 24 7,604 .04 .02, .06 23High interdependence 19 1,668 .10 .05, .15 23

Team type (Hypothesis 4)c

Relations-oriented diversity 222.91**Short-term teams 23 7,733 .09 .07, .12 89Long-term teams 43 9,730 �.14 �.16, �.12 305

Task-oriented diversity 0.63Short-term teams 13 684 .08 �.01, .16Long-term teams 34 8,373 .04 .02, .07 45

a N is the total number of teams counted by effect sizes; failsafe k indicates the number of unpublished studies reporting null resultsneeded to reduce the cumulative effect across studies to the point of nonsignificance (p � .05) and is only reported for statisticallysignificant results (p � .05); QB is the between-group heterogeneity statistic indicating the statistical significance of the categoricalmoderator model.

b Low-interdependence teams included, for example, production or sales teams with sequentially related activities and/or individual-based goals or rewards; high-interdependence teams included, for example, cross-functional R&D teams that frequently exchanged ideasand shared common goals.

c Short-term teams involved, for example, project-based teams existing for a limited period of time; permanent work teams or generalwork teams existing for longer than two years were generally considered relatively long-term teams.

* p � .05** p � .01

618 JuneAcademy of Management Journal

ticular context may provide team members belong-ing to this group with an “expertise advantage”;these group members may be considered morecompetent, and the contributions of the members oflow-status demographic groups may be devalued inthese settings (Berger, Ridgeway, Fisek, & Norman,1998; Berger, Ridgeway, & Zelditch, 2002). Indeed,a rich body of sociopsychological research hasshown that, in small groups, interpersonal interac-tions tend to replicate larger societal status differ-entials (Berger et al., 1998, 2002) and may accountfor the suboptimal performance outcomes of gen-der-diverse and ethnically diverse teams in major-ity male and white settings.

We propose that a closer examination of the oc-cupational context in which teams are embeddedcan provide researchers with an understanding ofbroader structural inequality and its implicationsfor team-level diversity outcomes. Further, al-though we were not able to test the role of organi-zational practices in these settings, research sug-gests that these status differentials are furtherlegitimized by organizational practices that favorsome demographic groups over others (Ridgeway,1991, 1997). Thus, another direction for futureresearch could be to examine the role that organ-izational practices play in either enhancing ormitigating status-based processes in a specificoccupational context.

Although statistically inconclusive, our findingwas that task-oriented diversity had weakly posi-tive effects in male/white dominated settings.This pattern is contrary to our theoretical argu-ments. Although we caution against overinterpret-ing this finding, we acknowledge that in these morehomogeneous settings, categorization based on de-mographic characteristics may not have been suf-ficient to mitigate the elaborative potential of task-oriented diversity. We call for more research onpossible mechanisms mediating task-oriented di-versity effects in these occupational settings.

Our findings indicated that the industry settingsin which teams were embedded also had interest-ing implications for team-level diversity-based out-comes. We found that relations-oriented diversityhad positive effects in service industry settings andslightly negative effects in manufacturing settings.In addition to the market competence perspective(Richard et al., 2007) discussed in developing ourhypotheses, some additional considerations shouldalso perhaps be taken into account. Service settings(for example, retail establishments and restaurants)involve front-line customer contact, and the costsof interactions based on negative categorizationsare high in this context. Hence, firms embedded inthese industries may engage in proactive diversity

management efforts to address gender-, ethnicity-,and age-based issues in the workplace.5 Specificforms of training intended at changing behaviorstargeted at demographically dissimilar employeesand customers may be implemented in these set-tings. In contrast, in manufacturing settings, wherewe found negative effects, firms do not face similarpressures and may be less likely to directly addressthese aspects of team diversity.

Surprisingly, we found that in high-technologysettings, relations-oriented diversity had a moresubstantial negative effect. Recently, DiTomaso andcolleagues found that, in 24 firms that fit our defi-nition of high-technology settings, white men re-ceived more training, mentoring, and coaching andalso the most favorable performance evaluations,relative to any other demographic group (DiTo-maso, Post, Smith, Farris, & Cordero, 2007). Thesecharacteristics of high-technology firms may en-hance ethnicity- and gender-based status differen-tials. It would appear that white men occupy ahigh-status position (reflected in these favorableemployment outcomes) in the high-technology sec-tor (DiTomaso, Post, Smith et al., 2007) and wouldenjoy an “expertise advantage” relative to womenor minorities. Therefore, the status-based processesthat we discussed earlier could also account for thenegative effects of relations-oriented diversity inthis industry. Furthermore, like manufacturingfirms, high-technology firms may not face pres-sures to proactively adopt diversity managementpractices. These findings call for a finer-grainedunderstanding of the industry context framing di-versity-based outcomes. For example, future re-search could apply contingency-based perspectivesand examine whether a fit between firms’ diversitymanagement strategies and industry-level contin-gencies explains the outcomes of diversity at theteam level. Longitudinal studies that examine howspecific industry trends influence firms’ responsesto diversity and their implications for team perfor-mance could be an avenue for future inquiry.

We conclude this section by noting that our find-ings challenge the assumption, born from social-categorization theory, that some aspects of diver-sity necessarily have detrimental effects on teamperformance. Although relations-oriented diversityattributes have been typically associated with cat-egorization-based processes (Jackson et al., 1995;

5 Consider as an example the negative publicity re-ceived by Denny’s Restaurants as a result of discrimina-tion charges and their subsequent efforts to train andreward appropriate behaviors toward demographicallydissimilar employees and customers.

2009 619Joshi and Roh

Jehn et al., 1999), we found that the effects of theseattributes on team performance ranged from signif-icantly negative (in the high-technology industryand in male/white dominated occupations) to sig-nificantly positive (in the service industry and ingender-/ethnicity-balanced occupations). The ef-fects of task-oriented diversity, however, were in-variant across these contexts. Current applicationsof social identity theory or social categorizationtheory in diversity research are insufficient for ex-plaining these findings. These theoretical frame-works were originally developed to explain the out-comes of differences in minimal group settings andmay not capture the effects of demographic at-tributes among naturally occurring work groups inthese contexts (see also DiTomaso, Post, Smith,Farris, & Cordero, 2007; Linnehan & Konrad, 1999).Therefore, these findings suggest that a priority forfuture research is to investigate how the demogra-phy of a work context and the characteristics of theindustry settings in which teams are embedded caninfluence the performance outcomes of gender orethnic diversity. We call for more context-focuseddiversity research, which would entail targetingparticular settings in which either the negative orthe positive consequences of team diversity aremore likely. Such efforts could facilitate the devel-opment of new theoretical approaches that incor-porate the status-based or strategic managementperspectives we outlined above.

Team-Level Contextual Influences

Our review of past literature suggested that therehas been a growing emphasis on the role of moder-ators in diversity research and that, predominantly,past research has focused on team-level moderatorssuch as team tenure, task interdependence, andtask complexity (see Van Knippenberg & Schippers[2007] for a review). In view of this research, weproposed that contextual factors such as team in-terdependence and team type would moderate theeffects of diversity on performance in such a waythat elaboration-based processes would be morelikely to be observed in more interdependent orshorter-lived teams (Bowers et al., 2000; Jehn et al.,1999). Contrary to our predictions, we found thatrelations-oriented diversity had the most negativeeffects in moderately interdependent teams. Thesefindings suggest that the interactive effects of teamdiversity and interdependence may be more com-plex than acknowledged in the past. Moderatelyinterdependent tasks may impose constraints thatinterfere with the elaborative potential of diversityattributes but are not so demanding that team mem-bers must overcome categorization-based processes

and collaborate with each other. Although the ef-fect sizes were smaller, a lower level of interdepen-dence was associated with positive performanceoutcomes for relations-oriented diversity. In teamswith low interdependence, the distinct nature ofeach individual team member’s task and a lowerneed to exchange information may create a situa-tion in which team members are not frustrated withdissimilar group members and are able to recognizethe contributions of other members to the teamgoals (Pelled et al., 1999).

In support of our propositions, findings indi-cated that in short-term teams (such as temporaryproject teams), relations-oriented diversity had apositive effect on performance. In these teams, theexpectation of a finite amount of time to completetasks may force group members to exploit the fullelaborative potential of diversity attributes. In long-term teams, team members may feel less compelledto identify and utilize diverse perspectives andmay also have more opportunity to bring up con-flicts that may be detrimental to group functioning.We note that we had limited ability to test temporalinfluences on diversity effects more directly in thismeta-analysis. Measures of team tenure or longev-ity may yield a different pattern of findings (Harri-son et al., 2001; Schippers et al., 2003).

Overall, it is interesting to note that these team-level moderators, which have received the mostattention in past research, had weaker moderatingeffects on diversity outcomes than industry andoccupational variables. Moreover, these variableshad a more significant moderating effect on rela-tions-oriented diversity effects than on task-ori-ented diversity effects. In general, we found thatmoderately interdependent teams and long-termteams could face challenges with increasing diver-sity; less interdependent teams and short-termteams would benefit from greater demographic di-versity among team members. These findings sug-gest that more research on the psychological pro-cesses underlying these moderating effects isneeded in these specific team contexts. For exam-ple, understanding whether role stress or role over-load associated with greater team interdependencehas implications for the manner in which teammembers deal with demographic differences wouldbe useful. These team-level moderators may alsoneed to be considered in conjunction with othermacro-level contextual variables. For example, itwould be interesting to understand whether task-related contingencies are more likely to induceconflict in demographically diverse groups whenthe macrolevel context enhances categorization-based processes.

620 JuneAcademy of Management Journal

Caveats, Limitations, and Future Directions

The meta-analytic approach we applied in thisstudy has some limitations. One concern may bethat the sample is small and may be associated withsecond-order sampling error, which is particularlyrelevant for moderator analysis (Hunter & Schmidt,1990). However, the failsafe k’s we report are fairlylarge and suggest that, even at the rate at whichdiversity research is being conducted, at least adecade more of research may be required to over-turn the significant findings reported here.

We note that we were also limited in our abilityto code contextual variables, as the studies’ de-scriptions of research settings and team tasks wereoften scanty. For example, despite the increasingattention to temporal effects on team diversity out-comes, the lack of information regarding team ten-ure or team duration was especially troubling. Wepropose that future studies provide more detaileddescriptions of research settings and acknowledgethe role that these factors play in explaining re-search findings. Because of theoretical consider-ations and data availability, we considered only asubset of the possible moderating influences onteam diversity outcomes. Missing from this analy-sis is the role that organizational context and otherextraorganizational factors can play in shaping di-versity outcomes. Aspects of organizational contextsuch as managerial demography, climate, culture,and leadership also merit closer scrutiny in futureresearch. Other extraorganizational factors, such associetal and political events (e.g., immigrationtrends, passage of key legislation), may also be im-portant to consider in the future but were outsidethe scope of this study. Further, we note that fortesting the effects of occupational demography, wecould only include U.S.-based samples. These find-ings may vary in other cultural contexts.

Finally, we used a simplified relations-orientedversus task-oriented typology while consideringcontextual effects on diversity outcomes; thisdichotomy corresponded to the categorization-versus elaboration-based processes underlying di-versity effects. Other typologies, such as surface-versus deep-level diversity, may also be pertinentto this analysis and should be considered in thefuture. Contextual effects on the outcomes of deep-level diversity variables, such as personality, cog-nitive ability, values, and attitudes, were also out-side the scope of the present study and may be ofinterest in future research. We were also unable toincorporate the faultlines approach that has been agrowing focus in diversity research. Of the studieswe initially reviewed, 5 percent applied a fault-lines-based operationalization of diversity. We also

note that, although conceptually interesting, fieldresearch directly applying faultlines-based mea-sures is not abundant and has yielded inconsistentresults (e.g., Gibson & Vermeulen, 2003; Greer &Jehn, 2008; Lau & Murnighan, 2005; Li & Hambrick,2005). We propose that future research also exam-ine how contextual moderators may enhance thestrength and number of faultlines in teams to re-solve these mixed findings.

Harrison and Klein (2007) offered several guide-lines for the measurement of diversity in terms ofthree dimensions: separation, variety, and dispar-ity. These authors proposed that the measurementof various diversity attributes should be based onthe specific dimension of diversity considered. Inthis meta-analytic review, because we relied onmeasurement of diversity that predated this 2007article, we were unable to adopt these guidelines.However, complementing Harrison and Klein’s(2007) arguments, we note that the malleability ofrelations-oriented diversity to occupational, indus-try, and team-level contextual variables under-scores the importance of clarifying the specificdimension associated with these demographic at-tributes in the future. A comprehensive contextualanalysis may also help researchers determinewhether a specific demographic attribute is likelyto manifest as “separation,” “variety,” or “dispar-ity” in teams. For example, diversity could be con-ceptualized as variety in short-term teams, in ser-vice settings, and in demographically balancedsettings but as disparity in long-term teams, inmale/white dominated occupations, and in high-technology settings.

Implications for Diversity Management

A context-based approach to workplace diversityresearch can potentially provide practical insightsthat might enhance the effectiveness of diversitymanagement practices. In some work contexts, it isimportant to recognize that diversity at the workgroup level may be problematic. In these settings,diversity interventions aimed directly at targetingbehaviors driven by stereotyping and bias againstunderrepresented groups and ensuring representa-tion at higher levels may be necessary to reversenegative diversity-based outcomes. Furthermore, or-ganizations can directly address the dominance of ademographic majority in a particular labor marketby proactively implementing partnerships with ed-ucational institutions to increase gender- and race-based diversity in the applicant pool. These obser-vations may be particularly important in the case ofthe high-technology settings included in our anal-ysis. At a time when this sector of the economy is

2009 621Joshi and Roh

growing rapidly, there is also growing concernamong policy makers regarding the declining ratioof women and minorities in this industry (Informa-tion Technology Association of America, 2003).The negative performance effects of diversity thatwe found in high technology may also reflect an“unfriendly” atmosphere for women and minori-ties in this setting. Partnerships with high schoolsand college campuses to increase gender and ethnicdiversity in mathematics, science, and engineeringcould enhance the diversity of the applicant pooland provide a more balanced representation of var-ious demographic groups in firms, thereby mitigat-ing the problematic consequences of demographicdiversity at the work group level. We note that afew prominent firms in this industry are alreadymaking these proactive efforts, and our findingsunderscore the value of these initiatives.

With respect to our findings regarding the effectsof team-level moderators, we propose that diversitymanagement practices need to specifically addressthe level of interdependence or longevity that teammembers will encounter. Teams performing moreinterdependent tasks over the long term may needongoing training interventions or team coachingthat facilitates group decision making and con-flict resolution. A more tailored team-focusedapproach to designing diversity managementpractices may enhance the relevance and effec-tiveness of these practices in teams. In general, a“contextual diagnosis” will allow firms to de-velop diversity management practices that aretailored to reduce categorization-based processesand enhance elaboration-based processes at theteam level. Overall, long-term teams that performinterdependent tasks in less demographicallybalanced occupational settings may be the mostvulnerable to categorization-based processes andmay need more direct and focused interventionsthan other types of teams.

Conclusion

Past reviews have labeled differing theoreticalperspectives and contradictory findings in diver-sity research as the “double-edged sword of diver-sity” (e.g., Webber & Donahue, 2001). In this article,we propose a new agenda for diversity research—one that moves beyond a debate regarding the po-tential benefits or costs of diversity and highlightsthe inherent context dependence of diversity ef-fects in organizations. We note that inadequatelyreporting and acknowledging context not only ob-scures the important consequences of diversity inorganizations, but also hampers efforts to synthe-size and integrate the cumulative evidence from the

past; handicaps future theory building; and limitsresearchers’ ability to distill the practical implica-tions of findings. This study presents a roadmap forcontext-focused research that we hope encouragesresearchers to more carefully account for contextin future studies, facilitates greater theoreticalintegration of the macro and micro levels of anal-ysis, and paves the way for new theoretical andmethodological developments.

REFERENCES

*Ainoya, N. 2004. Demographic diversity, team pro-cess, and team performance: Assessing moderatoreffects of cognitive conflict management practicesand task interdependence. Unpublished doctoraldissertation, University of Southern California.

Allison, P. D. 1978. Measures of inequality. AmericanSociological Review, 43: 865–880.

Allport, G. 1954. The nature of prejudice. Cambridge,MA: Addison-Wesley.

*Ancona, D. G., & Caldwell, D. F. 1992. Demography anddesign: Predictors of new product team performance.Organization Science, 3: 321–341.

*Balkundi, P., Kilduff, M., Barsness, Z. L., & Michael,J. D. 2007. Demographic antecedents and perfor-mance consequences of structural holes in workteams. Journal of Organizational Behavior, 28:241–260.

Bamberger, P. 2008. Beyond contextualization: Usingcontext theories to narrow the micro-macro gap inmanagement research. Academy of ManagementJournal, 51: 839–846.

Batt, R. 2002. Managing customer service: Human re-source practices, quit rates, and sales growth. Acad-emy of Management Journal, 45: 587–597.

*Baugh, S. G., & Graen, G. B. 1997. Effects of team genderand racial composition on perceptions of team per-formance in cross-functional teams. Group and Or-ganization Management, 22: 366–383.

Berger, J., Fisek, M. H., Norman, R. Z., & Zelditch, M.1977. Status characteristics in social interactions:An expectations states approach. New York:Elsevier.

Berger, J., Ridgeway, C., Fisek, M. H., & Norman, R. Z.1998. The legitimation and delegitimation of powerand prestige orders. American Sociological Review,63: 379–405.

Berger, J., Ridgeway, C., & Zelditch, M. 2002. The con-struction of status and referential structures. Socio-logical Theory, 20: 157–179.

Blau, P. M. 1977. Inequality and heterogeneity: A prim-itive theory of social structure. New York: FreePress.

Bourgeois, L. J. 1985. Strategic goals, perceived uncer-

622 JuneAcademy of Management Journal

tainty, and economic performance in volatile envi-ronments. Academy of Management Journal, 28:548–573.

Bowers, C. A., Pharmer, J. A., & Salas, E. 2000. Whenmember homogeneity is needed in work teams: Ameta-analysis. Small Group Research, 31: 305–327.

Brewer, M. B. 1988. A dual process model of impressionformation. In T. K. Srull & R. S. Wyer (Eds.), Ad-vances in social cognition, vol. 1: 1–36. Hillsdale,NJ: Erlbaum.

*Bunderson, J. S. 2003. Recognizing and utilizing exper-tise in work groups: A status characteristics perspec-tive. Administrative Science Quarterly, 48: 557–591.

Burns, T., & Stalker, G. M. 1961. The management ofinnovation. London: Tavistock.

*Cady, S. H., & Valentine, J. 1999. Team innovation andperceptions of consideration: What difference doesdiversity make? Small Group Research, 30: 730–750.

*Campion, M. A., Medsker, G. J., & Higgs, A. C. 1993.Relations between work group characteristics andeffectiveness: Implications for designing effectivework groups. Personnel Psychology, 46: 823–850.

*Campion, M. A., Papper, E. M., & Medsker, G. J. 1996.Relations between work team characteristics and ef-fectiveness: A replication and extension. PersonnelPsychology, 49: 429–452.

Cappelli, P., & Sherer, P. 1991. The missing role of con-text in OB: The need for a meso-level approach. InL. L. Cummings & B. M. Staw (Eds.), Research inorganizational behavior, vol.13: 55–110. Green-wich, CT: JAI Press.

Carpenter, M. A., Geletkanycz, M. A., & Sanders, W. G.2004. Upper echelons research revisited: Anteced-ents, elements, and consequences of top manage-ment team composition. Journal of Management,30: 749–778.

Chatman, J. A., & Flynn, F. 2001. The influence of demo-graphic heterogeneity on the emergence and conse-quences of cooperative norms in work teams. Acad-emy of Management Journal, 44: 956–974.

*Choi, J. N. 2007. Group composition and employeecreative behavior in a Korean electronics company:Distinct effects of relational demography and groupdiversity. Journal of Occupational and Organiza-tional Psychology, 80: 213–234.

Choi, J. N., Price, R. H., & Vinokur, A. D. 2003.Self-efficacy changes in groups: Effects of diversity,leadership, and group climate. Journal of Organiza-tional Behavior, 24: 357–372.

Cleveland, J. N., Festa, R. M., & Montgomery, L. 1988.Applicant pool composition and job perceptions:Impact on decisions regarding an older applicant.Journal of Vocational Behavior, 32: 112–125.

Cleveland, J. N., Montgomery, L., & Festa, R. M. 1984.

Group composition and job perceptions: Impact ondecision about sustainability of applicants forgroup membership. Unpublished manuscript, Ba-ruch College, City University of New York.

Collins, C. J., & Smith, K. G. 2006. Knowledge exchangeand combination: The role of human resource prac-tices in the performance of high-technology firms.Academy of Management Journal, 49: 544–560.

*Colquitt, J. A., Noe, R. A., & Jackson, C. L. 2002. Justicein teams: Antecedents and consequences of proce-dural justice climate. Personnel Psychology, 55: 83–109.

Combs, J., Liu, Y., Hall, A., & Ketchen, D. 2006. Howmuch do high-performance work practices matter? Ameta-analysis of their effects on organizational per-formance. Personnel Psychology, 59: 501–528.

Datta, D. K., Guthrie, J. P., & Wright, P. M. 2005. Humanresource management and labor productivity: Doesindustry matter? Academy of Management Journal,48: 135–145.

De Dreu, C. K. W., & Weingart, L. R. 2003. Task versusrelationship conflict, team performance, and teammember satisfaction: A meta-analysis. Journal ofApplied Psychology, 88: 741–749.

Devine, D. J., & Phillips, J. L. 2001. Do smarter teams dobetter? A meta-analysis of cognitive ability and teamperformance. Small Group Research, 32: 507–532.

DiTomaso, N., Post, C., & Parks-Yancy, R. 2007. Work-force diversity and inequality: Power, status, andnumbers. Annual Review of Sociology, 33: 473–501.

DiTomaso, N., Post, C., Smith, D. R., Farris, G. F., &Cordero, R. 2007. Effects of structural position onallocation and evaluation decisions for scientistsand engineers in industrial R&D. AdministrativeScience Quarterly, 52: 175–207.

*Drach-Zahavy, A., & Somech, A. 2002. Team heteroge-neity and its relationship with team support andteam effectiveness. Journal of Educational Admin-istration, 40: 44–66.

Drazin, R., & Van de Ven, A. 1985. Alternative forms of fitin contingency theory. Administrative ScienceQuarterly, 30: 514–539.

Dubin, R. 1976. Theory building in applied areas. InM. D. Dunnette (Ed.), Handbook of industrial andorganizational psychology: 17–39. Chicago: RandMcNally.

Eagly, A. H. 1995. The science and politics of comparingwomen and men. American Psychologist, 50: 145–158.

Eagly, A. H., & Steffen, V. J. 1984. Gender stereotypesstem from the distribution of women and men intosocial roles. Journal of Personality and SocialPsychology, 46: 735–754.

Ely, R. J. 1994. The effects of organizational demograph-ics and social identity on relationships among pro-

2009 623Joshi and Roh

fessional women. Administrative Science Quar-terly, 39: 203–238.

Ely, R. J. 1995. The power in demography: Women’ssocial constructions of gender identity at work.Academy of Management Journal, 38: 589–634.

*Ely, R. J. 2004. A field of group diversity, participationin diversity education programs, and performance.Journal of Organizational Behavior, 25: 755–780.

Fiske, S. T. 1993. Controlling other people: The impact ofpower on stereotyping. American Psychologist, 48:621–628.

Fiske, S. T. 1998. Stereotyping, prejudice, and discrimi-nation. In D. T. Gilbert, S. T. Fiske, & G. Lindzey(Eds.), Handbook of social psychology: 357–411.Boston: McGraw-Hill.

Fiske, S. T., Cuddy, A. J. C., Glick, P., & Xu, J. 2002. Amodel of (often mixed) stereotype content: Compe-tence and warmth respectively follow from per-ceived status and competition. Journal of Personal-ity and Social Psychology, 82: 878–902.

Frink, D. D., Robinson, R. K., Reithel, B., Arthur, M. M.,Ferris, G. R., Kaplan, D. M. et al. 2003. Gender de-mography and organizational performance: A twostudy investigation with convergence. Group andOrganization Management, 28: 127–147.

Gaertner, S. L., & Dovidio, J. F. 2000. Reducing inter-group bias: The common intergroup identitymodel. Philadelphia: Psychology Press.

Gibson, C., & Vermeulen, F. 2003. A healthy divide:Subgroups as a stimulus for team learning behavior.Administrative Science Quarterly, 48: 202–239.

Gomez-Mejia, L., & Balkin, D. 1992. Determinants of fac-ulty pay: An agency theory perspective. Academy ofManagement Journal, 35: 921–955.

Greer, L. L., & Jehn, K. A. 2008. Where perception meetsreality: The effects of different types of faultlineperceptions, asymmetries, and realities on inter-subgroup conflict and workgroup functioning.Working paper, Leiden University, The Netherlands.

Gully, S. M., Incalcaterra, K., Joshi, A., & Beaubien, J. M.2002. A meta-analysis of team efficacy, potency andperformance: Interdependence and level of analysisas moderators of observed relationships. Journal ofApplied Psychology, 87: 819–832.

Hackman, J. R. 1987. The design of work teams. In J. W.Lorsch (Ed.), Handbook of organizational behav-ior: 315–342. Englewood Cliffs, NJ: Prentice-Hall.

Haleblian, J., & Finkelstein, S. 1993. Top managementteam size, CEO dominance, and firm performance:The moderating role of environmental turbulenceand discretion. Academy of Management Journal,36: 844–863.

Hambrick, D. C., Cho, T. S., & Chen, M. 1996. The influ-ence of top management team heterogeneity onfirms’ competitive move. Administrative ScienceQuarterly, 41: 659–684.

Hambrick, D. C., & Finkelstein, S. 1987. Managerial dis-cretion: A bridge between polar views on organiza-tions. In L. L. Cummings & B. M. Staw (Eds.), Re-search in organizational behavior, vol. 9: 369–406.Greenwich, CT: JAI Press.

Hambrick, D. C., & Mason, P. A. 1984. Upper echelons:The organization as a reflection of its top managers.Academy of Management Review, 9: 193–206.

Harrison, D. A., & Klein, K. J. 2007. What’s the differ-ence? Diversity constructs as separation, variety, ordisparity in organizations. Academy of Manage-ment Review, 32: 1199–1228.

Harrison, D. A., Price, K. H., & Bell, M. P. 1998. Beyondrelational demography: Time and the effects of sur-face- and deep-level diversity on work group cohe-sion. Academy of Management Journal, 41: 96–107.

Harrison, D. A., Price, K. H., Gavin, J. H., & Florey, A. T.2002. Time, teams, and task performance: Changingeffects of surface- and deep-level diversity on groupfunctioning. Academy of Management Journal, 45:1029–1045.

Hedge, L. V., & Olkin, I. 1985. Statistical methods formeta-analysis. Orlando, FL: Academic Press.

*Herron, M. T. 1993. The effects of the ethnic andgender diversity of work team on the perceptionsof performance outputs. Unpublished doctoral dis-sertation, California School of Professional Psychol-ogy, Los Angeles.

Hilton, J. J., & von Hippel, W. 1996. Stereotypes. In J. T.Spence, J. M. Darley, & D. J. Foss (Eds.), Annualreview of psychology, vol. 47: 237–271. Palo Alto,CA: Annual Reviews.

Homan, A. C., & Van Knippenberg, D. 2003. The benefi-cial effects of cross-categorizing informational anddemographical diversity in groups. Paper pre-sented at the 11th European Congress of Work andOrganizational Psychology, Lisbon.

Horwitz, S. K., & Horwitz, I. B. 2007. The effects of teamdiversity on team outcomes: A meta-analytic reviewof team demography. Journal of Management, 33:967–1015.

Huffcutt, A. I., & Arthur, W., Jr. 1995. Development of anew outlier statistic for meta-analytic data. Journalof Applied Psychology, 80: 327–334.

Hunter, J. E., & Schmidt, F. L. 1990. Methods of meta-analysis: Correcting error and bias in researchfindings. Newbury Park, CA: Sage.

Hultin, M., & Szulkin, R. 1999. Wages and unequal accessto organizational power: An empirical test of genderdiscrimination. Administrative Science Quarterly,44: 453–472.

Ilgen, D. R., Hollenbeck, J. R., Johnson, M., & Jundt, D.2005. Teams in organization: From input-process-output models to IMOI models. In S. Fiske, D. L.Schacter, & A. Kasdin (Eds.), Annual review of psy-

624 JuneAcademy of Management Journal

chology, vol. 56: 517–543. Palo Alto, CA: AnnualReviews.

Information Technology Association of America. 2003.Report of the Information Technology Associationof America (ITAA) blue ribbon panel on IT diver-sity. Presented at the National IT Workforce Convo-cation, Arlington, VA.

*Jackson, S. E., & Joshi, A. 2004. Diversity in socialcontext: A multi-attribute, multi-level analysis ofteam diversity and sales performance. Journal ofOrganizational Behavior, 25: 675–702.

Jackson, S. E., Joshi, A., & Erhardt, N. L. 2003. Recentresearch on team and organizational diversity:SWOT analysis and implications. Journal of Man-agement, 29: 801–830.

Jackson, S. E., May, K. E., & Whitney, K. 1995. Under thedynamics of diversity in decision-making teams. InR. A. Guzzo & E. Salas (Eds.), Team effectivenessand decision making in organizations: 204–261.San Francisco: Jossey-Bass.

*Jarrell, K. A. 2002. Job-related diversity and serviceteam outcomes: New insights into the roles of taskstructure and process conflict. Unpublished doc-toral dissertation, Syracuse University.

*Jehn, K. A., & Bezrukova, K. 2004. A field study of groupdiversity, workgroup context, and performance.Journal of Organizational Behavior, 25: 703–729.

*Jehn, K. A., Northcraft, G. B., & Neale, M. A. 1999.Whydifferences make a difference: A field study of diver-sity, conflict, and performance in workgroups.Administrative Science Quarterly, 44: 741–763.

Johns, G. 2006. The essential impact of context on organ-izational behavior. Academy of Management Re-view, 31: 386–408.

*Joshi, A. 2002. How does context matter? Examiningthe process and performance outcomes of workteam heterogeneity. Unpublished doctoral disserta-tion, Rutgers University, New Brunswick, NJ.

Joshi, A., Liao, H., & Jackson, S. E. 2006. Cross-leveleffects of workplace diversity on sales performanceand pay. Academy of Management Journal, 49:459–481.

Jost, J. T., & Banaji, M. R. 1994. The role of stereotypingin system-justification and the production of falseconsciousness. British Journal of Social Psychol-ogy, 33: 1–27.

*Kearney, E., & Gerbert, D. 2009. Managing diversity andenhancing team outcomes: The promise of transfor-mational leadership. Journal of Applied Psychol-ogy: In press.

*Kearney, E., Gerbert, D., & Voelpel, S. 2009. When andhow diversity benefits teams: The importance ofteam members’ need for cognition. Academy ofManagement Journal, 52: 581–598.

Keck, S. 1997. Top management team structure: Differ-

ential effects by environmental context. Organiza-tion Science, 8: 143–156.

*Keller, R. T. 2001. Cross-functional project groups inresearch and new product development: Diversity,communications, job stress and outcomes. Academyof Management Journal, 44: 547–555.

*Kirkman, B. L., Tesluk, P. E., & Rosen, B. 2004. Theimpact of demographic heterogeneity and team lead-er-team member demographic fit on team empower-ment and effectiveness. Group and OrganizationManagement, 29: 334–368.

Kochan, T., Bezrukova, K., Ely, R., Jackson, S., Joshi, A.,Jehn, K. et al. 2003. The effects of diversity on busi-ness performance: Report of the diversity researchnetwork. Human Resource Management, 42: 3–21.

Larkey, L. K. 1996. Toward a theory of communicativeinteractions in culturally diverse workgroups. Acad-emy of Management Review, 21: 463–491.

Lau, D. C., & Murnighan, J. K. 1998. Demographic diver-sity and faultlines: The compositional dynamics oforganizational groups. Academy of ManagementReview, 23: 325–340.

Lau, D. C., & Murnighan, J. K. 2005. Interactions withingroups and subgroups: The effects of demographicfaultlines. Academy of Management Journal, 48:645–659.

Lawrence, P., & Lorsch, J. 1967. Organization and itsenvironment. Cambridge, MA: Harvard UniversityPress.

*Leonard, J. S., Levine, D. I., & Joshi, A. 2004. Do birds ofa feather shop together? The effects on performanceof employees’ similarity with one another and withcustomers. Journal of Organizational Behavior, 25:731–754.

LePine, J. A., Hanson, M. A., Borman, W. C., & Motow-idlo, S. J. 2000. Contextual performance and team-work: Implications for staffing. In G. R. Ferris & K. M.Rowland (Eds.), Research in personnel and humanresources management, vol. 19: 53–90. Greenwich,CT: JAI Press.

Li, J., & Hambrick, D. C. 2005. Factional groups: A newvantage on demographic faultlines, conflict, and dis-integration in work teams. Academy of Manage-ment Journal, 48: 794–813.

Linnehan, F., & Konrad, A. M. 1999. Diluting diversity:Implications for intergroup inequality in organiza-tions. Journal of Management Inquiry, 8: 399–414.

Lipsey, M. W., & Wilson, D. B. 2001. Practical meta-analysis. Thousand Oaks, CA: Sage.

*Lovelace, K., Shapiro, D. L., & Weingart, L. R. 2001.Maximizing cross-functional new product teams’ in-novativeness and constraint adherence: A conflictcommunications perspective. Academy of Manage-ment Journal, 44: 779–793.

*Marsteller, J. A. 2003. The relationship between non-racial diversity in team composition and perfor-

2009 625Joshi and Roh

mance in creativity in a chronic illness carequality improvement intervention. Unpublisheddoctoral dissertation, University of California,Berkeley.

Martins, L. L., Miliken, F. J., Wiesenfeld, B. M., & Sal-gado, S. R. 2003. Racioethnic diversity and groupmember’s experience. Group and OrganizationManagement, 28: 75–106.

Mayo, M., Pastor, J., & Meindl, J. R. 1996. The effects ofgroup heterogeneity on the self-perceived efficacy ofgroup leaders. Leadership Quarterly, 7: 265–284.

McGrath, J. E. 1984. Groups: Interaction and perfor-mance. Englewood Cliffs, NJ: Prentice-Hall.

Milkovich, G. T. 1987. Compensation systems in hightechnology companies. In D. B. Balkin & L. R. Go-mez-Mejia (Eds.), New perspectives on compensa-tion. Englewood Cliffs, NJ: Prentice-Hall.

Milliken, F. J., & Martins, L. L. 1996. Searching for com-mon threads: Understanding the multiple effects ofdiversity in organizational groups. Academy ofManagement Review, 21: 402–433.

Organisation for Economic Cooperation and Develop-ment. 2006. OECD science, technology and indus-try outlook. Organisation for Economic Cooperationand Development.

Orwin, R. G. 1983. A fail-safe N for effect size inmeta-analysis. Journal of Educational Statistics, 8:157–159.

Pelled, L. H. 1996. Relational demography and percep-tions of group conflict and performance: A fieldinvestigation. International Journal of ConflictManagement, 7: 230–246.

*Pelled, L. H., Eisenhardt, K. M., & Xin, K. R. 1999.Exploring the black box: An analysis of work groupdiversity, conflict, and performance. AdministrativeScience Quarterly, 44: 1–28.

Perry, E. L. 1997. A cognitive approach to understandingdiscrimination: A closer look at applicant genderand race. In G. R. Ferris (Ed.), Research in personneland human resource management, vol. 15: 175–240. Greenwich, CT: JAI Press.

Pettigrew, T. F. 1998. Intergroup contact theory. In J. T.Spence, J. M. Darley, & D. J. Foss (Eds.), Annualreview of psychology, vol. 49: 65–85. Palo Alto, CA:Annual Reviews.

Porter, M. E. 1980. Competitive strategy. New York: FreePress.

*Portero-Brown, R. A. 1999. The missing link betweenteam heterogeneity and effectiveness: Individualand contextual influences on social categorysalience. Unpublished doctoral dissertation, Uni-versity of California, Berkeley.

*Puck, J., Rygl, D., & Kittler, M. 2006. Cultural anteced-ents and performance consequences of open commu-nication and knowledge transfer in multiculturalprocess-innovation teams. Journal of Organiza-

tional Transformation and Social Change, 3:223–241.

Quinn, J. B., Anderson, P., & Finkelstein, S. 1996. Leverag-ing intellect. Academy of Management Executive,10(3): 7–27.

Ragins, B., & Sundstorm, E. 1990. Gender and powerorganizations: A longitudinal perspective. Psycho-logical Bulletin, 105: 51–88.

*Raver, J. L., & Gelfand, M. J. 2005. Beyond the individualvictim: Linking sexual harassment, team processes,and team performance. Academy of ManagementJournal, 48: 387–400.

*Reagans, R., & Zuckerman, E. W. 2001. Networks, diver-sity, and productivity: The social capital of corporateR&D teams. Organization Science, 12: 502–517.

Reagans, R., Zuckerman, E. W., & McEvily, B. 2004. Howto make the team: Social networks vs. demographyas criteria for designing effective teams. Administra-tive Science Quarterly, 49: 101–133.

Reskin, B. F., McBrier, D. B., & Kmec, J. 1999. The deter-minants and consequences of workplace sex andrace composition. In J. Hagan & K. Cook (Eds.), An-nual review of sociology, vol. 25: 355–361. PaloAlto, CA: Annual Reviews.

Richard, O. C. 2000. Racial diversity, business strategy,and firm performance: A resource-based view. Acad-emy of Management Journal, 43: 164–177.

Richard, O. C., Barnett, T., Dwyer, S., & Chadwick, K.2004. Cultural diversity in management, firm perfor-mance, and the moderating role of entrepreneurialorientation dimensions. Academy of ManagementJournal, 47: 255–266.

Richard, O. C., Murthi, B. P. S., & Ismail, K. 2007. Theimpact of racial diversity on intermediate and long-term performance: The moderating role of environ-mental context. Strategic Management Journal, 28:1213–1233.

Ridgeway, C. L. 1991. The social construction of statusvalue: Gender and other nominal characteristics. So-cial Forces, 70: 367–386.

Ridgeway, C. L. 1997. Interaction and the conservation ofgender inequality: Considering employment. Amer-ican Sociological Review, 62: 218–235.

Rosen, B., & Jerdee, T. H. 1977. Too old or not too old.Harvard Business Review, 55(6): 97–106.

Rosenthal, R. 1979. The “file drawer problem” and toler-ance of null results. Psychological Bulletin, 86:638–641.

Rousseau, D. M., & Fried, Y. 2001. Location, location,location: Contextualizing organizational research.Journal of Organizational Behavior, 22: 1–13.

Saavedra, R., Earley, P., & Van Dyne, L. 1993. Complexinterdependence in task-performing groups. Journalof Applied Psychology, 78: 61–72.

*Sacco, J. M., & Schmitt, N. A. 2005. A dynamic multi-

626 JuneAcademy of Management Journal

level model of demographic diversity and misfit ef-fects. Journal of Applied Psychology, 90: 203–231.

*Schippers, M. C., Den Hartog, D. N., Koopman, P. L., &Wienk, J. A. 2003. Diversity and team outcomes: Themoderating effects of outcome interdependence andgroup longevity and the mediating effect of reflexiv-ity. Journal of Organizational Behavior, 24: 779–802.

Shea, G. P., & Guzzo, R. A. 1987. Groups as humanresources. In K. M. Rowland & G. R. Ferris (Eds.),Research in personnel and human resource man-agement, vol. 5: 323–356. Greenwich, CT: JAI Press.

Shore, L. M., & Goldberg, C. B. 2005. Age discriminationin the workplace. In R. L. Dipboye & A. Colella(Eds.), Discrimination at work: The psychologicaland organizational bases: 203–225. Mahwah, NJ:Erlbaum.

Sidanius, J. 1993. The psychology of group conflict andthe dynamics of oppression: A social dominanceperspective. In S. Iyengar & W. J. McGuire (Eds.),Explorations in political psychology: 183–219.Durham, NC: Duke University Press.

Skaggs, S., & DiTomaso, N. 2004. Understanding the ef-fects of workforce diversity on employment out-comes: A multidisciplinary and comprehensiveframework. In N. DiTomaso & C. Post (Eds.), Re-search in the sociology of work: 279–306. NewYork: Elsevier.

*Somech, A. 2006. The effects of leadership style andteam process on performance and innovation infunctionally heterogeneous teams. Journal of Man-agement, 32: 132–157.

Teachman, J. D. 1980. Analysis of population diversity.Social Methods and Research, 8: 341–362.

Thompson, J. D. 1967. Organizations in action. NewYork: McGraw-Hill.

*Tyran, K. L., & Gibson, C. B. 2008. Is what you see, whatyou get? The relationship among surface- and deep-level heterogeneity characteristics, group efficacy,and team reputation. Group and Organization Man-agement, 33: 46–76.

U. S. Bureau of Labor Statistics. 2006. Labor force sta-tistics from the Current Population Survey. Wash-ington, DC: U. S. Department of Labor.

U. S. Bureau of Labor Statistics. 2007. Occupationaloutlook handbook, 2006–2007. Washington, DC:U.S. Department of Labor.

U. S. Census Bureau. 2002. North American IndustryClassification System (NAICS)–Revisions for 2002.Washington, DC: U. S. Census Bureau.

*Van der Vegt, G. S., & Bunderson, J. S. 2005. Learningand performance in multidisciplinary teams: Theimportance of collective team identification. Acad-emy of Management Journal, 48: 532–547.

Van der Vegt, G. S., & Janssen, O. 2003. Joint impact of

interdependence and group diversity on innovation.Journal of Management, 29: 29–51.

Van der Vegt, G. S., Van de Vliert, E., & Huang, X. 2005.Location-level links between diversity and innova-tive climate depend on national power distance.Academy of Management Journal, 48: 1171–1182.

Van de Ven, A. H., Delbecq, A. L., & Koenig, R. 1976.Determinants of coordination modes within organi-zations. American Sociological Review, 41: 322–338.

Van Knippenberg, D., De Dreu, C. K. W., & Homan, A. C.2004. Work group diversity and group performance:An integrative model and research agenda. Journalof Applied Psychology, 89: 1008–1022.

Van Knippenberg, D., & Schippers, M. C. 2007. Work-group diversity. In M. I. Posner & M. K. Rothbart(Eds.), Annual review of psychology, vol. 58: 2.1–2.27. Palo Alto, CA: Annual Reviews.

Watson, W. E., Johnson, L., & Merritt, D. 1998. Teamorientation, self-orientation, and diversity in taskgroups: Their connection to team performance overtime. Group and Organization Management, 23:161–282.

Webber, S. S., & Donahue, L. M. 2001. Impact of highlyand less job-related diversity on work group cohe-sion and performance: A meta-analysis. Journal ofManagement, 27: 141–162.

Whitener, E. M. 1990. Confusion of confidence intervalsand credibility intervals in meta-analysis. Journal ofApplied Psychology, 75: 315–321.

Williams, K., & O’Reilly, C. A. 1998. Demography anddiversity in organizations: A review of 40 years ofresearch. In B. M. Staw & L. L. Cummings (Eds.),Research in organizational behavior, vol. 20: 77–140. Greenwich, CT: JAI Press.

*Yeh, Y., & Chou, H. 2005. Team composition and learn-ing behaviors in cross-functional teams. SocialBehavior and Personality, 33: 391–402.

Aparna Joshi (aparnajo@illinois.edu) is an assistant pro-fessor of labor and employment relations at the Univer-sity of Illinois, Urbana-Champaign. She received herPh.D. from the School of Management and Labor Rela-tions, Rutgers University. She conducts research in thearea of work team diversity, global and distributedteams, team social capital, and generational issues inthe workplace.

Hyuntak Roh (hroh2@illinois.edu) is a Ph.D. candidatein the School of Labor and Employment Relations at theUniversity of Illinois at Urbana-Champaign. His researchinterests include workforce diversity, group process andperformance, employee turnover, and multilevel contex-tual issues in HR/OB research.

2009 627Joshi and Roh