The Role of Context in Work Team Diversity Research: a Meta-Analytic Review Author(S): Aparna Joshi and Hyuntak Roh Source: the Academy of Management Journal, Vol

The Role of Context in Work Team Diversity Research: A Meta-Analytic Review Author(s): Aparna Joshi and Hyuntak Roh Source: The Academy of Management Journal, Vol. 52, No. 3 (Jun., 2009), pp. 599-627 Published by: Academy of Management Stable URL: http://www.jstor.org/stable/40390306 Accessed: 22-02-2018 12:46 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://about.jstor.org/terms

Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to The Academy of Management Journal

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms © Academy of Management Journal 2009, Vol. 52, No. 3, 599-627.

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

APARNA JOSHI HYUNTAK ROH University of Illinois at Urbana-Champaign

Integrating macro and micro theoretical perspectives, we conducted a meta-analysis examining the role of contextual factors in team diversity research. Using data from 8,757 teams in 39 studies conducted in organizational settings, we examined whether contextual factors at multiple levels, including industry, occupation, and team, influ- enced the performance outcomes of relations-oriented and task-oriented diversity. The direct 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 in effect sizes across studies.

Research in the area of work team diversity has applications of these theoretical perspectives have grown exponentially in the last four decades. How- offered broad generalizations for why differences ever, several comprehensive reviews have noted within work groups may manifest in specific atti- that the findings in this area do not provide a clear tudinal outcomes, such as conflict or cohesion, or consensus regarding the performance effects of in behavioral outcomes, such as turnover or absen- work team diversity (Harrison & Klein, 2007; Jack- teeism (see Jackson et al., 2003). In addition to son, Joshi, & Erhardt, 2003; Milliken & Martins, asking why diversity manifests in specific out- 1996; Van Knippenberg & Schippers, 2007; Wil- comes, a careful examination of situational settings liams & O'Reilly, 1998). In some studies, research- would also ask when, where, and how diversity dy- ers have reported that team diversity is positively namics unfold in workplaces; these contextual con- associated with performance (e.g., Ely, 2004; Van siderations, not often captured (see Johns, 2006), are der Vegt, Van de Vliert, & Huang, 2005). In another pertinent for reconciling the mixed findings from past set of studies, team diversity has been found to research. Researchers have attempted to reconcile negatively predict performance (e.g., Jehn, North- these findings within prevalent theoretical traditions craft, & Neale, 1999; Leonard, Levine, & Joshi, by considering the influence of organization-level 2004). A majority of these studies, however, have and team-level factors on diversity outcomes (e.g., reported a nonsignificant, direct relationship be- Kirkman et al., 2004; Kochan et al., 2003). However, tween team diversity and performance. Further- these explanations are often offered post hoc; contex- more, even within studies, the effects of gender, tual factors have less often been incorporated in hy- race, age, and tenure diversity on team performance pothesis development or in study design. We propose have varied (e.g., Kirkman, Tesluk, & Rosen, 2004; that a unified and comprehensive contextual frame- Kochan et al, 2003). work for diversity research has the potential to re- Current theoretical perspectives framing diver- solve these inconsistent findings and can contribute sity research, such as social identity theory, social to further theoretical and empirical developments in categorization theory, and the attraction-selection- the field. attrition framework, appear to be insufficient for This article highlights contextual issues in team resolving these mixed findings. In general, current diversity research. Our study takes a substantively different approach from past meta-analytic reviews on this topic (Bowers, Pharmer, & Salas, 2000; Hor- A version of this paper was presented at the 2008 Annual witz & Horwitz, 2007; Webber & Donahue, 2001). Meeting of the Academy of Management and won the Do- Rather than test whether diversity attributes have a rothy Harlow Distinguished Paper Award. We are indebted positive or a negative effect on team performance, to Joseph Martocchio for advice and feedback on drafts of this work. We are also extremely grateful to Jason Colquitt we accounted for aspects of context at multiple levels and examined whether these contextual fac- and three anonymous reviewers for invaluable feedback throughout the review process. We thank Niti Pandey and tors shaped the team diversity-performance rela- Erik Young for research assistance. tionship. In proposing and testing the contextual

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 express written permission. Users may print, download or email articles for individual use only.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 600 Academy of Management Journal June framework presented in this KEY CONCEPTS article, AND we attempt to make several contributions THEORETICAL to BACKGROUNDwork team diversity research. First, we develop and test a theoretically We define diversity as an aggregate team-level driven framework for work team diversity context construct that represents differences among mem- across multiple levels bersof of ananalysis. interdependent workSeveral group with research- respect ers have acknowledged to a specific that personal contextual attribute (Jackson consider-et al., ations are critical in diversity2003). In line with pastresearch research, we distinguish(e.g., Jack- son et al., 2003; Martins, Milliken, Wiesenfeld, & between task-oriented and relations-oriented as- Salgado, 2003). However, a theoretically driven pects of diversity (Jackson, May, & Whitney, 1995). multilevel framework explicating contextual deter- Relations-oriented diversity attributes such as gen- minants of work team diversity outcomes has not der, race/ethnicity, and age are cognitively accessi- been forthcoming. Our study addresses this gap. ble, pervasive, and immutable and are associated Some scholars have also drawn attention to struc- with social categorization processes (Fiske, 1998; tural and institutional factors that give meaning to Van Knippenberg, De Dreu, & Homan, 2004). These demographic differences in organizations (DiTo- social categorization-based processes, which man- maso, Post, & Parks-Yancy, 2007; Ragins & Sund- ifest in intergroup bias and negative attitudes storm, 1990). Yet the prevalent theoretical perspec- toward dissimilar others in a group may have neg- tives we referred to in our introduction do not ative performance consequences. In contrast, task- account for these institutional and structural fac- oriented diversity attributes, such as education, tors. Therefore, a second contribution we attempt is function, and tenure, are associated with skill-based an integration of work team diversity research with and informational differences among work group macro theoretical perspectives that account for members (Jackson et al., 1995). These aspects of di- these factors and have received relatively scant at- versity are assumed to constitute a team's cognitive resource base and are associated with elaboration- tention in the past. Furthermore, team task characteristics can also shape the salience of diversity based processes, defined as the exchange of information and perspectives among group mem- attributes within teams (Van Knippenberg & Schip- bers, individual-level information processing, gaining pers, 2007); we also consider whether these characteristics are a relevant team-level context influenc- feedback, and integrating information and perspectives. These elaboration-based processes ex- ing the diversity-performance link. Third, we plain the positive performance outcomes of work integrate 15 years of field research on the perfor- group diversity (see Van Knippenberg et al., 2004). mance outcomes of work team diversity. Our re- In the subsequent sections, we discuss how va- view distinguishes between the effects of task-ori- rious aspects of diversity context can influence the ented (e.g., function, education, and tenure) and categorization-based processes associated with re- relations-oriented (e.g., gender, race/ethnicity, and lations-oriented diversity and the elaboration- age) aspects of diversity in relation to performance. based processes associated with task-oriented di- Finally, this study answers a call in the broader versity, and implications for team performance. domain of management research for a more context Although more sophisticated conceptualizations of based understanding of workplace phenomena team diversity have been developed (e.g., Harrison (Bamberger, 2008; Johns, 2006; Rousseau & Fried, & Klein, 2007; Lau & Murnighan, 1998), we draw on 2001). Context theorizing in management research a more simplified yet established typology of diver- can enhance its "market orientation," which not sity attributes to develop our key propositions re- only makes study findings more accessible to man- garding the role of contextual factors in diversity agers but is also important for future theory build- research.1 ing (Bamberger, 2008; Dubin, 1976). Team performance is defined as the extent to In the following section, we discuss in detail how we conceptualize diversity context and outline various aspects of this context. Next, we develop hy- 1 Some researchers have also conceptualized team di- potheses that delineate the moderating effects of versity by differentiating between surface- and deep- these contextual factors on team diversity in rela- level diversity (e.g., Harrison, Price, & Bell, 1998). Other researchers have proposed an approach based on "fault- tion to team performance. Finally, we present the lines" (e.g., Lau & Murnighan, 1998). Recently, Harrison findings from a meta-analytic review and consider and Klein (2007) posed "separation," "variety," and "dis- the implications of these contextual considerations parity" as diversity dimensions to consider. We discuss for future theoretical and empirical developments these alternative approaches later, in the Discussion in diversity research. section.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 601 which a team Among the studiesaccomplishes that did incorporate these vari- its goal or mission (De- vine & Phillips, ables, the focus was on2001). national culture, customer We focus on performance as an outcome base demography,of market diversity competition, and rate of because this outcome has received the technological most change (e.g., Ancona &research Caldwell, 1992; attention and represents an area in which Lovelace, Shapiro, & Weingart, the 2001; Leonard et mixedal., findings have been the most prevalent 2004; Reagans & Zuckerman, (Jackson 2001; Van der Vegt et et al., 2003). Results pertain- ing to performance al., 2005). On the basis of this review, weoutcomes aimed to of diversity may need the most integration, take stock of contextual effects that are conceptually and understanding the contextual factors distinct shaping and have been considered in pastperformance research outcomes may have the most value and also to incorporate for additional contextualresolving vari- past findings. Context has been defined as the situational set- ables that have received less attention in the past. ting in which workplace phenomena occur (Cap- Theoretical and practical considerations gov- pelli & Sherer, 1991). In recent theoretical advance- erned our efforts to identify key contextual factors. ments, it has been recognized that various aspects Sociopsychological theoretical perspectives sug- of context may serve as "situational opportunities gest that the demography of the job or occupation for and countervailing constraints against organiza- in which diverse work groups are embedded can tional behavior [and] be represented as a tension shape categorization-based processes in these system or force field comprising such opportunities groups (DiTomaso, Post, & Parks-Yancy, 2007; Lar- and constraints" (Johns, 2006: 387). Drawing on key, 1996; Reskin, McBrier, & Kmec, 1999).2 Unlike this perspective, we propose that context can set in a balanced setting, in an occupational context specific constraints and opportunities that either dominated by a single demographic group, negative enhance or minimize the direct effects of work stereotypes about underrepresented groups can in- team diversity on performance. Opening the dis- fluence categorization-based outcomes within course on team diversity to contextual influences is work groups (Hilton & Von Hippel, 1996; Larkey, challenging because it requires one to specify team, 1996). We extend these perspectives to focus on organizational, and extraorganizational factors that occupational demography as a contextual factor may comprise a "tension system" shaping diversity that can enhance or minimize the influence of effects. Therefore, we undertook an extensive re- team diversity on performance outcomes. view of team diversity research conducted in or- We also consider industry as an embedding con- ganizational settings from 1992 through 2008 to text for diversity-based outcomes. Strategic man- identify aspects of diversity context that past re- agement research suggests that interindustry varia- search has explicitly acknowledged as either mod- tion in levels of technological change, regulatory erator or as control variables. Table 1 represents the pressure, customer demands, and market competi- findings of this review. tion are greater than intraindustry variations (Bour- Our review indicated that approximately 60 per- geois, 1985; Porter, 1980). Furthermore, these in- cent of the direct effects reported in past research dustry-level contingencies can serve as situational were nonsignificant for various diversity attributes. enhancers or minimizers of diversity effects on per- Among the remainder, 20 percent of the effects formance (Hambrick, Cho, & Chen, 1996; Richard, reported were significantly positive, and 20 percent Murthi, & Ismail, 2007). Within the strategic man- were significantly negative. Researchers have con- agement domain, considerable research on top sidered contextual variables primarily at the team management teams and research on the firm diver- level to explicate these mixed effects of team diversity-performance link has incorporated industry- sity on performance. Among the studies we reviewed, level context (e.g., service versus manufacturing) as over 70 percent accounted for team-level contextual a key moderator (e.g., Haleblian & Finkelstein, factors such as task interdependence, complexity, cli- 1993; Hambrick et al., 1996; Keck, 1997; Richard et mate, and other team-level perceptual variables (e.g., al., 2007). We extend these perspectives to consider Jehn et al., 1999; Pelled, Eisenhardt, & Xin, 1999; industry setting as a relevant aspect of a team's Schippers, Den Hartog, Koopman, & Wienk, 2003; Van der Vegt & Bunderson, 2005). Approximately 20 2 We acknowledge that the proximal work context rep- percent of the studies represented in Table 1 exam- resented by organization-level demography, culture, and ined contextual moderators at the organizational climate has received some attention in past research and level, including organizational demography, diver- that these are also important contextual variables to con- sity training participation, and organizational culture sider. However, of the studies included in this review, only (e.g., Ely, 2004; Jackson & Joshi, 2004; Jehn & three provided information regarding organization-level de- Bezrukova, 2004). Less than 10 percent of the studies mography, culture, or climate, and therefore these variables reviewed examined extraorganizational factors. could not be included in the meta-analysis.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms S | "I

la I 1 § §' 1 g 8'

i • j ' ^ 1-1 II M I Ul I ¡ü!

co^ °u §R bCU o co § cd 5 w® u ©«S th P^

11!in îiïSZ Tf ! iîPL, Ufa tH 1 HlZ Sï^SCO 2t[» CO Sìa 2l!!l IO Sîîi CU tï OOi|ïS ihJ ila CDill tj ï|arHpL, ¡ ïïg-g rn 1|1 Pm ji?; rHl|i I PhII ii * !i!|| î ! f. Ia! î } i Ijll! i i ! i î

i -I i *

w l U 18g - i 5if l| II ¡s I if *•"•§ " Sh" CO ^pqSig5 CO 3 CO0 ^ü CO ß£ CO SS JÖ "SS, Jd oJd o> <

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms e?

I g

r . Jf ¡, Ì t . I §l!§ H 11

> f|Q* o T3|||| S TÍ DppClIjij S O DO'SI| §0!| § Q í O O O

!,■8§1 ■8 88 ihlês^fS ^fS pi a°§i!. a°§«lií «lií liU ilâi «lií■ ss |s JSJS ■Id SilSils |Hlilft I-Sil un I-Sil sSg^s malsSg^s «II ||||

5 • 9 « Ö^«n,T3O na .S ...S E« H ii« iï M ! îî il J ï If! .tïf 'U* }* * ï W Ü1 11? JI| : I, 1J1 f II lili ij{ ililllllll 1!J|

oo^ coed coro cjä ;o>2 th.íh 'Hcd cmS3 t^S cm.5 «S «.S? ri© oo^COZ filili "* coed PLh coro ^P-i£¡1 cjä 0î2i ¡il ;o>2COZ 5I| I r-l(Z)lili th.íh rHfl,'Hcd i I{1 O3W cmS3 Hg* tvpL, 5111 t^S

ïf I , 1 . Il i Í v I ï 5ÍT1tí a o• 5. oc »„-hoa? o»5 s . s« S Sa ^3 -83 ö if l1ï!fi1 Nî I II I11 ï1 -S 11 |¡Í I -SIIIJî ë§ljJ 'S I' "8 I Í35 -SI IsII

UQ11 ilwE s ÃI 1Ã iÃ |i Ã ïl Ã S WÜ Mil «^4i II

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms (4 +J OSO 4-1 CO i ^ |1 |i 1 |i J-alsl 1^1

.2I 3 cdob.2 cao Î w t:3 52 OTIt > ô cdob I cao lil w t: 52 H_ OT > ô .2 S & 3 lei- cdob il* cao llg w t: 52 1§! OT > ô T2 j¿ © M p p £v T2 O° j¿ ihJ 'S © -8 O li MPQ g |Ä-§ PQp |S1 p 1 i ,i«i |î Ìli "i Ph C OJ O O ïj «ïi V tí +J

51 tí tí * ^

íl Iftf J î i íit îtî L ¡îî it ut tí idi !î! î itr. un ü m ni i!i li Hit h di ii un sii

Ni-]Ni-] Sj« ^Z^Z »5|# -^CL.iDWrfZ'ifZ 11 -^CL.iDWrfZ'ifZ PI Jl| !|#«PL, ?1l «PL, N N h) P:P h)(nZ^ZcoZ (nZ^ZcoZ || |l| i|#thZ thZ CMC!«P|t CMC!« ||a

g I 1 | I g h I" lisi és. I il s 1 i i H ir 1 1 1 I1 í , I, i ! »! li „ f ì li ìì « . Ö tJ © T3 fl . S "S g4 o 03 co © CJ Ch © ¿j CÍ U fí ^¿ Œ pm rH M ^ O ü Ü CO p OOOcOœûO'S pm OTJcflcO CO

. í~ íl t -i Sfa 1 ¡i s ii i

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms - * ¡I - ■I lì * I 2

Si» & e 2 'ö 55 >■> 2 g o « ft 'S oo

Sr^Sr^ 'ö III 55| >■>IS 2 IS Ilallfl^ltì g§ o « •Ila ftJ 'S § 3J |B >oo 111! i limili gj

I«- .s lá « ¿ o! II I li l 4Ã « ** Si Jf lui ji Ji- :sïji fü in« « s ut î .it Jif Jin mí m 1 « 1

Ifl5t© 5t© illHiîco° co° ^«3 ^«3 MUt^^5 t^^5 ilillt^cd t^cd «ti i|l 03*3? 03*3 I info rnZ inCL, mZ cmPh ooc/3 rStí .a i 1 j § l S I ■II *. I li Î il 111 il |i II i ¡ l I iì Û0ti*X3§ö§ m S^-E m S^-E 8 I00 § ^§ «t-jO)Se-* S)- § S8 ^ S) § 2 S5 < fi ow < £îëol3 | ¡g

cO , - , O H DO ^ m m J3

¿tiEdöSm ï h3î g«si ti>2i Sii rfa î cö^ü-o im u o ?*> ™ ™ ®

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 606 Academy of Management Journal June

diversity context. On work a practicalanalysts). Thus, the note,term "occupation" information refers regarding occupational to this collectiveand descriptionindustry of a number setting of indi- was available from most vidualstudy jobs performed, descriptions. with minor variations, Addition- in ally, publicly accessible many establishments databases (U. S. Bureau provide of Labor Statis- a wealth of objective information tics, 2007). regarding A significant body the of sociological occupational re- and industry contexts search in shows which that occupation-level teams demographic are embedded. Therefore, we focus composition on can have industry important implications setting for and occupational demography gender-based as and contextual ethnicity- or race-based factors diversity that provide a comprehensive in organizations understanding (for a review, see Reskin of eta al.team's macrolevel context and[1999]). Wecould propose belowpotentially that occupational de-account for the mixed findings mography reported. serves as a situational setting that can The nature of a team's enhance task the effects can of relations-orientedhave a significant diversity influence on the extent to which team members are and minimize the effects of task-oriented diversity interdependent in terms of goals and task outcomes on team performance. (see Ilgen, Hollenbeck, Johnson, & Jundt, 2005; Considerable research on stereotype formation LePine, Hanson, Borman, & Motowidlo, 2000). As enables us to specify the psychological processes the review above suggests, team-level contextual by which an occupational demography can influ- variables have received the most attention in past ence diversity dynamics within teams. This re- research. Social categorization theory would sug- search suggests that stereotypes, which emerge in gest that aspects of a team's task can minimize the response to environmental factors - such as differ- salience of diversity attributes by reinforcing a ences in social roles (Eagly, 1995) and in power common group identity or by placing demands on (Fiske, 1993) - and as a means to justify the status the team's diverse cognitive resource base (Gaert- quo (Jost & Banaji, 1994; Sidanius, 1993), result in ner & Dovidio, 2000; Jehn et al., 1999). Information categorization-based responses toward targeted processing theory would suggest that the nature of groups (Allport, 1954). When a single demographic tasks would place requirements on a team's cogni- group dominates an occupation, negative stereo- tive resource base with implications for the sa- type-based categorization processes against under- lience of diversity attributes within the team (Jehn represented groups are likely (Fiske, 1993; Reskin et al., 1999; Williams & O'Reilly, 1998). Our review et al., 1999). Once categorization-based processes also indicated that information regarding aspects of are invoked, additional information regarding tar- teams' tasks, such as team interdependence and geted group members is filtered out, and individu- team type (i.e., short-term versus long-term), was ating processes are blocked (Allport, 1954; Brewer, more consistently available across studies. There- 1988). In diverse teams, when the environment fore, we chose to focus on task interdependence primes negative stereotypes against specific demo- and team type as aspects of team-level context. graphic groups, these categorization-based pro- Ultimately, we sought to examine the effects of cesses are likely to influence interactions. In work conceptually distinct contextual variables that to- environments, where stereotypes are less salient, gether provide a comprehensive picture of context individuating information regarding demographi- at the team level. In the subsequent sections, we cally dissimilar individuals is likely to be cogni- discuss in greater detail how these aspects of occu- tively acceptable and result in more positive inter- pational, industry, and team-level context may actions (see Larkey, 1996). Integrating these shape the performance outcomes of both task-ori- perspectives, we propose that in occupational set- ented and relations-oriented team diversity. tings dominated by a single demographic group, it is likely that Stereotypie reactions against underrepresented groups will be triggered (Fiske, 1993; A CONTEXTUAL FRAMEWORK FOR TEAM Reskin et al., 1999). These reactions result in cate- DIVERSITY RESEARCH gorization-based processes that hinder effective group interactions, with negative performance con- Occupational Demography sequences (Larkey, 1996). These categorization- In the United States, workers are classified basedinto processes are less likely in demographically occupational categories based on work performed, balanced occupational settings. skills, education, training, and credentials. Some A rich body of research on the age, race, and occupations are found in just one or two industries gender typing of jobs corroborates our argument (e.g., post office clerks); however, many occupa- that the demographic attributes of a job category or tions are found in a large number of industries occupation (e.g., are associated with job stereotypes that auditors, accountants, software developers, may net- trigger social categorization processes based

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 foshi and Roh 607

on these demographic role of occupational demography in shaping perfor- attributes (Eagly & Steffen, 1984; Perry, mance outcomes 1997). within diverse teams. At the firmFor example, consider the male- dominated level, occupational Frink et al. (2003) found that an inverted category of production engineers. On U-shaped relationshipthe between genderbasis composi- of the research reviewed above, we tionwould and performance was only observedsurmise in gender- that female engineers (an underrepresented balanced occupational settings and notgroup) in male- are likely to be targets of negative stereotypes dominated settings. The authors noted that these (e.g., "women are less techni- cally competent") findings may reflect the inability of organizationsenhancing the salience of gender as a basis for in male-dominated categorization contexts to capitalize on the when men and women work together benefits of gender diversity in (Frink aet al., 2003).production team (Eagly & Stef- fen, 1984; AlthoughEly, the researchers 1994, had not accounted for 1995). We propose that gender diversity thisis contextual likelyfactor in their theoretical frame- to have negative performance effects in work, this hypotheses, or study design,setting. they discussed In a more gender balanced occupational the importance of occupationalsetting, demography in ex- such as postsecondary science teaching, plaining their findings. on the other hand, this research suggests that Cumulatively, the genderresearch perspectives dis- may not emerge as a salient basis for categorization cussed above suggest that in occupational settings within a team of teachers working on dominated organizingby a single demographic group, diverse a school science fair. In this setting, gender teams may face performance losses diversityprimarily for is less likely to have a negative influence on team performance. A similar two reasons. First, these teams may perform subop- argument can be made with regard to the effects of timally because the work context enhances stereo- occupational race/ethnic composition on the per- typing and bias against underrepresented demo- formance outcomes of race/ethnicity-based diver- graphic groups that triggers social categorization sity in teams based on these attributes(Reskin within the teams (Di- et al., 1999). Regarding the effects of occupational age compo- Tomaso, Post, & Parks-Yancy, 2007; Skaggs & Di- sition, a significant body of research suggests that Tomaso, 2004; Van Knippenberg et al., 2004). Sec- negative stereotypes against older workers are ond, in these settings, teams with higher fairly prevalent and can have detrimental conse- proportions of underrepresented group members quences for these workers (Fiske, Cuddy, Glick, & Xu, 2002; (e.g.,Rosen women or ethnic minorities) may& be valued Jerdee, 1977; see Shore and less and receive poorer performance ratings or ac- Goldberg [2005] for a recent review). Although cess to resources, which is likely to impact subjec- older workers may experience more unfavorable tive or objective performance outcomes (Baugh & job outcomes overall, research also indicates that Graen, 1997; Hultin & Szulkin, 1999; Joshi, Liao, & the age composition of a job context can influence Jackson, 2006). In view of these considerations, performance ratings and advancement potential for older employees. we proposed that occupational demographyThis will research suggests that older workers may moderate the relationship face between relations-ori- more unfavorable outcomes in occupations ented diversity with and team performance andfewer tested older workers (Cleveland, Festa, & Montgomery, specific aspects ofthat general relationship meta- 1988; Cleveland, Montgom- ery, & Festa, analytically with the 1984).following: In view of this research, we would propose that among, for example, software programmers Hypothesis la. Occupational gender(an composi- occupation dominated by younger workers), tion moderates the negative effect of team gen-older programmers may be targets of negative der diversity on performance. The negative stereotypes ef- (e.g., viewed as less skilled or motivated); fect of gender diversity on performance is the implications of age diversity-based weakeroutcomes in gender-balanced settings. would thus be negative when older and youngerHypothesis lb. Occupational race/ethnic com- team members collaborate on a joint software position moderates the negative development effect of team project. In work settings with race/ethnica higher diversity on performance. The neg- proportion of older workers (for instance, among ative effect of race/ethnic diversity onwelfare perfor- service workers or post office clerks), mance is weaker in racially/ethnicallythese bal- workers are less likely to face stereotypes and discrimination (Shore & Goldberg, anced settings. 2005). In these settings, categorization-based effects based on ageHypothesis lc. Occupational are age composition less likely to disrupt groups' functioning. moderates the negative effect of team age di- To date, diversity research has not addressed the versity on performance. The negative effect of

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 608 Academy of Management Journal June

age diversity on performance text as a key contingency isthat influencesweaker the rela- in relatively age-balanced tionship settings. between organizational processes/practices and performance outcomes (see Combs, Lieu, So far we have argued that occupational demog- Hall, and Ketchen [2006] for a meta-analysis). In- raphy creates a context that may enhance or mini- dustry-level context has received some, albeit lim- mize categorization-based processes in work ited, attention in past diversity research (e.g., An- groups. Our arguments cona & Caldwell, raise 1992; Lovelacethe etpossibility al, 2001; that occupational demography will moderate the rela- Reagans & Zuckerman, 2001). On the basis of stra- tionship between task-oriented diversity and team tegic management perspectives and in line with our performance as well. We surmise that when a sin- conceptualization of context, we considered gle demographic group dominates an occupation, whether contingencies associated with three spe- relations-oriented diversity may be correlated with cific contexts - service, manufacturing, and high task-oriented diversity. For example, white men technology - may serve as situational enhancers or may be more tenured and have a specific educa- minimizers of diversity effects on performance out- tional background. In these settings, individuals' comes. Together, these three industrial settings in- task-relevant contributions may be confounded corporate a bulk of the research settings considered with their demographic attributes (Berger, Fisek, in past diversity research. Norman, & Zelditch, Service 1977) industries, so defined that as customer-oriented any positive outcomes of task-oriented diversity on performance industries that require front-line customer interac- may also be mitigated. Researchers have also noted tion and engagement, include retail trade, hospital- that when the negative effects of categorization- ity, and education (U.S. Census Bureau, 2002). based diversity are salient, information processing Relative to manufacturing industries, service in- in groups is disrupted (Van Knippenberg et al., dustries are characterized by more frequent and 2004). Some studies have shown that when rela- closer interactions with customers and a greater tions-oriented diversity is associated with greater emphasis on discretionary behavior that can man- conflict, task-oriented diversity is also less likely to ifest directly in performance outcomes such as have positive outcomes (Homan & Van Knippen- sales, customer satisfaction, and customer reten- berg, 2003; Jehn et al., 1999). In view of these per- tion (Datta et al., 2005). Researchers have argued spectives, to account for the possibility that occu- that increasing demographic attribute-based diver- pational demography may have implications for sity can enhance a firm's "market competence," the outcomes of task-oriented diversity, we propose which is a form of competitive advantage in the the following broad hypothesis: service industry (Richard et al., 2007). Consider as Hypothesis Id. Occupational demography an example that a retail store with a diverse group moderates the positive effect of task-oriented of store employees is more likely to attract diverse diversity on team performance. The effects of customers and thereby also more likely to have task-oriented diversity are stronger in more strong store sales. On the other hand, a competing balanced settings. store that is unable to enhance employee diversity to attract customers and increase market share is likely to perform poorly in comparison. On the Industry Setting basis of these considerations, we would expect that Industry settings refer to fairly specific business aspects of service settings such as direct customer environments in which teams are nested that may contact and higher levels of discretionary behaviors have important implications for team diversity dy- would serve as situational enhancers of relations- namics that go over and above the occupational oriented diversity effects on performance. In the effects discussed above (Batt, 2002; Datta, Guthrie, service industry, since these aspects of diversity & Wright, 2005). Porter's (1980) analysis identifies can be considered as a form of market competence, factors such as customers, suppliers, and regulatory we propose that any negative effects of gender, groups that vary by industry and pose varying con- race, and age diversity are likely to be reversed. tingencies for firms. Contingency theory suggests Manufacturing industries are based on the fabri- that these industry factors provide firms with op- cation, processing, or preparation of products from portunities as well as challenges for utilizing key raw materials and commodities. These industries organizational resources to enhance performance are typically highly capital-intensive and include (Burns & Stalker, 1961; Drazin & Van de Ven, 1985; automobile manufacturing, chemical manufactur- Lawrence & Lorsch, 1967; Porter, 1980). Drawing ing, and paper and wood manufacturing (U. S. Cen- on this perspective, a significant body of strategic sus Bureau, 2002). In contrast to service industry management research has identified industry con- firms, manufacturing firms rely more on plant and

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 609 equipment, technology, and raw materials to management teams has examined industrial envi- achieve business goals (Quinn, Anderson, & Finkel- ronment as a relevant influence on the outcomes of stein, 1996). Because manufacturing relies more on diversity (e.g., Hambrick et al., 1996; Hambrick & physical capital and equipment and less on direct Finkelstein, 1987; Keck, 1997), research on lower- customer-based interactions, we would expect di- level teams has rarely taken this contextual factor versity attributes to be less likely to directly impact into account. In general, top management team re- performance in this setting (Richard et al., 2007). search suggests that industry attributes, such as rate Further, Combs and colleagues (Combs, Hall, & of technological change, munificence, and environ- Ketchen, 2006) noted that manufacturing indus- mental uncertainty, can enhance the salience of tries are more likely than service industries to im- diversity in top management teams in relation to plement total quality management techniques, reg- firm performance (see Carpenter, Geletkanycz, and ular training, and formalized human resource (HR) Sanders [2004], and see Hambrick and Finkelstein practices because of the need to monitor employees [1987] for a review). Richard and colleagues (2007) and to develop their knowledge, skill, and ability to examined whether industry (service versus manu- utilize expensive and sometimes dangerous ma- facturing) moderated the relationship between ra- chinery. Consider, as an example, the automobile cial diversity and firm performance in a sample of manufacturing industry, which has implemented a over 800 large U.S. companies; in support of the number of total quality management techniques in- propositions, results indicated that the relationship volving team-based interventions and undertaken between racial diversity and firm performance was regular training for employees. These attributes of stronger in service than in manufacturing firms the automobile industry may serve to buffer any (Richard et al., 2007). direct effects of team diversity. Thus, we surmise On the basis of the theoretical and empirical that aspects of manufacturing settings, such as re- perspective detailed above, we would expect the liance on machinery and HR practices that involve three distinct industrial settings to impose varying greater monitoring of employee behavior, may act contingencies on diversity outcomes. In service set- as situational minimizers of diversity effects on tings, customer-based contingencies may enhance performance outcomes. positive performance outcomes of relations-oriented High-technology industries follow invention and diversity. In manufacturing settings, a reliance on innovation in business strategy and compete in physical equipment and standardized practices to global and short-cycle product-markets (cf. Milkov- monitor employee behavior may minimize any direct ich, 1987). In contrast to service and manufacturing performance outcomes of diversity; and in high- industries, these industries rely more heavily on technology settings, technological and intellectual intellectual capital and invest significantly more in capital-based contingencies may enhance positive research and development (Organisation for Eco- effects of task-oriented diversity. Thus, we propose nomic Cooperation and Development, 2006). In- that industry setting moderates the relationship be- dustries such as information technology, biomedi- tween relations- and task-oriented diversity and cai technology, telecommunication, and data team performance. Specifically: services are included in this category. Researchers have begun to examine this particular setting as a Hypothesis 2a. Relations-oriented diversity is distinct context framing firm practices and out- likely to have a positive effect on performance in comes (Collins & Smith, 2006). This research has service industries. In manufacturing and high- suggested that employers in this setting are also technology settings, relations-oriented diversity likely to adopt commitment-based practices aimed is less likely to have a significant effect on at recruiting and retaining highly skilled employ- performance. ees, which can create a climate fostering knowledge Hypothesis 2b. The positive effect of task-oriented exchange and combination (Collins & Smith, 2006). diversity on performance is stronger in high-tech- These situational contingencies may enhance task- nology industries than in manufacturing and ser- oriented diversity effects. Relative to service and vice settings. manufacturing settings, given contingencies such as rapidly changing technology, reliance on intellectual capital, and a need for creativity and inno- Team-Level Diversity Context vation in a dynamic environment, we would expect task-oriented attributes that form a team's cognitive We propose below that specific team character- resource base to significantly impact performance istics - team interdependence and team type - con- outcomes in this setting. stitute a team-level diversity context that can Although a significant body of research on top enhance or minimize the direct effects of relations-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 610 Academy of Management Journal June oriented and task-oriented tions-oriented diversity diversity and team performance. on team performance. The positive effect of task-oriented diversity on Team interdependence. Teams display varying team performance is stronger among highly levels of interdependence based on their tasks, interdependent teams. The negative effect of goals, and outcomes. Task interdependence is de- relations-oriented diversity on team perfor- fined as the extent to which team members rely on mance is weaker among highly interdependent each other to complete their task (Shea & Guzzo, teams. 1987). In highly task interdependent teams, team Team type. The durability of a team's member- members engage in both sequential and reciprocal ship - that is, whether the team has been assembled exchanges to accomplish the team tasks; in less task to accomplish a short-term goal or whether it is interdependent teams, team members' independent instead a stable and permanent unit in an organi- contributions are aggregated to accomplish the zation - is likely to have consequences for interper- team tasks (Saavedra, Earley, & Van Dyne, 1993; sonal interactions among diverse team members. Thompson, 1967; Van de Ven, Delbecq, & Koenig, Task-related contingencies are likely to differ in 1976). Teams may also vary on goal and outcome short- and long-term teams. In short-term teams, interdependence. Goal interdependence refers to the extent to which a team as a whole has a collec- greater urgency may surround goals and missions. On the other hand, in long-term teams, task require- tive goal. Outcome interdependence refers to the ments may be more stable, and distribution of tasks extent to which team members are interdependent and roles may also be more clearly defined (De in terms of rewards and feedback. These aspects of Dreu & Weingart, 2003). Another dimension along interdependence tend to be highly correlated and which these teams are likely to differ is the longev- therefore researchers have suggested combining ity of team membership. The members of short- them into an overall team interdependence con- term teams likely have shorter tenure, than the struct (Gully, Incalcaterra, Joshi, & Beaubien, 2002). members of long-term teams.3 Thus, temporal dy- Social categorization theory would predict that namics are also likely to vary in these two types of higher outcome and goal interdependence is likely teams. to unite team members to work toward a common Some studies that have examined temporal influ- goal and motivate them to cast aside differences ences on team diversity outcomes have shown that (Gaertner & Dovidio, 2000). Task interdependence the length of time team members spend together may facilitate intergroup contact conducive to re- may diminish the salience of visible aspects of di- ducing categorization-based processes in teams versity and enhance the salience of attitudinal or (Pettigrew, 1998). Team interdependence has re- value-based aspects of diversity (Harrison et al., ceived considerable attention as a moderating in- 1998; Harrison, Price, Gavin, & Florey, 2002). fluence in team diversity research. For example, Schippers and colleagues (2003) found that over a Schippers and colleagues (2003) found that out- longer duration, highly diverse teams were less come interdependence reinforces common group likely to display elaboration-based processes. In the goals that can counteract the negative effects of short run, however, diverse teams engaged in more diversity, so that highly outcome-interdependent task-relevant debates and discussions that had pos- teams with high levels of diversity display more task-related discussions and communication than itive consequences for team performance. The authors noted that in long-standing diverse teams, highly diverse teams with low outcome interdepen- team members may attribute conflicts to relational dence. Jehn and colleagues also reported that de- differences, and the motivation and willingness to mographic diversity was positively associated with satisfaction and commitment when task interde- resolve differences through greater communication may erode over time. In the short term, on the other pendence was high (Jehn et al., 1999). Other re- hand, the members of highly diverse teams are search has also supported this pattern of findings more likely to communicate across differences to (e.g., Van der Vegt & Janssen, 2003). We would accomplish the teams' tasks (Schippers et al., expect that high interdependence creates a context 2003). Corroborating this finding, Watson, Johnson, for elaboration-based processes within a team and thus that the outcomes of task- and relations- and Merrit (1998) also found that demographic di- oriented diversity may be more positive or less negative when the level of interdependence is high. 3 Although team tenure or longevity could also be im- Hence, we propose: portant variables to consider, our review indicated that these variables were sparsely reported in extant research Hypothesis 3. Team interdependence moder- and we were, therefore, unable to include these in the ates the relationship between task- and rela- present study.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 611 versity was O'Reilly, negatively1998) and quantitative (Bowers et al., associated with outcomes over time. 2000;We Horwitz & propose,Horwitz, 2007; Webber & on the basis of this empirical research, Donahue, that2001) reviews. Finally, in in an short-termeffort to teams, urgency sur- rounding the identify relevantaccomplishment unpublished studies, we searched of the tasks may require team ProQuest members Digital Dissertations and conference pro-to overlook differences and aim at utilizing ceedings for the annual meetingsdiversity of the Academy of based on task-relevant dimensions. Since short-term teams are also asso- Management and the Society of Industrial and Or- ciated with shorter team tenure, team members ganizational Psychology for the previous five years. may be less likely to attribute task-based differ- Researchers in related areas were also contacted to ences to deeper attitudinal or personality-based dif- obtain current and unpublished studies that might ferences (Harrison et al., 1998; Schippers et al., fit our criteria for inclusion. 2003). In long-term teams, divisions based on di- Since this study concerned detecting the moder- versity attributes may become more entrenched ating effects of several contextual variables embed- and self-reinforcing, so that conflicts and differ- ded at multiple levels, we focused only on studies ences based on relations-oriented attributes have a that had been conducted in field settings where more debilitating impact on team performance these variables were likely to influence diversity (Schippers et al., 2003; Watson et al., 1998). In view outcomes. We did not include studies that relied on of these considerations, we meta-analytically tested student samples, were conducted in laboratory set- the following: tings, or involved simulated tasks or tasks in artificial environments. Although these types of studies Hypothesis 4. Team type moderates the rela- are of great value in theory development, for the tionship between task- and relations-oriented purposes of this research we were interested in diversity and team performance. The positive identifying contextual moderators influencing di- effect of task-oriented diversity is stronger in versity outcomes in naturally occurring, intact short-term teams than in long-term teams. The work teams in business settings. Research suggests negative effect of relations-oriented diversity is that these types of teams differ substantively from stronger in long-term teams than in short-term teams. laboratory teams (e.g., McGrath, 1984). From an initial set of 95 field studies conducted in organizational settings, we applied the following addi- METHODS tional criteria to select articles for our meta-analy-

Literature Search sis. First, although diversity research has often been conducted at multiple levels of analysis, we We employed multiple search techniques to restricted the present meta-analysis to the team identify prior empirical research that examined the level. For a study to be included in the meta-anal- relationship between work team diversity and per- ysis, both diversity and performance had to be mea- formance. First, we searched the computerized da- sured at the team level. Second, team diversity tabases PsycINFO, ABI/Inform, and Soclndex using variables had to include either relations-oriented keywords such as "team/group diversity," "team/ diversity attributes (race/ethnicity, gender, or age) group composition," "team/group performance," or task-oriented diversity attributes (education, and "team/group effectiveness" as well as search functional background, or organizational tenure) terms associated with specific team diversity at- pertinent to our theoretical arguments. Third, we tributes (e.g., "gender," "race/ethnicity," "age," also excluded studies that examined top manage- "tenure," "education," and "functional background ment teams [n = 47) because outcomes for such diversity"). Second, the electronic search was sup- teams were often measured at the firm level (e.g., as plemented by a manual search of 19 major journals, organizational financial performance), and top including Academy of Management Journal, Ad- management teams were generally considered as ministrative Science Quarterly, Journal of Applied operating under different dynamics than lower- Psychology, Personnel Psychology, Organization level general work teams in organizations (e.g., Science, and Journal of Organizational Behavior, Hambrick & Mason, 1984; Webber & Donahue, and others considered the most highly cited jour- 2001). Furthermore, team effectiveness models also nals in the field of management (see Gomez-Mejia & suggest that performance is a proximal outcome of Balkin, 1992). Third, we also consulted the refer- team composition among general work teams; this ence lists from previous reviews on this topic, in- assumption may not hold among top management cluding both narrative (Harrison & Klein, 2007; teams (Hackman, 1987). Finally, a study had to Jackson et al., 2003; Milliken & Martins, 1996; Van report sample sizes and an appropriate statistic Knippenberg & Schippers, 2007; Williams & (e.g., mean and standard deviation, chi-square, f, F]

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 612 Academy of Management Journal June

that allowed the computation recent reconceptualizations of a (e.g., correlation Harrison & Klein, coefficient with the formula 2007), we were provided unable to incorporate by these Hunter newer and Schmidt (1990: 272). Based on these criteria, the directions in this study. final data set included 8,757 teams from 39 studies Team performance. Measures of team perfor- conducted between 1992 and 2009, yielding a total mance included financial and operational mea- of 117 effect sizes. The data included in the final sures (e.g., sales, productivity), product quality/ analyses represent approximately twice the num- quantity, team innovation, supervisor ratings of ber of effect sizes and three times the number of team performance/effectiveness, and self-ratings teams included in past published meta-analyses on by team members. Most studies provided objective this topic (e.g., Webber & Donahue, 2001). team performance measures or ratings by supervisors. For subjective performance measures, reliabilities of measurement instruments were re- Coding of Studies corded whenever reported. In those cases in which Our initial coding scheme was based on opera- no reliabilities were reported, we took the average tionalization of team characteristics (team interde- reliability of the same variable from all other stud- pendence and team type) and research setting ies. The average reliabilities were .82 for subjective (occupational and industry setting). Using this performance measures. To maintain the statistical scheme, each author and two additional raters in- independence of the data set, when multiple mea- dependently coded a random selection of five arti- sures for team performance were available, we in- cles; initial interrater agreement ranged between 75 cluded only the most objective external measure; and 94 percent. To resolve disagreements, we went when no other source was available, we relied on back to the studies and reached consensus through team members' own assessments of their perfor- discussion. From this discussion, we developed a mance. In a study measuring multiple dimensions detailed set of decision rules and used them to code of team performance (e.g., its quality and quantity), an additional seven articles. Interrater agreement we calculated a composite effect by averaging each was almost 100 percent for these articles. The au- correlation coefficient. thors then coded the remaining articles using the Occupational demography. We relied on the ar- decision rules we had developed. chival data from the Labor Force Statistics of the Current Population Survey (U.S. Bureau of Labor Statistics, 2006) to obtain definitions and statistics Measures of occupational categories. Initially we coded the Team diversity. Drawing on past research (e.g.,research settings (based on sample descriptions) of Jackson et al., 1995; Webber & Donahue, 2001), studies we included in the meta-analysis using Bureau classified diversity attributes into two categories: of Labor Statistics (BLS) definitions; then we ob- relations-oriented (race/ethnicity, gender, and tainedage) the occupational gender, race/ethnicity, and and task-oriented (organizational tenure, educa- age composition data from the BLS data set and tion, and functional background). In the studies assigned this demographic composition to the oc- included in the analyses, categorical diversity cupations at- included in the studies. For example, if a tributes were measured using Blau's (1977) indexstudy's respondents were described as members of or Teachman's (1980) entropy measure; Allison's production teams from multiple electronics manu- (1978) coefficient of variation was used for contin- facturing firms, we obtained data on electrical, uous variables such as age and tenure. Because electronics, and electro-mechanical assemblers' measures based on "faultlines" (hypothetical gender,lines ethnicity, and age from the BLS website splitting a group into attribute-based subgroups and used them to measure occupational demogra- [Lau & Murnighan, 1998: 328; 2005: 645]) werephy [n = 16). For teams that included multiple used less frequently and in some cases combined occupations (e.g., cross-functional product devel- demographic and task-oriented attributes, theseopment teams consisting of engineers and produc- were not included in the present analysis. In tiongen- supervisors), we found demographic data for eral, the operationalization of relations- and task-each occupational category involved in a team and oriented diversity in this study is theoretically then calculated a composite value by averaging the driven and consistent with past research. Our demographic ini- information on all participating occu- tial review also indicated that the diversity pations at- from the BLS data set [n = 8). In aggregating tributes included in this meta-analysis were data the over multiple occupational categories, we most commonly studied variables in past research. were careful not to include studies that incorpo- However, since the measurement of diversity rated at- occupations that were differently skewed de- tributes in studies involved in the sample predates mographically. For example, when a research sam-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 613

pie involved manufacturing. High technologymultiple included semi- occupations in single teams - combining, conductor/electronics, information processing, say, the white-male-dominated occupation telecommunication, and professional of R&D ser- network analyst and the relatively balanced vices. Studies that included samples fromoccupation multi- of insurance under- writer or ple industryHR settings wereprofessional excluded from our (Campion, Papper, & Medsker, analyses1996) [n = 5). - we performed no aggregation and excluded Team interdependence. the Drawing on studyconceptual from this set of analyses. Moreover, when studies were unclear about the definitions of interdependence presented in previ- samples used, had non-U.S. samples for which ous research (e.g., Campion, Medsker, & Higgs, data were not available, or had samples with 1993; Saavedra et al., 1993; Shea & Guzzo, 1987), multiple types of teams, we also excluded them we determined team interdependence using three from the analyses. separate ratings for task, goal, and outcome inter- Among the studies included in these analyses, dependence. Each dimension was rated on a scale the occupational percent female ranged from 20.6 ranging from 1 to 3 (low, moderate, and high) and to 73.4 percent, and the occupational percent eth- averaged to provide a composite score of overall nic minority (i.e., nonwhite) ranged from 15.7 to team interdependence. Low-interdependence teams in- 37.6 percent. If either of these percentages was cluded, for example, production or sales teams below the overall mean for all U.S. occupations with sequentially related activities or individual- (i.e., 46.3 percent for female composition and 29 based goals and rewards; cross-functional R&D percent for minority composition), the occupa- teams whose members frequently exchanged ideas tional setting was characterized as "majority male" and shared common goals were deemed highly or "majority white" for the purposes of categorical interdependent. analyses. Occupational settings with above-average Team type. We also coded team type based on percents female or minority were categorized as the length of time a team was expected to exist "balanced" settings. In our sample, we did not en- (Schippers et al., 2003). Two broad categories, counter enough study settings that could be cate- short-term and long-term, were used. Cross-func- gorized as majority female or majority ethnic mi- tional project teams existing for a limited period nority; therefore, we were unable to test the effects were considered short-term teams, for example. of diversity in these settings. Occupational age Permanent work teams and general work teams ex- composition data were based on the information isting for longer than two years were generally con- provided by the BLS, which offers data on four age sidered long-term. categories by occupation. We considered an occupation as "majority younger worker" if the proportion of older (over 55 years) workers in that occu- Meta- Analytic Techniques pation was lower than the average across occupations (i.e., 18%) and as "balanced" if the We used Hedge and Olkin's (1985) meta-analytic proportion of older workers was above the average procedures to analyze the data. Zero-order correla- across occupations. tions between work team diversity and perfor- Industry setting. Industry was also coded when- mance were taken or calculated from each study ever it was reported and whenever the research and corrected for measurement error. Following setting or sample was described. Following strate- Hunter and Schmidt's (1990) formula, we also cor- gic management research (e.g., Collins & Smith, rected the correlations for unreliability using the 2006; Datta et al., 2005) and drawing from the de- artifact distributions for subjective team perfor- tailed industry descriptions based on the North mance measures. We calculated weighted mean American Industry Classification System (NAICS; correlations by adopting the inverse variance U. S. Census Bureau, 2002), we classified indus- weights and applying Fisher's Z transformation tries examined in previous research into three procedures (Hedge & Olkin, 1985; Lipsey & Wilson, broad categories: service, manufacturing, and 2001). Ninety-five percent confidence intervals high technology. Service industries in the sample were calculated around the sample-weighted corre- included wholesale/retail trade, finance/insur- lation as a measure of accuracy of the effect size ance, health care, educational service, moving/ (Whitener, 1990). The failsafe Jc was also calculated transportation, and government service; manu- to identify the number of "file drawer" (unknown) facturing industries included automobile studies of the same relationship with a true corre- manufacturing, paper and wood manufacturing, lation of zero needed to widen the reported confi- textile manufacturing, oil and gas, chemical dence interval enough to include zero (Orwin, product manufacturing, and general equipment 1983; Rosenthal, 1979).

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 614 Academy of Management Journal June

Heterogeneity of gressioneffect model explains sizes. significant To variability determine in whether effect sizes were consistent over the stud- the correlations of the relationship between team ies reviewed, we tested the homogeneous distri- diversity variables and team performance (Lipsey bution of the effect sizes by calculating the Q- & Wilson, 2001).4 statistic (Hedge & Olkin, 1985). A significant Q indicates the likelihood of moderators that ex- RESULTS plain variability in correlations over studies (Lip- sey & Wilson, 2001). We also examined the po-Main Effects: Work Team Diversity tential impact of outliers by calculating the and Performance sample-adjusted meta-analytic deviancy statistic Using the meta-analytic techniques described suggested by Huffcut and Arthur (1995). Three above, we tested the main effects of work team outliers were identified by use of this statistic, diversity on performance outcomes as well as the but a detailed review of the potential outliers moderating effects of contextual factors embedded revealed no problematic correlations, so these at multiple levels. Table 2 presents the main effect were not eliminated from the analyses. results. We first examined the correlations between Moderator analysis. We conducted detailed all types of diversity and performance and obtained moderator analyses to determine whether multilevel contextual variables were related to the het- a near-zero, nonsignificant result (r = -.01, k = 117, 95% CI = -.02 to .00). This initial result erogeneity of effect sizes (Hedge & Olkin, 1985; corroborated past meta-analytic findings (e.g., Web- Lipsey & Wilson, 2001). The logic of the categorical ber & Donahue, 2001). We then conducted separate model moderator test is analogous to analysis of analyses for relations- and task-oriented diversity variance (ANO VA). Calculating the categorical and found a different pattern of results for each models results in (1) the between-group goodness- type of diversity. For relations-oriented diversity, of-fit statistic QB, which has an approximate chi- we found a very weak negative but significant re- square distribution with p - 1 degrees of freedom, lationship with performance (r = -.03, k = 69, where p is the number of groups, and (2) the with- 95% CI = -.05 to -.02). The relationship between in-group goodness-of-fit statistic Qw, which has an task-oriented diversity and performance was also approximate chi-square distribution with m - 1 very weak but positive and significant (r = .04, k = degrees of freedom, where m is the number of effect 48, 95% CI = .02 to .06). The failsafe k's in Table 2 sizes in the group. That is, Qq is analogous to a suggest that, although the effect sizes are very main effect in an ANOVA, and Qw indicates homo- small, at least 118 (for relations-oriented diversity) geneity within each group in an ANOVA. In the and 73 (for task-oriented diversity) file-drawer null present analysis, as recommended in previous re- effects would need to be reported before those two search on this topic (e.g., Webber & Donahue, relationships would lose statistical support. We 2001), we used a QB statistic to test whether the also conducted additional analyses for each diver- categorical moderator model was statistically sig- sity attribute and found that functional background nificant and then examined each subgroup within diversity was most positively related to perfor- the sample by testing the confidence intervals for mance (r = .13, k = 20, 95% CI = .09 to .17) and statistical significance and by comparing the effect that age diversity showed the most negative perfor- sizes across subgroups whenever possible. For the continuous moderators (i.e., occupational percent female and occupational percent minority), weighted least squares (WLS) regression was also 4 We undertook additional OLS regressions whereby we used, as suggested by Hedge & Olkin (1985). This regressed the effect of diversity-performance on the key procedure involves weighting each observed effect independent variables. Results indicated that the pattern of size by the inverse of its variance, as with formulae findings for OLS regressions mirrors the pattern of findings for weighted means and confidence intervals (Lip- reported here. We also undertook a stepwise regression sey & Wilson, 2001). Using such an approach procedure and found that adding each contextual variable avoids the artificial categorization of continuous was associated with significant incremental variance in the moderating variables. Two indexes assessing the diversity-performance effect. Collectively, in these analyses, contextual variables accounted for 56 percent of the overall fit of the weighted regression model can variance in relations-oriented effect sizes across studies. be calculated: a Q attributable to the regression Further, the variance inflation factors in all cases were well and a Q error or residual (denoted as QR and QE, below 10, which is the rule-of-thumb cutoff for high mul- respectively, and both distributed as a chi- ticollinearity. The full results are available from the authors square). QR is analogous to an F for a regression upon request. We thank our action editor, Jason Colquitt, model and, if significant, indicates that the re- for this suggestion.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 615

TABLE 2 Main Effects: The Relationship between Team Diversity and Performance0

95% Effect Sizes Total Teams Weighted Confidence Failsafe Diversity Type [k) (A/) Meanr Interval 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, .01 Race/ethnicity 22 7,089 -.01 -.04, .01 Age 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 65 Education 9 2,863 -.02 -.06, .01 Tenure 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 results needed to reduce the cumulative effect across studies to the point of nonsignificance (p ^ .05) and is only reported for statistically significant results (p < .05); Q is the effect size heterogeneity statistic indicating the possibility of moderators. **p<.01 mance effect (r = -.06, k = 21, 95% CI = -.09 to phy variables were measured as continuous vari- -.04). ables (i.e., occupational percent female and occu- Table 2 also shows that considerable heterogene- pational percent minority), to further understand ity among effect sizes exists, as indicated by the the patterns of these moderating effects, we also Q-statistic. Both Q values for relations- and task- performed WLS regressions. In support of both oriented diversity are highly significant [p's < .01), hypotheses, we found that occupational percent indicating that correlations vary across studies and female and minority accounted for significant that potential moderators might exist that can ex- amounts of variance in the correlations between plain these correlations. gender and race/ethnicity diversity and team performance (ß = .32, p < .01, R2 = .10, for gender diversity; ß = .37, p < .01, R2 = .14, for race/ Occupational Demography ethnicity diversity). Significant values of QR for Hypotheses la and lb predicted that the negative gender diversity (QR[1] = 16.78, p < .01) and effects of gender and race/ethnicity diversity would race/ethnicity diversity (QR[1] = 22.71, p < .01) be weaker in more gender-balanced and ethnically also indicate that both regression models were balanced settings, respectively. We conducted cat- statistically significant. egorical moderator analyses and contrasted the dif- Hypothesis lc predicted that the negative effects ference between majority male or majority white of age diversity would be strengthened in occupa- occupations and relatively gender-balanced and tions composed of relatively younger workers and ethnically balanced occupations. Table 3 summa- weakened in relatively age-balanced occupations. rizes the results. The categorical model testing the We conducted a categorical analysis and did not moderating effect of occupational percent female find a strong moderating relationship (QB[1] = was highly significant [QB[1] = 39.19, p < .01). As0.85, p > .10) (see Table 3). Although occupations hypothesized, gender diversity had a significant, composed of younger workers displayed slightly negative effect on team performance in majority more negative effects than occupations that were male occupational settings (r = -.09, k = 12, 95% more balanced in terms of age, considering the CI = -.12 to -.05). The effect of gender diversity nonsignificant QB and significantly overlapping was significantly positive in relatively gender bal- confidence intervals, the two occupational groups anced settings (r = .11, k = 7, 95% CI = .06 to .15). were not statistically different. The moderating effect of occupational percent Hypothesis Id proposed that the positive perfor- minority was also significant (QB[1] = 48.65, p mance< outcome of task-oriented diversity would .01). The average correlation was significantly be stronger in more balanced occupational set- negative in majority white occupations (r = -.07, tings. Contrary to this proposition, task-oriented k = 10, 95% CI = -.10 to -.04) and positive in diversity showed more positive performance ef- relatively balanced occupations (r = .11, k = fects6, in majority male and white settings. How- 95% CI = .07 to .14). In addition, because the ever, considering the nonsignificant subgroup re- occupational gender and race/ethnicity demogra- sults for balanced occupational settings across all

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 616 Academy of Management Journal June

TABLE 3 Contextual Influences: Occupational Demography0' b

95% Team Diversity x Occupational Effect Sizes Total Teams Weighted Confidence Failsafe Demography0 [k] (N) Mean r Interval k QB

Gender diversity (Hypothesis la) 39.19* * Majority male settings 12 2,952 -.09 -.12, -.05 20 Balanced settings 7 1,832 .11 .06, .15 12 Race/ethnicity diversity (Hypothesis lb) 48.65** Majority white settings 10 4,071 -.07 -.10, -.04 17 Balanced settings 6 2,584 .11 .07, .14 13 Age diversity (Hypothesis lc) 0.85 Majority younger worker settings 6 4,758 -.08 -.10, -.05 13 Balanced settings 9 1,869 -.05 -.10, -.00 3

Task-oriented diversity (Hypothesis Id) Occupational gender demography 9.67** Majority male settings 22 6,866 .06 .03, .09 39 Balanced settings 14 1,823 -.03 -.08, .02 Occupational race/ethnicity demography 0.60 Majority white settings 32 8,497 .04 .02, .06 39 Balanced settings 4 192 -.02 -.17, .13 Occupational age demography 5.36* Majority younger worker settings 9 5,199 .06 .03, .09 13 Balanced settings 27 3,490 .01 -.03, .04

diversity attributes and the overlapping confi- -.01) and interestingly, relations-oriented diver- dence intervals in general, we were unable inter- sity displayed the strongest negative performance pret this finding meaningfully. effect in high-technology industry settings (r = -.18, k = 21, 95% CI = -.20 to -.15). This finding is also resistant to unpublished null effects with a Industry Setting failsafe k of 152. Hypotheses 2a and 2b proposed that industry Regarding Hypothesis 2b, we found weak sup- setting moderated the relationship between team port for the moderating effects of industry setting diversity and performance outcomes. Table 4 pre- on the performance outcome of task-oriented diver- sents the results. With regard to relations-oriented sity. The overall categorical model was modestly diversity, the overall categorical model testing in- significant [QB[2] = 7.28, p < .05). Task-oriented dustry impact was highly significant (QB[2] = diversity was positively related to performance in 209.89, p < .01). As predicted in Hypothesis 2a, the high-technology industry setting, as hypothe- relations-oriented diversity had a positive effect on sized (r = .06, k = 23, 95% CI = .04 to .09), but this performance in service industries (r = .07, k = 21, effect was only slightly larger than the overall main 95% CI = .05 to .09). Inconsistently with Hypoth- effect of task-oriented diversity (r = .04). We did esis 2a, however, in the manufacturing industry not find any significant support for industry mod- setting, the effect of relations-oriented diversity erating effects on task-oriented diversity effects in was negative (r = -.04, k = 16, 95% CI = -.07 to the manufacturing and service settings.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 617

TABLE 4 Contextual Influences: Industry Setting3

95% Effect Sizes Total Teams Weighted Confidence Failsafe Team Diversity x Industry Setting*1 (k) (AT) Mean r Interval k QB

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

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

Team Interdependence and Team Type the relationship between team diversity and performance. Hypothesis 4 proposed that the negative Table 5 presents the results regarding the effects effects of relations-oriented diversity would be of team-level moderators on the relationship be- strengthened in long-term teams. We found strong tween team diversity and performance. Hypothesis support for relations-oriented diversity: the cate- 3 proposed that effects of task-oriented diversity gorical moderator model was highly significant will be stronger and the effects of relations-oriented (QB[1] = 222.91, p < .01). The performance effect of diversity will be weaker among highly interdepen- relations-oriented diversity was positive in rela- dent teams. The findings with regard to relations- tively short-term teams (r = .09, k = 23, 95% CI = oriented diversity were contrary to this hypothesis. .07 to .12) but became negative in more stable or Although the categorical model for relations-ori- long-term teams (r = -.14; k = 43, 95% CI = -.16 ented diversity was significant (QB[2] = 174.21, to -.12). Both findings are resistant to unpublished p < .01), among teams with low interdependence null effects, with failsafe it's of 89 and 305, respec- relations-oriented diversity was positively related tively. For task-oriented diversity, we did not find a to performance (r = .08), and among teams with statistical support for the hypothesis (QB[1] = 0.63, moderate and high interdependence, relations-ori- p > .10). Although the performance effect of task- ented diversity was negatively related to perfor- oriented diversity was more positive in short-term mance (r = -.12 and r = -.04, respectively). The teams (r = .08) than in long-term teams (r = .04), categorical analysis for task-oriented diversity sup- these subgroup results were not statistically differ- ported the hypothesized pattern, although the over- ent from each other. all moderating model only showed limited statistical support (QB[2] = 7.14, p < .05). The positive performance effect of task-oriented diversity in- DISCUSSION creased as team tasks, goals, and outcomes became more interdependent; however, these results also This meta-analytic review took stock of past re- need to be interpreted cautiously because the 95 search on the diversity-performance relationship percent confidence interval of the low-interdepen- conducted in organizational settings over the past dence subgroup included zero and overlapped with 15 years and examined the sensitivity of this rela- those for the other two subgroups. tionship to contextual variables at multiple levels. Our final hypothesis addressed whether team Our findings revealed that when one considers the type - the length of time a team was expected to relationship between all types of diversity (i.e., col- exist (i.e., long-term versus short-term) - affected lapsing relations- and task-oriented distinction)

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 618 Academy of Management Journal June

TABLE 5 Contextual Influences: Team Interdependence and Team Type*

95% Effect Sizes Total Teams Weighted Confidence Failsafe Team Diversity x Team Context (1c) (iV) Mean r Interval k QB

Team interdependence (Hypothesis 3)h Relations-oriented diversity 174.21** Low interdependence 17 8,051 .08 .06, .10 56 Moderate interdependence 38 10,770 -.12 -.14, -.10 238 High interdependence 14 958 -.04 -.11, .03 Task-oriented diversity 7.14* Low interdependence 5 557 -.03 -.11, .06 Moderate interdependence 24 7,604 .04 .02, .06 23 High interdependence 19 1,668 .10 .05, .15 23

Team type (Hypothesis 4f Relations-oriented diversity 222.91 * * Short-term teams 23 7,733 .09 .07, .12 89 Long-term teams 43 9,730 -.14 -.16, -.12 305 Task-oriented diversity 0.63 Short-term teams 13 684 .08 -.01, .16 Long-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 results needed to reduce the cumulative effect across studies to the point of nonsignificance (p ^ .05) and is only reported for statistically significant results (p < .05); QB is the between-group heterogeneity statistic indicating the statistical significance of the categorical moderator 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 ideas and shared common goals. c Short-term teams involved, for example, project-based teams existing for a limited period of time; permanent work teams or general work teams existing for longer than two years were generally considered relatively long-term teams. * p < .05 **p< .01 and performance, the direct effect of diversity on discuss the theoretical, empirical, and practical im- performance is essentially zero. Relations-oriented plications of our findings. diversity attributes such as gender, race/ethnicity, and age had very small, though significantly nega- The Role of Occupational Demography and tive, effects on team performance. Although func- Industry Setting as Context tional diversity had a more substantial positive effect, other forms of task-oriented diversity (i.e., We found that in occupations dominated by male education and tenure) also had very small effects or by white employees, gender and ethnic diversity on team performance. These findings make it ap- had more negative effects on performance out- pear that diversity does not really matter for team comes. These findings draw attention to the impor- performance. However, we contend that these weak tance of extraorganizational context in shaping di- direct relationships may be obscuring the specific versity outcomes, and we call for a more detailed conditions under which diversity can have benefi- and comprehensive acknowledgement of macro- cial or detrimental effects on performance out- level context in future research. Apart from stereo- comes. Our study developed and tested a frame- types associated with underrepresented groups in a work identifying boundary conditions under which particular occupational context, implicit status dif- the diversity-performance relationship is likely to ferences between demographic groups may also be be significant. Specifically, we found that after we a mechanism by which contextual factors such as accounted for moderating variables at multiple lev- occupational demography filter into team-level in- els, diversity effects doubled or tripled in size. Fur- teractions (see Ridgeway, 1991, 1997). The domi- ther, industry and occupational moderators, which nance of a particular demographic group within a have received relatively scant attention in past re- particular occupational setting can signal greater search, explained significant variance in relations- access to resources and privilege for this group. The oriented diversity effects across studies. Below we privilege accrued by a demographic group in a par-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 619 ticular context management efforts to addressmay gender-, ethnicity-, provide team members belong- ing to this and age-based group issues in the workplace.5 withSpecific an "expertise advantage"; these group forms of training members intended at changing behaviors may be considered more competent, targeted and at demographically the dissimilar employeescontributions of the members of low-status anddemographic customers may be implemented in these set- groups may be devalued in these settings tings. In contrast, in(Berger, manufacturing settings, where Ridgeway, Fisek, & Norman, 1998; Berger, we found negative Ridgeway, effects, firms do not face similar & Zelditch, 2002). Indeed, a rich body pressures and mayof be less likely tosociopsychological directly address research has shown that, these aspects in of team diversity.small groups, interpersonal interactions tend Surprisingly,to wereplicate found that in high-technology larger societal status differentials (Berger settings, relations-oriented et diversity al., had a more 1998, 2002) and may account for the suboptimal substantial negative effect. Recently, DiTomaso performance and outcomes of gender-diverse colleagues and found that, in 24 firmsethnically that fit our defi- diverse teams in majority male and nition of high-technology white settings, white mensettings. re- We propose ceived morethat training, mentoring, aand coachingcloser and examination of the occupational also thecontext most favorable performance evaluations, in which teams are embedded can provide relative to researchersany other demographic group (DiTo- with an understanding of broader structural maso, Post, Smith, Farris, & Cordero, 2007). Theseinequality and its implications for team-level characteristics of high-technology diversity firms may en- outcomes. Further, although we hance were ethnicity- and gender-based not status differen- able to test the role of organizational practices tials. It would appear that white men in occupy a these settings, research suggests that high-status these position (reflected in statusthese favorable differentials are further legitimized employment by outcomes) in organizationalthe high-technology sec- practices that favor some demographic tor (DiTomaso, Post, Smith et al., 2007) and wouldgroups over others (Ridgeway, 1991, 1997). enjoy an "expertiseThus, advantage" relative toanother women direction for future research could or minorities. Therefore, bethe status-based to processes examine the role that organizational practicesthat we discussed earlier could also account for the play in either enhancing or mitigating negative effectsstatus-based of relations-oriented diversity in processes in a specific occupational this industry. Furthermore,context. like manufacturing Although firms,statistically high-technology firms may not face pres- inconclusive, our finding was that task-oriented sures to proactively adopt diversity management diversity had weakly positive effects practices. These in findings call male/whitefor a finer-grained dominated settings. This pattern understanding of theis industry contrarycontext framing di- to our theoretical arguments. Although versity-based outcomes. For example, we future re- caution against overinterpret- ing this finding, search could apply contingency-based we perspectives acknowledge that in these more homogeneous and examine whether settings, a fit between firms' diversity categorization based on demographic management characteristics strategies and industry-level contin- may not have been suf- ficient to mitigate gencies explains the outcomes of diversity the at the elaborative potential of task- oriented diversity. team level. Longitudinal studies that examine howWe call for more research on possible mechanisms specific industry trends influence firms' responses mediating task-oriented diversity effects to diversity and their implications in for teamthese perfor- occupational settings. Our findings mance could be an indicatedavenue for future inquiry. that the industry settings in which teams were embedded also had interest- We conclude this section by noting that our finding implications for team-level diversity-based out- ings challenge the assumption, born from social- comes. We found that relations-oriented diversity categorization theory, that some aspects of diver- had positive effects in service industry settings and sity necessarily have detrimental effects on team slightly negative effects in manufacturing settings. performance. Although relations-oriented diversity In addition to the market competence perspective attributes have been typically associated with cat- (Richard et al., 2007) discussed in developing our egorization-based processes (Jackson et al., 1995; hypotheses, some additional considerations should also perhaps be taken into account. Service settings (for example, retail establishments and restaurants) 5 Consider as an example the negative publicity re- involve front-line customer contact, and the costs ceived by Denny's Restaurants as a result of discrimina- of interactions based on negative categorizations tion charges and their subsequent efforts to train and are high in this context. Hence, firms embedded in reward appropriate behaviors toward demographically these industries may engage in proactive diversity dissimilar employees and customers.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 620 Academy of Management Journal June

Jehn et al., 1999), we andfound collaborate withthat each other. the Although effects the ef- of these attributes on team performance fect sizes were smaller, a lower ranged level of interdepen- from significantly negative (in thedence was high-technologyassociated with positive performance industry and in male/white dominated outcomes for relations-oriented occupations) diversity. In teams to significantly positive (in with lowthe interdependence, service the distinct industry nature of and in gender-/ethnicity-balanced each individual team member'soccupations). task and a lower The effects of task-oriented need diversity, to exchange information however, may create a situa- were in- variant across these contexts. tion in which team members Current are not frustrated with applications of social identity theory dissimilar group membersor socialand are able to recognizecategorization theory in diversity research the contributions ofare other members insufficient to the team for explaining these findings. goals (Pelled These et al., 1999). theoretical frame- works were originally In developed support of our propositions, to explainfindings indi- the outcomes of differences catedin that minimal in short-term teams group (such as temporary settings and may not capture the project effects teams), relations-oriented of demographic diversity had a attributes among naturally positive effect occurring on performance. In thesework teams, the groups in these contexts (see also DiTomaso, Post, Smith, expectation of a finite amount of time to complete Farris, & Cordero, 2007; Linnehan & Konrad, 1999). tasks may force group members to exploit the full Therefore, these findings suggest that a priority for elaborative potential of diversity attributes. In long- future research is to investigate how the demogra- term teams, team members may feel less compelled phy of a work context and the characteristics of the to identify and utilize diverse perspectives and industry settings in which teams are embedded can may also have more opportunity to bring up con- influence the performance outcomes of gender or flicts that may be detrimental to group functioning. ethnic diversity. We call for more context-focused We note that we had limited ability to test temporal diversity research, which would entail targeting influences on diversity effects more directly in this particular settings in which either the negative or meta-analysis. Measures of team tenure or longev- the positive consequences of team diversity are ity may yield a different pattern of findings (Harri- more likely. Such efforts could facilitate the devel- son et al., 2001; Schippers et al., 2003). opment of new theoretical approaches that incor- Overall, it is interesting to note that these team- porate the status-based or strategic management level moderators, which have received the most perspectives we outlined above. attention in past research, had weaker moderating effects on diversity outcomes than industry and Team-Level Contextual Influences occupational variables. Moreover, these variables had a more significant moderating effect on rela- Our review of past literature suggested that there has been a growing emphasis on the role of moder- tions-oriented diversity effects than on task-ori- ators in diversity research and that, predominantly, ented diversity effects. In general, we found that past research has focused on team-level moderators moderately interdependent teams and long-term such as team tenure, task interdependence, and teams could face challenges with increasing diver- task complexity (see Van Knippenberg & Schippers sity; less interdependent teams and short-term [2007] for a review). In view of this research, we teams would benefit from greater demographic di- proposed that contextual factors such as team in- versity among team members. These findings sug- terdependence and team type would moderate the gest that more research on the psychological pro- effects of diversity on performance in such a way cesses underlying these moderating effects is that elaboration-based processes would be more needed in these specific team contexts. For exam- likely to be observed in more interdependent or ple, understanding whether role stress or role over- shorter-lived teams (Bowers et al., 2000; Jehn et al., load associated with greater team interdependence 1999). Contrary to our predictions, we found that has implications for the manner in which team relations-oriented diversity had the most negative members deal with demographic differences would effects in moderately interdependent teams. These be useful. These team-level moderators may also findings suggest that the interactive effects of team need to be considered in conjunction with other diversity and interdependence may be more com- macro-level contextual variables. For example, it plex than acknowledged in the past. Moderately would be interesting to understand whether task- interdependent tasks may impose constraints that related contingencies are more likely to induce interfere with the elaborative potential of diversity conflict in demographically diverse groups when attributes but are not so demanding that team mem- the macrolevel context enhances categorization- bers must overcome categorization-based processes based processes.

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 621

Caveats, Limitations, note that, although conceptually interesting, field and Future Directions research directly applying faultlines-based mea- The meta-analytic approach we applied in this sures is not abundant and has yielded inconsistent study has some limitations. One concern may be results (e.g., Gibson & Vermeulen, 2003; Greer & that the sample is small and may be associated with Jehn, 2008; Lau & Murnighan, 2005; Li & Hambrick, second-order sampling error, which is particularly 2005). We propose that future research also exam- relevant for moderator analysis (Hunter & Schmidt, ine how contextual moderators may enhance the 1990). However, the failsafe Jc's we report are fairly strength and number of faultlines in teams to re- large and suggest that, even at the rate at which solve these mixed findings. diversity research is being conducted, at least a Harrison and Klein (2007) offered several guide- decade more of research may be required to over- lines for the measurement of diversity in terms of turn the significant findings reported here. three dimensions: separation, variety, and dispar- We note that we were also limited in our ability ity. These authors proposed that the measurement to code contextual variables, as the studies' de- of various diversity attributes should be based on scriptions of research settings and team tasks were the specific dimension of diversity considered. In often scanty. For example, despite the increasing this meta-analytic review, because we relied on attention to temporal effects on team diversity out- measurement of diversity that predated this 2007 comes, the lack of information regarding team ten- article, we were unable to adopt these guidelines. ure or team duration was especially troubling. We However, complementing Harrison and Klein's propose that future studies provide more detailed (2007) arguments, we note that the malleability of descriptions of research settings and acknowledge relations-oriented diversity to occupational, indus- the role that these factors play in explaining re- try, and team-level contextual variables under- search findings. Because of theoretical consider- scores the importance of clarifying the specific ations and data availability, we considered only a dimension associated with these demographic at- subset of the possible moderating influences on tributes in the future. A comprehensive contextual team diversity outcomes. Missing from this analy- analysis may also help researchers determine sis is the role that organizational context and other whether a specific demographic attribute is likely extraorganizational factors can play in shaping di- to manifest as "separation," "variety," or "dispar- versity outcomes. Aspects of organizational context ity" in teams. For example, diversity could be con- such as managerial demography, climate, culture, ceptualized as variety in short-term teams, in ser- and leadership also merit closer scrutiny in future vice settings, and in demographically balanced research. Other extraorganizational factors, such as settings but as disparity in long-term teams, in societal and political events (e.g., immigration male/white dominated occupations, and in high- trends, passage of key legislation), may also be im- technology settings. portant to consider in the future but were outside the scope of this study. Further, we note that for Implications for Diversity Management testing the effects of occupational demography, we could only include U.S.-based samples. These find- A context-based approach to workplace diversity ings may vary in other cultural contexts. research can potentially provide practical insights Finally, we used a simplified relations-oriented that might enhance the effectiveness of diversity versus task-oriented typology while considering management practices. In some work contexts, it is contextual effects on diversity outcomes; this important to recognize that diversity at the work dichotomy corresponded to the categorization- group level may be problematic. In these settings, versus elaboration-based processes underlying di- diversity interventions aimed directly at targeting versity effects. Other typologies, such as surface- behaviors driven by stereotyping and bias against versus deep-level diversity, may also be pertinent underrepresented groups and ensuring representa- to this analysis and should be considered in the tion at higher levels may be necessary to reverse future. Contextual effects on the outcomes of deep- negative diversity-based outcomes. Furthermore, or- level diversity variables, such as personality, cog- ganizations can directly address the dominance of a nitive ability, values, and attitudes, were also out- demographic majority in a particular labor market side the scope of the present study and may be of by proactively implementing partnerships with ed- interest in future research. We were also unable to ucational institutions to increase gender- and race- incorporate the faultlines approach that has been a based diversity in the applicant pool. These obser- growing focus in diversity research. Of the studies vations may be particularly important in the case of we initially reviewed, 5 percent applied a fault- the high-technology settings included in our anal- lines-based operationalization of diversity. We also ysis. At a time when this sector of the economy is

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 622 Academy of Management Journal June growing rapidly, there past; handicaps is future also theory growing building; and limits concern among policy makers researchers' regarding ability to distill the the practical declining implica- ratio of women and minorities tions of findings. in this This study industry presents a roadmap for (Informa- tion Technology Association context-focused research of that weAmerica, hope encourages 2003). The negative performance researchers to effects more carefully account of fordiversity context that we found in high technology in future studies, facilitates may greater also theoretical reflect an "unfriendly" atmosphere integration of forthe macro womenand micro levels of andanal- minori- ties in this setting. Partnerships ysis, and paves the way for with new theoretical high and schools and college campuses methodologicalto increase developments. gender and ethnic diversity in mathematics, science, and engineering could enhance the diversity of the applicant pool and provide a more balanced REFERENCES representation of various demographic groups in firms, thereby mitigat- *Ainoya, N. 2004. Demographic diversity, team pro- ing the problematic consequences cess, and team performance: Assessingof demographic moderator diversity at the work effects group of cognitive level. conflict management We practices note that a few prominent firms inand taskthis interdependence. industry Unpublished are doctoral already making these proactive dissertation, efforts, University of Southern and California. our findings underscore the value of these initiatives. Allison, P. D. 1978. Measures of inequality. American With respect to our findings regarding the effects Sociological Review, 43: 865-880. of team-level moderators, we propose that diversity Allport, G. 1954. The nature of prejudice. Cambridge, management practices need to specifically address MA: Addison-Wesley. the level of interdependence or longevity that team members will encounter. Teams performing more *Ancona, D. C, & Caldwell, D. F. 1992. Demography and interdependent tasks over the long term may need design: Predictors of new product team performance. Organization Science, 3: 321-341. ongoing training interventions or team coaching that facilitates group decision making and con- *Balkundi, P., Kilduff, M., Barsness, Z. L., & Michael, flict resolution. A more tailored team-focused J. D. 2007. Demographic antecedents and perfor- approach to designing diversity management mance consequences of structural holes in work teams. Journal of Organizational Behavior, 28: practices may enhance the relevance and effec- 241-260. tiveness of these practices in teams. In general, a "contextual diagnosis" will allow firms to Bamberger,de- P. 2008. Beyond contextualization: Using velop diversity management practices that are context theories to narrow the micro-macro gap in management research. Academy of Management tailored to reduce categorization-based processes Journal, 51: 839-846. and enhance elaboration-based processes at the team level. Overall, long-term teams that perform Batt, R. 2002. Managing customer service: Human re- interdependent tasks in less demographically source practices, quit rates, and sales growth. Acad- emy of Management Journal, 45: 587-597. balanced occupational settings may be the most vulnerable to categorization-based processes and*Baugh, S. G., & Graen, G. B. 1997. Effects of team gender may need more direct and focused interventions and racial composition on perceptions of team performance in cross-functional teams. Group and Or- than other types of teams. ganization Management, 22: 366-383. Berger, J., Fisek, M. H., Norman, R. Z., & Zelditch, M. Conclusion 1977. Status characteristics in social interactions: Past reviews have labeled differing theoretical An expectations states approach. New York: Elsevier. perspectives and contradictory findings in diversity research as the "double-edged sword of diver- Berger, J., Ridgeway, C, Fisek, M. H., & Norman, R. Z. sity" (e.g., Webber & Donahue, 2001). In this article, 1998. The legitimation and delegitimation of power we propose a new agenda for diversity research - and prestige orders. American Sociological Review, 63: 379-405. one that moves beyond a debate regarding the potential benefits or costs of diversity and highlights Berger, J., Ridgeway, C, & Zelditch, M. 2002. The construction of status and referential structures. Socio- the inherent context dependence of diversity effects in organizations. We note that inadequately logical Theory, 20: 157-179. reporting and acknowledging context not only ob- Blau, P. M. 1977. Inequality and heterogeneity: A prim- scures the important consequences of diversity in itive theory of social structure. New York: Free Press. organizations, but also hampers efforts to synthe- size and integrate the cumulative evidence from the Bourgeois, L. J. 1985. Strategic goals, perceived uncer-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 foshi and Roh 623

tainty, and Groupeconomic composition and job perceptions: Impact onperformance in volatile environments. decisionAcademy about sustainability of applicants for of Management Journal, 28: 548-573. group membership. Unpublished manuscript, Ba- Bowers, C. A., Pharmer, J. A., & Salas, E. 2000. ruch When College, City University of New York. member homogeneity is needed in work teams:Collins, C. A J., & Smith, K. G. 2006. Knowledge exchange meta-analysis. Small Group Research, 31: 305-327. and combination: The role of human resource prac- Brewer, M. B. 1988. A dual process model of impression tices in the performance of high-technology firms. formation. In T. K. Srull & R. S. Wyer (Eds.), Academy Ad- of Management Journal, 49: 544-560. vances in social cognition, vol. 1: 1-36. Hillsdale, *Colquitt, J. A., Noe, R. A., & Jackson, C. L. 2002. Justice NJ: Erlbaum. in teams: Antecedents and consequences of proce- *Bunderson, J. S. 2003. Recognizing and utilizing exper- dural justice climate. Personnel Psychology, 55: 83- 109. tise in work groups: A status characteristics perspective. Administrative Science Quarterly, 48: 557- Combs, J., Liu, Y., Hall, A., & Ketchen, D. 2006. How 591. much do high-performance work practices matter? A Burns, T., & Stalker, G. M. 1961. The management of meta-analysis of their effects on organizational per- innovation. London: Tavistock. formance. Personnel Psychology, 59: 501-528.

*Cady, S. H., & Valentine, J. 1999. Team innovation and Datta, D. K., Guthrie, J. P., & Wright, P. M. 2005. Human perceptions of consideration: What difference does resource management and labor productivity: Does diversity make? Small Group Research, 30: 730- industry matter? Academy of Management Journal, 48: 135-145. 750.

*Campion, M. A., Medsker, G. J., & Higgs, A. C. 1993. De Dreu, C. K. W., & Weingart, L. R. 2003. Task versus Relations between work group characteristics and relationship conflict, team performance, and team effectiveness: Implications for designing effective member satisfaction: A meta-analysis. Journal of work groups. Personnel Psychology, 46: 823-850. Applied Psychology, 88: 741-749. *Campion, M. A., Papper, E. M., & Medsker, G. J. 1996. Devine, D. J., & Phillips, J. L. 2001. Do smarter teams do Relations between work team characteristics and ef- better? A meta-analysis of cognitive ability and team fectiveness: A replication and extension. Personnel performance. Small Group Research, 32: 507-532. Psychology, 49: 429-452. DiTomaso, N., Post, C, & Parks-Yancy, R. 2007. Work- Cappelli, P., & Sherer, P. 1991. The missing role of con- force diversity and inequality: Power, status, and text in OB: The need for a meso-level approach. In numbers. Annual Review of Sociology, 33: 473- 501. L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior, vol.13: 55-110. Green- DiTomaso, N., Post, C, Smith, D. R., Farris, G. F., & wich, CT: JAI Press. Cordero, R. 2007. Effects of structural position on allocation and evaluation decisions for scientists Carpenter, M. A., Geletkanycz, M. A., & Sanders, W. G. 2004. Upper echelons research revisited: Anteced- and engineers in industrial R&D. Administrative ents, elements, and consequences of top manage- Science Quarterly, 52: 175-207. ment team composition. Journal of Management, *Drach-Zahavy, A., & Somech, A. 2002. Team heteroge- 30: 749-778. neity and its relationship with team support and Chatman, J. A., & Flynn, F. 2001. The influence of team demo- effectiveness. Journal of Educational Admin- istration, 40: 44-66. graphic heterogeneity on the emergence and consequences of cooperative norms in work teams. Drazin, Acad- R., & Van de Ven, A. 1985. Alternative forms of fit emy of Management Journal, 44: 956-974. in contingency theory. Administrative Science *Choi, J. N. 2007. Group composition and employee Quarterly, 30: 514-539. creative behavior in a Korean electronics company: Dubin, R. 1976. Theory building in applied areas. In Distinct effects of relational demography and M.group D. Dunnette (Ed.), Handbook of industrial and diversity. Journal of Occupational and Organiza- organizational psychology: 17-39. Chicago: Rand tional Psychology, 80: 213-234. McNally. Choi, J. N., Price, R. H., & Vinokur, A. D. 2003. Eagly, A. H. 1995. The science and politics of comparing Self-efficacy changes in groups: Effects of diversity, women and men. American Psychologist, 50: 145- leadership, and group climate. Journal of Organiza- 158. tional Rehavior, 24: 357-372. Eagly, A. H., & Steffen, V. J. 1984. Gender stereotypes Cleveland, J. N., Festa, R. M., & Montgomery, L. 1988. stem from the distribution of women and men into Applicant pool composition and job perceptions: social roles. Journal of Personality and Social Impact on decisions regarding an older applicant. Psychology, 46: 735-754. Journal of Vocational Rehavior, 32: 112-125. Ely, R. J. 1994. The effects of organizational demograph- Cleveland, J. N., Montgomery, L., & Festa, R. M. 1984. ics and social identity on relationships among pro-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 624 Academy of Management Journal June

fessional women. Administrative Science Quar- Hambrick, D. C, & Finkelstein, S. 1987. Managerial dis- terly, 39: 203-238. cretion: A bridge between polar views on organiza- Ely, R. J. 1995. The power in demography: Women's tions. In L. L. Cummings & B. M. Staw (Eds.), Re- social constructions of gender identity at work. searchin organizational behavior, vol. 9: 369-406. Academy of Management Journal, 38: 589-634. Greenwich, CT: JAI Press. *Ely, R. J. 2004. A field of group diversity, participation Hambrick, D. C, & Mason, P. A. 1984. Upper echelons: in diversity education programs, and performance. The organization as a reflection of its top managers. Journal of Organizational Behavior, 25: 755-780. Academy of Management Review, 9: 193-206. Harrison, D. A., & Klein, K. J. 2007. What's the differ- Fiske, S. T. 1993. Controlling other people: The impact of power on stereotyping. American Psychologist, 48: ence? Diversity constructs as separation, variety, or 621-628. disparity in organizations. Academy of Manage- ment Review, 32: 1199-1228. Fiske, S. T. 1998. Stereotyping, prejudice, and discrimination. In D. T. Gilbert, S. T. Fiske, & G. Harrison,Lindzey D. A., Price, K. H., & Bell, M. P. 1998. Beyond (Eds.), Handbook of social psychology: 357-411. relational demography: Time and the effects of sur- Boston: McGraw-Hill. face- and deep-level diversity on work group cohesion. Academy of Management Journal, 41: 96- Fiske, S. T., Cuddy, A. J. C, Glick, P., & Xu, J. 2002. A 107. model of (often mixed) stereotype content: Compe- tence and warmth respectively follow from per- Harrison, D. A., Price, K. H., Gavin, J. H., & Florey, A. T. ceived status and competition. Journal of Personal- 2002. Time, teams, and task performance: Changing ity and Social Psychology, 82: 878-902. effects of surface- and deep-level diversity on group functioning. Academy of Management Journal, 45: Frink, D. D., Robinson, R. K., Reithel, B., Arthur, M. M., 1029-1045. Ferris, G. R., Kaplan, D. M. et al. 2003. Gender demography and organizational performance: A two Hedge, L. V., & Olkin, I. 1985. Statistical methods for study investigation with convergence. Group and meta-analysis. Orlando, FL: Academic Press. Organization Management, 28: 127-147. *Herron, M. T. 1993. The effects of the ethnic and Gaertner, S. L., & Dovidio, J. F. 2000. Reducing inter- gender diversity of work team on the perceptions group bias: The common intergroup identity of performance outputs. Unpublished doctoral dis- model Philadelphia: Psychology Press. sertation, California School of Professional Psychol- ogy, Los Angeles. Gibson, C, & Vermeulen, F. 2003. A healthy divide: Subgroups as a stimulus for team learning behavior. Hilton, J. J., & von Hippel, W. 1996. Stereotypes. In J. T. Administrative Science Quarterly, 48: 202-239. Spence, J. M. Darley, & D. J. Foss (Eds.), Annual review of psychology, vol. 47: 237-271. Palo Alto, Gomez-Mejia, L., & Balkin, D. 1992. Determinants of fac- CA: Annual Reviews. ulty pay: An agency theory perspective. Academy of Management Journal, 35: 921-955. Homan, A. C, & Van Knippenberg, D. 2003. The benefi- cial effects of cross-categorizing informational and Greer, L. L., & Jehn, K. A. 2008. Where perception meets demographical diversity in groups. Paper pre- reality: The effects of different types of faultline sented at the 11th European Congress of Work and perceptions, asymmetries, and realities on inter- Organizational Psychology, Lisbon. subgroup conflict and workgroup functioning. Working paper, Leiden University, The Netherlands. Horwitz, S. K., & Horwitz, I. B. 2007. The effects of team diversity on team outcomes: A meta-analytic review Gully, S. M., Incalcaterra, K., Joshi, A., & Beaubien, J. M. of team demography. Journal of Management, 33: 2002. A meta-analysis of team efficacy, potency and 967-1015. performance: Interdependence and level of analysis as moderators of observed relationships. Journal of Huffcutt, A. L, & Arthur, W., Jr. 1995. Development of a Applied Psychology, 87: 819-832. new outlier statistic for meta-analytic data. Journal of Applied Psychology, 80: 327-334. Hackman, J. R. 1987. The design of work teams. In J. W. Lorsch (Ed.), Handbook of organizational behav- Hunter, J. E., & Schmidt, F. L. 1990. Methods of meta- ior: 315-342. Englewood Cliffs, NJ: Prentice-Hall. analysis: Correcting error and bias in research Haleblian, J., & Finkelstein, S. 1993. Top management findings. Newbury Park, CA: Sage. team size, CEO dominance, and firm performance: Hultin, M., & Szulkin, R. 1999. Wages and unequal access The moderating role of environmental turbulence to organizational power: An empirical test of gender and discretion. Academy of Management Journal, discrimination. Administrative Science Quarterly, 36: 844-863. 44: 453-472.

Hambrick, D. C, Cho, T. S., & Chen, M. 1996. The Ilgen, influ- D. R., Hollenbeck, J. R., Johnson, M., & Jundt, D. ence of top management team heterogeneity 2005. on Teams in organization: From input-process- firms' competitive move. Administrative Science output models to IMOI models. In S. Fiske, D. L. Quarterly, 41: 659-684. Schacter, & A. Kasdin (Eds.), Annual review ofpsy-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 625

chology, vol. ential effects 56: by environmental 517-543. context. Organiza- Palo Alto, CA: Annual Reviews. tion Science, 8: 143-156. Information Technology Association of America. 2003. *Keller, R. T. 2001. Cross-functional project groups in Report of the Information Technology Association research and new product development: Diversity, of America (ITAA) blue ribbon panel on IT diver- communications, job stress and outcomes. Academy sity. Presented at the National IT Workforce Convo- of Management Journal, 44: 547-555. cation, Arlington, VA. *Kirkman, B. L., Tesluk, P. E., & Rosen, B. 2004. The *Jackson, S. E., & Joshi, A. 2004. Diversity in social impact of demographic heterogeneity and team lead- context: A multi-attribute, multi-level analysis of er-team member demographic fit on team empower- team diversity and sales performance. Journal of ment and effectiveness. Group and Organization Organizational Behavior, 25: 675-702. Management, 29: 334-368. Jackson, S. E., Joshi, A., & Erhardt, N. L. 2003. Recent Kochan, T., Bezrukova, K., Ely, R., Jackson, S., Joshi, A., research on team and organizational diversity: Jehn, K. et al. 2003. The effects of diversity on busi- SWOT analysis and implications. Journal of Man- ness performance: Report of the diversity research agement, 29: 801-830. network. Human Resource Management, 42: 3-21. Jackson, S. E., May, K. E., & Whitney, K. 1995. Under the Larkey, L. K. 1996. Toward a theory of communicative dynamics of diversity in decision-making teams. In interactions in culturally diverse workgroups. Acad- R. A. Guzzo & E. Salas (Eds.), Team effectiveness emy of Management Review, 21: 463-491. and decision making in organizations: 204-261. Lau, D. C, & Murnighan, J. K. 1998. Demographic diver- San Francisco: Jossey-Bass. sity and faultlines: The compositional dynamics of *Jarrell, K. A. 2002. Job-related diversity and service organizational groups. Academy of Management Review, 23: 325-340. team outcomes: New insights into the roles of task structure and process conflict Unpublished doc- Lau, D. C, & Murnighan, J. K. 2005. Interactions within toral dissertation, Syracuse University. groups and subgroups: The effects of demographic faultlines. Academy of Management Journal, 48: *Jehn, K. A., & Bezrukova, K. 2004. A field study of group 645-659. diversity, workgroup context, and performance. Journal of Organizational Behavior, 25: 703-729. Lawrence, P., & Lorsch, J. 1967. Organization and its environment. Cambridge, MA: Harvard University *Jehn, K. A., Northcraft, G. B., & Neale, M. A. 1999.Why Press. differences make a difference: A field study of diversity, conflict, and performance in workgroups. *Leonard, J. S., Levine, D. I., & Joshi, A. 2004. Do birds of Administrative Science Quarterly, 44: 741-763. a feather shop together? The effects on performance of employees' similarity with one another and with Johns, G. 2006. The essential impact of context on organ- customers. Journal of Organizational Behavior, 25: izational behavior. Academy of Management Re- 731-754. view, 31: 386-408. LePine, J. A., Hanson, M. A., Borman, W. C, & Motow- *Joshi, A. 2002. How does context matter? Examining idlo, S. J. 2000. Contextual performance and team- the process and performance outcomes of work work: Implications for staffing. In G. R. Ferris & K. M. team heterogeneity. Unpublished doctoral disserta- Rowland (Eds.), Research in personnel and human tion, Rutgers University, New Brunswick, NJ. resources management, vol. 19: 53-90. Greenwich, Joshi, A., Liao, H., & Jackson, S. E. 2006. Cross-level CT: JAI Press. effects of workplace diversity on sales performance Li, J., & Hambrick, D. C. 2005. Factional groups: A new and pay. Academy of Management Journal, 49: vantage on demographic faultlines, conflict, and dis- 459-481. integration in work teams. Academy of Manage- Jost, J. T., & Banaji, M. R. 1994. The role of stereotyping mentjournal, 48: 794-813. in system-justification and the production Linnehan, of false F., & Konrad, A. M. 1999. Diluting diversity: consciousness. British Journal of Social Psychol- Implications for intergroup inequality in organiza- ogy, 33: 1-27. tions. Journal of Management Inquiry, 8: 399-414. *Kearney, E., & Gerbert, D. 2009. Managing diversity Lipsey, andM. W., & Wilson, D. B. 2001. Practical meta- enhancing team outcomes: The promise of transfor- analysis. Thousand Oaks, CA: Sage. mational leadership. Journal of Applied Psychol- *Lovelace, K., Shapiro, D. L., & Weingart, L. R. 2001. ogy: In press. Maximizing cross-functional new product teams' in- *Kearney, E., Gerbert, D., & Voelpel, S. 2009. When novati and veness and constraint adherence: A conflict how diversity benefits teams: The importance communications of perspective. Academy of Manage- team members' need for cognition. Academy ment ofJournal, 44: 779-793. Management Journal, 52: 581-598. *Marsteller, J. A. 2003. The relationship between non- Keck, S. 1997. Top management team structure: racialDiffer- diversity in team composition and perfor-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 626 Academy of Management Journal June

mance in creativity in a chronic illness care tional Transformation and Social Change, 3: quality improvement intervention. Unpublished 223-241. doctoral dissertation, University of California, Quinn, J. B., Anderson, P., & Finkelstein, S. 1996. Leverag- Berkeley. ing intellect. Academy of Management Executive, Martins, L. L., Miliken, F. J., Wiesenfeld, B. M., & Sal- 10(3): 7-27. gado, S. R. 2003. Racioethnic diversity and group Ragins, B., & Sundstorm, E. 1990. Gender and power member's experience. Group and Organization organizations: A longitudinal perspective. Psycho- Management, 28: 75-106. logical Bulletin, 105: 51-88. Mayo, M., Pastor, J., & Meindl, J. R. 1996. The effects of *Raver, J. L., & Gelfand, M. J. 2005. Beyond the individual group heterogeneity on the self-perceived efficacy of victim: Linking sexual harassment, team processes, group leaders. Leadership Quarterly, 7: 265-284. and team performance. Academy of Management McGrath, J. E. 1984. Groups: Interaction and perfor- Journal, 48: 387-400. mance. Englewood Cliffs, NJ: Prentice-Hall. *Reagans, R., & Zuckerman, E. W. 2001. Networks, diver- Milkovich, G. T. 1987. Compensation systems in high sity, and productivity: The social capital of corporate technology companies. In D. B. Balkin & L. R. Go- R&D teams. Organization Science, 12: 502-517. mez-Mejia (Eds.), New perspectives on compensa- Reagans, R., Zuckerman, E. W., & McEvily, B. 2004. How tion. Englewood Cliffs, NJ: Prentice-Hall. to make the team: Social networks vs. demography Milliken, F. J., & Martins, L. L. 1996. Searching for com- as criteria for designing effective teams. Administra- mon threads: Understanding the multiple effects of tive Science Quarterly, 49: 101-133. diversity in organizational groups. Academy of Reskin, B. F., McBrier, D. B., & Kmec, J. 1999. The deter- Management Review, 21: 402-433. minants and consequences of workplace sex and Organisation for Economic Cooperation and Develop- race composition. In J. Hagan & K. Cook (Eds.), An- ment. 2006. OECD science, technology and indus- nual review of sociology, vol. 25: 355-361. Palo try outlook. Organisation for Economic Cooperation Alto, CA: Annual Reviews. and Development. Richard, O. C. 2000. Racial diversity, business strategy, Orwin, R. G. 1983. A fail-safe N for effect size in and firm performance: A resource-based view. Acad- meta-analysis. Journal of Educational Statistics, 8: emy of Management Journal, 43: 164-177. 157-159. Richard, O. C, Barnett, T., Dwyer, S., & Chadwick, K. Pelled, L. H. 1996. Relational demography and percep- 2004. Cultural diversity in management, firm perfor- tions of group conflict and performance: A fieldmance, and the moderating role of entrepreneurial investigation. International Journal of Conflict orientation dimensions. Academy of Management Management, 7: 230-246. Journal, 47: 255-266. *Pelled, L. H., Eisenhardt, K. M., & Xin, K. R. 1999. Richard, O. C, Murthi, B. P. S., & Ismail, K. 2007. The Exploring the black box: An analysis of work group impact of racial diversity on intermediate and long- diversity, conflict, and performance. Administrative term performance: The moderating role of environ- Science Quarterly, 44: 1-28. mental context. Strategic Management Journal, 28: Perry, E. L. 1997. A cognitive approach to understanding 1213-1233. discrimination: A closer look at applicant gender Ridgeway, C. L. 1991. The social construction of status and race. In G. R. Ferris (Ed.), Research in personnel value: Gender and other nominal characteristics. So- and human resource management, vol. 15: 175- cial Forces, 70: 367-386. 240. Greenwich, CT: JAI Press. Ridgeway, C. L. 1997. Interaction and the conservation of Pettigrew, T. F. 1998. Intergroup contact theory. In J. T. gender inequality: Considering employment. Amer- Spence, J. M. Darley, & D. J. Foss (Eds.), Annual ican Sociological Review, 62: 218-235. review of psychology, vol. 49: 65-85. Palo Alto, CA: Annual Reviews. Rosen, B., & Jerdee, T. H. 1977. Too old or not too old. Harvard Business Review, 55(6): 97-106. Porter, M. E. 1980. Competitive strategy. New York: Free Press. Rosenthal, R. 1979. The "file drawer problem" and toler- ance of null results. Psychological Bulletin, 86: *Portero-Brown, R. A. 1999. The missing link between 638-641. team heterogeneity and effectiveness: Individual Rousseau, D. M., & Fried, Y. 2001. Location, location, and contextual influences on social category salience. Unpublished doctoral dissertation, Uni- location: Contextualizing organizational research. Journal of Organizational Behavior, 22: 1-13. versity of California, Berkeley. *Puck, J., Rygl, D., & Kittler, M. 2006. Cultural anteced- Saavedra, R., Earley, P., & Van Dyne, L. 1993. Complex ents and performance consequences of open commu- interdependence in task-performing groups. Journal nication and knowledge transfer in multicultural of Applied Psychology, 78: 61-72. process-innovation teams. Journal of Organiza- *Sacco, J. M., & Schmitt, N. A. 2005. A dynamic multi-

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms 2009 Joshi and Roh 627

level model interdependenceof demographic and group diversity on innovation. diversity and misfit effects. Journal Journal of Management,of 29:Applied 29-51. Psychology, 90: 203-231. *Schippers, VanM. der Vegt, G.C, S., Van de Vliert,Den E., & Huang, X. Hartog,2005. D. N., Koopman, P. L., & Wienk, J. A. Location-level 2003. links between diversityDiversity and innova- and team outcomes: The moderating tiveeffects climate depend on national power of distance. outcome interdependence and group longevity Academy of Management and Journal, 48: the1171-1182. mediating effect of reflexiv- ity. Journal Van de Ven, ofA. H., Delbecq, Organizational A. L., & Koenig, R. 1976. Behavior, 24: 779- 802. Determinants of coordination modes within organi- Shea, G. P., & Guzzo, R. A. 1987. Groups as human zations. American Sociological Review, 41: 322- resources. In K. M. Rowland & G. R. Ferris (Eds.), 338. Research in personnel and human resource man- Van Knippenberg, D., De Dreu, C. K. W., & Homan, A. C. agement, vol. 5: 323-356. Greenwich, CT: JAI Press. 2004. Work group diversity and group performance: Shore, L. M., & Goldberg, C. B. 2005. Age discrimination An integrative model and research agenda. Journal in the workplace. In R. L. Dipboye & A. Colella of Applied Psychology, 89: 1008-1022. (Eds.), Discrimination at work: The psychological Van Knippenberg, D., & Schippers, M. C. 2007. Work- and organizational bases: 203-225. Mahwah, NJ: group diversity. In M. I. Posner & M. K. Rothbart Erlbaum. (Eds.), Annual review of psychology, vol. 58: 2.1- 2.27. Palo Alto, CA: Annual Reviews. Sidanius, J. 1993. The psychology of group conflict and the dynamics of oppression: A social dominance Watson, W. E., Johnson, L., & Merritt, D. 1998. Team perspective. In S. Iyengar & W. J. McGuire (Eds.), orientation, self-orientation, and diversity in task Explorations in political psychology: 183-219. groups: Their connection to team performance over Durham, NC: Duke University Press. time. Group and Organization Management, 23: 161-282. Skaggs, S., & DiTomaso, N. 2004. Understanding the effects of workforce diversity on employment out- Webber, S. S., & Donahue, L. M. 2001. Impact of highly comes: A multidisciplinary and comprehensive and less job-related diversity on work group cohe- framework. In N. DiTomaso & C. Post (Eds.), Re- sion and performance: A meta-analysis. Journal of search in the sociology of work: 279-306. New Management, 27: 141-162. York: Elsevier. Whitener, E. M. 1990. Confusion of confidence intervals *Somech, A. 2006. The effects of leadership style and and credibility intervals in meta-analysis. Journal of team process on performance and innovation in Applied Psychology, 75: 315-321. functionally heterogeneous teams. Journal of Man- Williams, K., & O'Reilly, C. A. 1998. Demography and agement, 32: 132-157. diversity in organizations: A review of 40 years of Teachman, J. D. 1980. Analysis of population diversity. research. In B. M. Staw & L. L. Cummings (Eds.), Social Methods and Research, 8: 341-362. Research in organizational behavior, vol. 20: 77- 140. Greenwich, CT: JAI Press. Thompson, J. D. 1967. Organizations in action. New York: McGraw-Hill. *Yeh, Y., & Chou, H. 2005. Team composition and learning behaviors in cross-functional teams. Social *Tyran, K. L., & Gibson, C. B. 2008. Is what you see, what Behavior and Personality, 33: 391-402. you get? The relationship among surface- and deep-

level heterogeneity characteristics, group efficacy, /AAV and team reputation. Group and Organization Man- agement, 33: 46-76. Aparna Joshi ([email protected]) is an assistant pro- U. S. Bureau of Labor Statistics. 2006. Labor force sta- fessor of labor and employment relations at the Univer- tistics from the Current Population Survey. Wash- sity of Illinois, Urbana-Champaign. She received her ington, DC: U. S. Department of Labor. Ph.D. from the School of Management and Labor Rela- U. S. Bureau of Labor Statistics. 2007. Occupational tions, Rutgers University. She conducts research in the outlook handbook, 2006-2007. Washington, DC: area of work team diversity, global and distributed U.S. Department of Labor. teams, team social capital, and generational issues in the workplace. U. S. Census Bureau. 2002. North American Industry Classification System (NAICS)-Revisions for 2002. Hyuntak Roh ([email protected]) is a Ph.D. candidate Washington, DC: U. S. Census Bureau. in the School of Labor and Employment Relations at the University of Illinois at Urbana-Champaign. His research *Van der Vegt, G. S., & Bunderson, J. S. 2005. Learning interests include workforce diversity, group process and and performance in multidisciplinary teams: The performance, employee turnover, and multilevel contex- importance of collective team identification. Acad- tual issues in HR/OB research. emy of Management Journal, 48: 532-547.

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

This content downloaded from 165.123.34.86 on Thu, 22 Feb 2018 12:46:41 UTC All use subject to http://about.jstor.org/terms