THE EFFECTS OF DIVERSITY ON BUSINESS ...web.mit.edu/cortiz/www/Diversity/Kochan et al, 2003.pdfTHE...

19
THE EFFECTS OF DIVERSITY ON BUSINESS PERFORMANCE: REPORT OF THE DIVERSITY RESEARCH NETWORK Thomas Kochan, Katerina Bezrukova, Robin Ely, Susan Jackson, Aparna Joshi, Karen Jehn, Jonathan Leonard, David Levine, and David Thomas This article summarizes the results and conclusions reached in studies of the relationships be- tween race and gender diversity and business performance carried out in four large firms by a research consortium known as the Diversity Research Network. These researchers were asked by the BOLD Initiative to conduct this research to test arguments regarding the "business case"for diversity. Few positive or negative direct effects of diversity on performance were observed. In- stead a number of different aspects of the organizational context and some group processes mod- erated diversity-performance relationships. This suggests a more nuanced view of the "business case" for diversity may be appropriate. © 2003 Wiley Periodicals, Inc. Introduction Since 1996, a group of industry chief execu- tives and human resource professionals have been working together under the auspices of a nonprofit organization called the Business Opportunities for Leadership Diversity (BOLD) Initiative to help American corpora- tions learn how to leverage their cultural di- versity for competitive advantage. These leaders espouse the now popular "business case" for diversity-the view that a more di- verse workforce will increase organizational effectiveness. For them, providing more op- portunities for women and minorities is a business imperative. Realizing, however, that they lacked clear evidence to support this view, either within their own organizations or more generally across American industry, these business leaders called for definitive re- search to assess the diversity-performance link. An initial study commissioned by BOLD found that no organizations were collecting the data needed to assess the effects of their diversity practices on firm performance (Cor- porate Leadership Council, 1997). There- fore, in 1997, the BOLD Initiative asked a group of researchers from a cross section of universities to design a large-scale field re- search project to examine the relationships between gender and racial diversity and busi- ness performance. This paper presents our conclusions from this five-year research effort. We believe Correspondence to: Thomas Kochan, MIT Sloan School of Management, Building #E52, Room 583, Cam- bridge, MA 02142; telephone: 617-253-6689; e-mail: [email protected] Human Resource Management, Spring 2003, Vol. 42, No. 1, Pp. 3-21 © 2003 Wiley Periodicals, Inc. Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI: 10.1002/hrm.10061

Transcript of THE EFFECTS OF DIVERSITY ON BUSINESS ...web.mit.edu/cortiz/www/Diversity/Kochan et al, 2003.pdfTHE...

THE EFFECTS OF DIVERSITY ON BUSINESSPERFORMANCE: REPORT OF THE DIVERSITYRESEARCH NETWORK

Thomas Kochan, Katerina Bezrukova, Robin Ely, Susan Jackson,Aparna Joshi, Karen Jehn, Jonathan Leonard, David Levine, andDavid Thomas

This article summarizes the results and conclusions reached in studies of the relationships be-tween race and gender diversity and business performance carried out in four large firms by aresearch consortium known as the Diversity Research Network. These researchers were asked bythe BOLD Initiative to conduct this research to test arguments regarding the "business case"fordiversity. Few positive or negative direct effects of diversity on performance were observed. In-stead a number of different aspects of the organizational context and some group processes mod-erated diversity-performance relationships. This suggests a more nuanced view of the "businesscase" for diversity may be appropriate. © 2003 Wiley Periodicals, Inc.

Introduction

Since 1996, a group of industry chief execu-tives and human resource professionals havebeen working together under the auspices ofa nonprofit organization called the BusinessOpportunities for Leadership Diversity(BOLD) Initiative to help American corpora-tions learn how to leverage their cultural di-

• versity for competitive advantage. Theseleaders espouse the now popular "businesscase" for diversity-the view that a more di-

• verse workforce will increase organizationaleffectiveness. For them, providing more op-portunities for women and minorities is abusiness imperative. Realizing, however, thatthey lacked clear evidence to support this

view, either within their own organizations ormore generally across American industry,these business leaders called for definitive re-search to assess the diversity-performancelink. An initial study commissioned by BOLDfound that no organizations were collectingthe data needed to assess the effects of theirdiversity practices on firm performance (Cor-porate Leadership Council, 1997). There-

fore, in 1997, the BOLD Initiative asked agroup of researchers from a cross section ofuniversities to design a large-scale field re-search project to examine the relationshipsbetween gender and racial diversity and busi-ness performance.

This paper presents our conclusionsfrom this five-year research effort. We believe

Correspondence to: Thomas Kochan, MIT Sloan School of Management, Building #E52, Room 583, Cam-bridge, MA 02142; telephone: 617-253-6689; e-mail: [email protected]

Human Resource Management, Spring 2003, Vol. 42, No. 1, Pp. 3-21© 2003 Wiley Periodicals, Inc. Published online in Wiley InterScience ( www.interscience.wiley.com ).DOI: 10.1002/hrm.10061

4 •

HUMAN RESOURCE MANAGEMENT, Spring 2003

During the1990s, diversityrhetoric shiftedto emphasizethe businesscase for sup-porting work-force diversity.

this to be the largest field-based researchproject on this topic undertaken to date. Wesummarize our results here and their implica-tions for managers and human resource prac-titioners and describe the challenges we en-countered along the way in the hopes ofadvancing the study and practice of diversityin organizations in the future.' Our resultssuggest the need to move beyond thebusiness-case argument for advancing thepractice of diversity in industry and how tomodify this argument to reflect the complexi-ties we discovered in our research. We alsopropose steps that industry professionals cantake to leverage the potential benefits of di-versity and to strengthen their ability to as-sess diversity initiatives in their organizations.

Historical Context of the Business-CasePerspective

The recognition that diversity is a reality inthe workforce has generated an enormousamount of activity over the years among lead-ers in business, government, and civil societyalike (for an extended discussion, see Jack-son & Joshi, 2001). An outcome of the civilrights movement, Title VII of the 1964 CivilRights Act, made it illegal for organizationsto engage in employment practices that dis-criminated against employees on the basis ofrace, color, religion, sex, and national origin(age and disability were legislated after1964). Through these government actions,society made a statement: employers mustprovide equal employment opportunities toall people of similar qualifications and ac-complishments. In addition, Executive Order11246, issued in 1965, required governmentcontractors to take affirmative actions toovercome past patterns of exclusion or dis-crimination. These societal mandates cur-tailed formal policies that discriminatedagainst certain classes of workers and raisedthe costs to organizations that failed to im-plement fair employment practices. The lawsremain a part of the legal responsibilitiesunder which firms and other labor-marketinstitutions, such as unions or job-matchingorganizations, operate today.

By the late 1970s and into the 1980s,there was growing recognition within the pri-

vate sector that, while the legal mandates werenecessary, they were not sufficient for ensuringthe effective management of diversity withinorganizations. Although the workforces ofmany organizations became more diverse, en-trenched organizational cultures, which re-mained inhospitable to traditionally underrep-resented groups, were slow to change. Topromote the development of more positive or-ganizational cultures that would support theeffective development of a more diverse work-force, many companies and consulting firmsbegan to offer training programs aimed at"valuing diversity". These efforts focused onchanging employees' attitudes and eliminatingbehaviors that reflected more subtle forms ofdiscrimination and exclusion, which often in-hibited effective interactions among people.The widespread adoption of such training pro-grams expanded the concept of "diversity" aspeople began to realize that visible, legally rec-ognized, demographic differences such as raceand gender were not the only types of differ-ences that affected work relationships amongemployees. Gradually, training initiatives pro-liferated, encouraging employees to value thewide range of physical, cultural, and interper-sonal differences, which would presumablyenhance decision making, problem solving,and creativity at work. Unfortunately, however,most studies show that such training rarelyleads to the desired long-term changes in atti-tudes and behavior (Bezrukova & Jehn, 2001).

During the 1990s, diversity rhetoricshifted to emphasize the business case forsupporting workforce diversity. Figure 1 re-ports how the former CEO of HewlettPackard described the new rhetoric. Essen-tially, he was looking for a way to convince hisfellow executives and managers that to man-age diversity effectively is a business necessitynot only because of the nature of labor andproduct markets today, but also because amore diverse work force-relative to a homo-geneous one-produces better business re-sults. He believed that providing evidence tosupport these claims would accelerate therate of progress employers would make in hir-ing and developing a more diverse workforceand produce organizations that are more fullyintegrated across occupations and levels ofhierarchy. Likewise, for diversity advocates,

The Business Case for Diversity

"I see three main points to make the busi-ness case for diversity:

I. A talent shortage that requires us toseek out and use the full capabilities ofall our employees.

2. The need to be like our customers, in-cluding the need to understand andcommunicate with them in terms thatreflects their concerns.

3. Diverse teams produce better results.

This last point is not as easy to sell as thefirst two-especially to engineers who wantthe data. What I need is the data, evidencethat diverse groups do better."

Figure 1. Lew Platt, former CEO of Hewlett Packard,comments to the Diversity Research Network, StanfordBusiness School, March 18, 1998.

the new imperative was to find evidence tosupport the business-case argument.

In fact, as both the study commissionedby BOLD and our own reviews of the re-search literature (see, e.g., Richard & John-son, 1999; Richard et al., 2002; Williams &O'Reilly, 1998) have shown, there is little re-search conducted in actual organizationsthat addresses the impact of diversity or di-versity-management practices on financialsuccess. While there are a large number oflaboratory experiments that test specific di-versity-performance hypotheses, there arefew such studies in real organizations andfewer still that assess this hypothesis usingobjective performance measures. An excep-tion is a study that compared companies withexemplary diversity-management practices tothose that had paid legal damages to settlediscrimination lawsuits. The results of thisstudy showed that the exemplary firms alsoperformed better as measured by their stockprices (Wright et al., 1995). Overall, how-ever, the search for evidence that directlysupports the business-case hypothesis hasproved elusive.

Two reasons might explain this lack ofevidence. First, diversity is extremely diffi-cult to study in organizational settings be-

Effects of Diversity on Business Performance

cause it raises sensitive issues that are diffi-cult to discuss. In addition, organizations, in-cluding many we contacted during this proj-ect, are reluctant to share their experiencesor data, given the legal climate and the po-tential for litigation. Another reason for thelack of evidence linking workforce diversityto business performance may be that the re-lationship between diversity and the bottomline is more complex than is implied by thepopular rhetoric. Decades of research on theeffects of diversity within teams and smallgroups indicate that diversity can have nega-tive effects, as well as positive ones. The em-pirical literature does not support the simplenotion that more diverse groups, teams, orbusiness units necessarily perform better,feel more committed to their organizations,or experience higher levels of satisfaction(Jackson et al., 1995; Millikin & Martins,1996; Williams & O'Reilly, 1998). Instead,the evidence suggests that diversity may si-multaneously produce more conflict and em-ployee turnover as well as more creativityand innovation (Jehn et al., 1999; cf.Williams & O'Reilly, 1998). For example, thispattern of mixed results was found in twostudies that examined diversity within topmanagement teams in the banking industry.In one study, diversity in top managementteams was associated with greater innovationwithin bank branches (Bantel & Jackson,1989). In another, diversity also associatedwith higher rates of turnover among topmanagement team members (Jackson et al.,1991). Thus, the research literature paints amore complex picture of the consequencesof diversity than does the popular rhetoric.

The Diversity Research Network andResearch Project

It was this mismatch between research re-sults and diversity rhetoric that led us toagree to form a Diversity Research Network.Our purpose was to conduct a multi-firmstudy of the effects of racial and gender di-versity on organizational performance. TheAlfred P. Sloan Foundation, the Society ofHuman Resource Management (SHRM)Foundation, and the BOLD Initiative pro-vided the funding for the Network's research.

6 HUMAN RESOURCE MANAGEMENT, Spring 2003

Figure 2 presents the model that guidedthe design of the studies discussed in this ar-ticle. Members of the Diversity ResearchNetwork developed the model collaborativelybased on a comprehensive review of the largenumber of laboratory studies and the smallnumber of field studies concerning the ef-fects of diversity on group dynamics andgroup performance (Richard et al., 2002;Williams & O'Reilly, 1998). The model re-flects both theory and empirical research,which suggest that whether diversity has apositive or negative impact on performancemay depend on several aspects of an organi-zation's strategy, culture, and human re-source (HR) practices. In addition, themodel proposes that these effects are likelyto operate through group or team processes,such that under facilitating conditions, di-versity is associated with positive group orteam processes and is therefore beneficial toperformance; whereas under inhibiting con-ditions, diversity is associated with negativegroup or team processes and is thereforedetrimental to performance.

Specifically, the literature suggests thatdiversity, if unattended, is likely to have an

adverse effect on group processes, such ascommunications, conflict, and cohesion(Williams & O'Reilly, 1998). More specifi-cally, diversity in a work group can producelower cohesion and miscommunicationamong group members, which can lead togroup conflict (Jehn, 1995). Some of thisconflict may be productive-if, for example,it avoids "groupthink" and brings additionalpoints of view into the discussion-whereasother forms may worsen group performance.

The connections shown in the model thatlink diversity to team processes and then toperformance seem logical, but we recognizethat past research has not always found stronglinkages between diversity and performanceoutcomes. In fact, past research suggests thatthere may be no direct positive or negative re-lationship between diversity and performanceoutcomes. In some groups, diversity may im-prove performance, while in other groups, di-versity may be detrimental to performance(Jackson, 1992; Jehn et al., 1999; O'Reilly &Flatt, 1989; Richard, 2001; Steiner, 1972). Ifdiversity has inconsistent effects acrossgroups, then in studies that examine the rela-tionship between diversity and performance

Organizational Context

Organizational Culture

Business Strategy

Human Resource Policies and Practices

Diversity Group/Team Processes Outcomes

Cultural

Communications

Performance

Demographic

Conflict

Satisfaction

Technical

Cohesion

Turnover

Cognitive

Information

Creativity

Figure 2. The model: the effects of diversity on group processes and outcomes.

across many groups, the positive and negativeeffects may cancel each other out so that noeffect obtains. Therefore, our research modelsuggests that the relationship between diver-sity and performance may depend on the or-ganizational context in which the work takesplace. For example, the effects of diversity onorganizational performance might be more fa-vorable if group leaders and members build onteam members' creativity and information. Di-versity may also be more likely to improve per-formance when group members and leadersare trained to deal with group process issues,particularly those involved in communicatingand problem solving in diverse teams. Pre-sumably, HR practices for recruiting, select-ing, training, motivating, and rewarding em-ployees partially determine whether teammembers and leaders are skilled in communi-cating with and coordinating among membersof diverse teams. When HR practices supportthe creation of a workforce that has the skillsneeded to turn diversity into an advantage, di-versity is more likely to lead to positive perfor-mance outcomes. In other organizations,however, HR practices may inadvertently re-sult in teams that are diverse but unskilled indiversity management. Such organizations aremore likely to experience negative outcomes,such as disruptive conflict and increasedturnover, which can harm performance.

Although not shown in the model, thereis another hypothesis implicit in the businesscase for diversity, which is related to the sec-ond point in Figure 1: a match between thedemographic composition of the workforceand the firm's customer base will yield highersales, thus enhancing organizational perfor-mance. The evidence bearing on this hypoth-esis is limited, however, and derived largelyfrom laboratory studies. One of the four stud-ies included in this project contains what webelieve to be the largest and most completeexamination of this hypothesis to date.

Data Collection

The BOLD Initiative first approached severalof us in early 1997 with the idea of forminga collaborative industry-university researchproject to explore the business case for di-versity. Researchers from a number of uni-

Effects of Diversity on Business Performance

versities met twice to discuss whether such aproject was feasible. We were well aware ofthe difficulties associated with field researchon this topic. It raises politically and emo-tionally charged issues, as well as legal con-cerns. In addition, it would be difficult to de-velop and even more difficult to implement aresearch design that would enable us to drawvalid, convincing conclusions about thecausal effects of diversity on organizationaloutcomes. In an ideal world, such a designwould entail longitudinal data collected froma large and representative sample of organi-zations, enabling us to track how changes indemography influence performance overtime. We knew from past organizational re-search that the data needed to examine theimpact of diversity on performance are quiteextensive, seldom collected, or relatively in-accessible, and unlikely to be comparableacross organizations. Yet we all shared theview that if we were to generate knowledgethat would be useful to practitioners on therelationship between diversity and perfor-mance, we needed to move the research onthis topic from the laboratory to the field.Moreover, we were intrigued by the potentialbenefits of forming a research networkamong those working on this topic, generat-ing a common framework to guide research,and comparing results across multiple organ-izations. Some of us had had positive experi-ences in building and participating in a sim-ilar network to explore the relationshipbetween human resource practices and firmperformance, a topic that individual re-searchers had previously found equally diffi-cult to study (Cohen, 1998; Ichniowski etal., 1998). Therefore, we decided to moveahead with the support of the Sloan Founda-tion, SHRM, and the BOLD Initiative.

The process of recruiting companies toprovide the data and participate in the re-search proved to be even more difficult thanwe had expected. Over a two-year period(1998-2000), members of the research net-work and leaders of the BOLD Initiative initi-ated discussions with over 20 large and well-known Fortune 500 companies, all of whichexpressed considerable interest in the topic.However, after often considerable discussionof the data, confidentiality, and time commit-

When HR prac-tices support thecreation of aworkforce thathas the skillsneeded to turndiversity into anadvantage, di-versity is morelikely to lead topositive perfor-manceoutcomes.

8 •

HUMAN RESOURCE MANAGEMENT, Spring 2003

While our orig-inal hope wasto collect thesame kind ofdata in eachcompany andto use the sameinstruments formeasuring thegroup processand contextvariables in ourmodel, itquickly becameapparent thatthis was notfeasible.

ments, all but four companies declined to par-ticipate. In some cases, the diversity advocatesand professionals in the company lacked suf-ficient influence to convince line managers tospend the time required to collect the neces-sary data. In other cases, these professionalswere reluctant to examine the effects of theirorganization's policies, with the view that theyhad sufficient top management support fortheir current initiatives and did not need todemonstrate a business case to maintain thissupport. In still other cases, we found that,even with the support of the CEO, objectionsraised by internal or external legal counseland/or by other managers who would need toprovide the data proved insurmountable. Inthe end, each of the firms that agreed to par-ticipate had a prior relationship with one ormore members of the research team and/orleaders of BOLD Initiative and, therefore, hadalready established a high level of trust. Thus,the first lesson of this research was that notonly had none of the organizations we con-tacted ever conducted a systematic examina-tion of the effects of their diversity efforts onbottom-line performance measures, very fewwere interested in doing so.

While our original hope was to collectthe same kind of data in each company andto use the same instruments for measuringthe group process and context variables inour model, it quickly became apparent thatthis was not feasible. Each company had itsown particular ways of collecting and storinghuman resource data and three of four firmsindicated a strong preference for using theirown internal survey measures to capture thevariables in the model. Therefore, eachstudy draws on somewhat different kinds ofdata to address common questions about theeffects of diversity on performance. All fourcompanies are large and highly respected,and each has a long history of success inachieving a diverse workforce and commit-ment to leveraging diversity to enhance or-ganizational performance.

Overall, our conclusions are based onanalyses of a mix of qualitative and quantita-tive data collected across the four studies.Within each company, we identified compa-rable teams, work groups, or business unitsto serve as the unit of analysis. In one study,

we collected qualitative data on business-unit cultures, HR and managerial practices,and business strategies and on the quality ofbusiness-unit processes. In others, we reliedon survey data to assess these aspects of theorganizational context and the quality ofgroup processes. In all four studies, we usedarchival data on the demographic composi-tion of teams, work groups, or business units;in one study, we also had census data on thedemographic composition of the communi-ties from which the business units drew theircustomers. And finally, in all four companies,we used a variety of objective measures to as-sess performance.

Results

Study 1: A Large Information-Processing Firm

Study 1 was conducted by the research teamof Karen Jehn and Katerina Bezrukova fromthe Wharton School at the University ofPennsylvania. This study is based on datafrom a large Fortune 500 information-processing company with over 26,000 em-ployees. Diversity has been at the forefrontof this company's social and business agendafor over half a century. In the early 1940s,when the company chairman took severalsales representatives, including one AfricanAmerican, to an awards ceremony, the hotelhosting the event refused to register theAfrican American, so the chairman left withhis sales force in tow. In 1987, managementrealized that many minority employees werenot advancing through the ranks, so theWomen's Resource Group and the MinorityResource Group were created to address theissue. In 1992, a Diversity Task Force wascreated to develop a strategic plan for pro-moting diversity. Each year since then theDiversity Leadership Council, comprised ofemployees at every level and in every depart-ment of the company, helps to create diver-sity action plans, which outline initiativeslinked to the strategic goals of the DiversityTask Force.

Each business unit of the company is re-quired to submit an end-of-year report mea-suring how well the unit performed againstits diversity objectives. This report details

r

quantitative information on diversity initia-tives, including the hiring and promotion ofwomen and minorities, succession planning,development, retention, and training. Thisapproach has resulted in a number of toolsdesigned to support the company's diversityobjectives. For instance, all business-unitnewsletters cover diversity, employees haveaccess to a lecture series on diversity, and"Managing Diversity" is taught at every man-ager orientation session.

The company's external practices arealso shaped by their belief that diversity is es-sential to innovation and growth. Thesestrategies include establishing alliances withexternal women's and minority organizations,supporting small businesses, and supportingregional minority business councils acrossthe country. Additionally, the company ac-tively supports minority scholarships, coop-erative education, and internship programs;and it provides substantial funding to na-tional organizations concerned with profes-sional, social, and educational goals in mi-nority communities.

To generate measures of the groupprocess and context variables, we content an-alyzed qualitative data contained in companydocuments that were part of a human re-sources-sponsored program created to iden-tify high potential employees and to recog-nize and hold accountable leaders at alllevels of management and supervision. Aspart of this program, managers create Devel-opment Plans (DPs) for their work groupsand discuss these plans with their supervi-sors. The Plans capture the dominant groupprocesses occurring in groups. In addition toDPs, managers and supervisors evaluatetheir leadership competencies (i.e., values,goals, skills, and knowledge). The competen-cies assessed in the supervisor reports serveas indirect evidence of the actual context re-garding organizational cultures, businessstrategies, and human resource practices(Dozy et al., 1993).

To specify our group process and con-text variables, we developed respective listsof key words based on relevant group andorganizational theories as well as the con-cepts used in the company's rhetoric. Team-focused processes relating to building com-

Effects of Diversity on Business Performance

mitment and group spirit, change-focusedprocesses relating to innovation and explor-ing new perspectives, and career-focusedprocesses relating to career advancementsand professional success exemplify con-structive group processes we investigated inthis study. Context variables are defined bybusiness strategies (i.e., growth, stability,and customer-oriented), culture (i.e., peo-ple- and competition-oriented), and humanresource practices (i.e., training- and diver-sity-oriented). Following the method ofJehn and Werner (1993), which was suc-cessfully employed in past organizationalresearch, two independent raters reviewedthe surrounding context and coded the tex-tual data for each work group on each ofthese dimensions. Two measures of teamperformance were available for this study:average performance-appraisal ratings pro-vided by the managers of each business unitand average bonuses of team members.Using these measures, our test of the modelguiding the overall research project re-vealed the following:

1. There were no significant direct ef-fects of race or gender diversity oneither team performance-appraisalratings or bonuses.

2. Diversity had a significant effect ongroup processes, but the nature of theeffect depended on whether the diver-sity was in gender or race. Specifi-cally, gender diversity increased con-structive group processes, while racialdiversity inhibited them.

3. Training- and development-focusedHR practices, including coaching,open communications and interac-tive listening, and providing chal-lenging assignments and opportuni-ties for development, reduced thenegative effects of racial diversity onconstructive group processes.

4. Diversity-focused HR practices en-hanced the positive effects of genderdiversity on constructive groupprocesses.

5. Constructive group processes, inturn, had a positive impact on per-formance ratings and group bonuses.

To specify ourgroup processand contextvariables, wedeveloped re-spective lists ofkey wordsbased on rele-vant group andorganizationaltheories as wellas the conceptsused in thecompany'srhetoric.

JO • HUMAN RESOURCE MANAGEMENT, Spring 2003

To ensure thattheir workforcereflects thecommunitiesthey serve, thecompanyaggressivelyrecruitscandidates ofall backgrounds.

We also conducted additional analyseson a larger sample of groups for which nogroup process data were available. With thissample, we examined the role that organiza-tional context played in shaping the relation-ship between diversity and performance. Wefound the following:

6. Gender diversity was positively associ-ated with group bonuses in businessunits with a people-oriented organiza-tional culture, diversity-focusedhuman resource practices, and cus-tomer-oriented business strategy butnot in units that lacked those specificcultures, practices, and strategies.

7. Racial diversity was negatively re-lated to performance in businessunits with competitive organizationalcultures, growth-oriented businessstrategies, and training-focusedhuman resource practices but not inunits that lacked those specific cul-tures, practices, and strategies.

In summary, our results in this organiza-tion showed no significant direct effects ofeither racial or gender diversity on perfor-mance. Gender diversity had positive effectson group processes while racial diversity hadnegative effects. The negative relationshipbetween racial diversity and group processeswas, however, largely absent in groups thathad received high levels of training in careerdevelopment and diversity management.

Study 2: A Financial Services Firm

Robin Ely and David Thomas at the HarvardBusiness School carried out Study 2. The datacome from the retail branches of a large fi-nancial services firm that is highly respectedfor integrating a commitment to diversity intoits managerial policies and strategies. At thecorporate level, the company has imple-mented four practices in particular which theybelieve represent "best practice" in the field.First, all senior managers in the company, in-cluding regional managers of the retailbranches, are held accountable for managinga formal diversity plan and for linking diversityto education, recruiting, succession planning,

career development, and business growth.These plans cascade down to individualbranch managers who, as part of their re-gional manager's plan, have diversity objec-tives they are required to meet. Second, in ad-dition to a company-wide Diversity Council,chaired by the CEO, each business hosts itsown diversity council chaired by its respectivebusiness executive, ensuring hands-on em-ployee involvement in their diversity initia-tives. The company now has 45 diversitycouncils around the world, involving some1000 employees, including many retail-branch employees. Third, the company con-siders itself unique in extending the focus ofits diversity efforts beyond race and gender.Their education efforts and dialogue with em-ployees across the company include race, gen-der, disability, religion, sexual orientation, andage. Finally, they have an aggressive construc-tion plan for back-up childcare centers withthe belief that they need to support the diversework-life needs of their employees. They arecurrently constructing a network of 17 on-siteback-up childcare centers in major companysites across the United States, to which mostretail-branch employees have access.

To ensure that their workforce reflectsthe communities they serve, the company ag-gressively recruits candidates of all back-grounds. In their entry management trainingprograms, they have established close re-cruiting ties with historically Black colleges.They participate in a number of internshipprograms that provide opportunities for bothhigh school and college level students. Theirintercept programs include A Better Chance,Smart Start, a Fellows Program, and Sum-mer Jobs for Youth, with the hope that manyof their interns will choose full-time employ-ment with their firm upon graduation. Dueat least in part to the success of these efforts,there are two distinctive features of thebranches' demographic composition. First,the racial composition of the branches iswide ranging, including branches that arepredominantly Black, Hispanic, Asian, orWhite as well as branches with virtuallyevery possible mix of these groups. Thus, thisstudy overcomes a common limitation to ex-isting research on diversity in which racialheterogeneity is often confounded with the

h

proportion of minorities, and comparisonsare limited to Whites and Blacks, or Whitesand "others" (Williams & O'Reilly, 1998).

Second, the branches in this bank are largelyfemale dominated. The male-dominated set-ting, typical of many corporations, is nonex-istent in the retail sector of the bank. Thismeans that the sex composition of thebranches is narrow, ranging from all womento about equal proportions of women andmen, and that a more diverse branch is onethat has a greater than average proportion ofmen employees, or, stated another way, a sexcomposition that is closer to a 50:50 ratio ofmen to women, relative to one that is pre-dominantly female. Our sample consisted of480 retail branches of the firm located pri-marily in a large city in the northeasternUnited States. Branches in this sampleranged in size from 4 to 70 employees, withan average of 15 and a standard deviation of10. The average proportion of Whites in thebranches was 49%; the average proportion ofwomen was 83%.

The data came from three sources: (1)archival data on the demographic character-istics of each employee in each branch, (2)employee attitude-satisfaction data from anannual employee attitude-satisfaction survey,and (3) branch performance data used to al-locate bonuses to branches on a semiannualbasis. Most of the results presented here arebased on analyses of data collected at theend of 1999; some analyses included datafrom 2000 and 2001, as well.

We obtained performance measuresfrom the firm's branch bonus-award systemwhereby branches are assessed semiannuallyon six areas of performance, relative to goalsset for the branch in each area. The six per-formance areas were: (1) revenue from NewSales, (2) revenue from growing the Con-sumer Portfolio (growth over the six-monthassessment period in revenue from retail cus-tomers), (3) revenue from growing the Busi-ness Portfolio (growth over the six-month as-sessment period in revenue from businesscustomers), (4) Customer Satisfaction (acomposite score assessed from independentlyconducted surveys of approximately 50 ran-domly selected customers for each branch),(5) number of Qualified Referrals to bank

Effects of Diversity on Business Performance

services (referrals by employees from oneproduct to another that resulted in sale tothe customer), and (6) Sales Productivity(total revenue from new sales relative tototal-salary expense). There was also a TotalPerformance score, which is generated froma weighted point system that the bank usesto assign bonuses to branches. Averagingacross five items from the company's em-ployee attitude-satisfaction survey, we devel-oped a branch-level measure of the quality ofa branch's team processes.

As proposed in the model guiding thisproject, we expected that the relationshipsbetween diversity and team processes andbetween team processes and outcomeswould likely depend on contextual factorsthat differentiated the branches. Using theemployee attitude-satisfaction survey, we de-veloped measures of two such factors. First,we examined the proportion of branch em-ployees who had attended at least one of thefirm's diversity-education programs. Most ofthese programs address multiple dimensionsof diversity, such as race, ethnicity, gender,age, religion, disability, and sexual orienta-tion. The primary purpose of these programsis to increase awareness o£ cultural differ-ences and how people's perceptions, biases,and stereotypes of others influence their be-havior, and teach skills for addressing con-flicts and managing discussions of issues re-lated to diversity. We hypothesized that highlevels of both gender and racial diversitywould lead to more positive outcomes inbranches in which a higher proportion ofemployees had attended at least one diversityprogram. Second, we developed a measure ofthe branches' diversity perspectives, a featureof work groups that we identified in previousresearch as a crucial factor in determiningwhether racial diversity has a positive or neg-ative impact on group performance (Ely &Thomas, 2001; Thomas & Ely, 1996). In par-ticular, we used four items from the survey todevelop a measure of the degree to whichbranches evidenced an integration-and-learning perspective on diversity. In our priorresearch, we identified three different per-spectives on diversity that culturally diversework groups, or groups aspiring to be cultur-ally diverse, hold. A work group's diversity

1 1

The primarypurpose of theseprograms is toincrease aware-ness of culturaldifferences andhow people'sperceptions, bi-ases, andstereotypes ofothers influ-ence theirbehavior...

1 2 • HUMAN RESOURCE MANAGEMENT, Spring 2003

Participation indiversity-educa-tion programsdid not foster apositive rela-tionship be-tween racialand gender di-versity andperformance.

perspective is the rationale that guides mem-bers' efforts to create and respond to culturaldiversity in the group. The three perspectivesare the discrimination-and-fairness perspec-tive, the access-and-legitimacy perspective,and the integration-and-learning perspective.We found that, whereas all three perspec-tives can be successful in motivating man-agers to diversify their staffs, only the inte-gration-and-learning perspective providesthe rationale and guidance needed to achievesustained benefits from diversity. Accordingto the integration-and-learning perspective,the insights, skills, and experiences employ-ees have developed as members of variouscultural identity groups are potentially valu-able resources that the work group can useto rethink its primary tasks, and redefine itsmarkets, products, strategies, and businesspractices in ways that will advance its mis-sion. In branches with an integration-and-learning perspective, we expected that em-ployees of all races would be encouraged tobring all relevant insights and perspectives tobear on their work and, hence, that high lev-els of racial diversity would be associatedwith better performance in branches that en-acted an integration-and-learning perspec-tive on their diversity and worse performancein branches that did not.

The key findings from these analyses areas follows:

1 Across the six separate performancemeasures and the overall perfor-mance index, we found only one sig-nificant direct effect of racial diver-sity on performance: racial diversitywas positively associated with growthin branches' business portfolios.Gender diversity had no significantdirect effects on any of the perfor-mance measures.

2. Our notion that the effects of diver-sity on performance would dependon certain context variables was par-tially supported. In particular, racialdiversity was associated with higheroverall performance in branches thatenacted an integration-and-learningperspective on diversity, relative toracially diverse branches that did not

enact this perspective and relative toracially homogeneous branches.

3. Employee participation in diversity-education programs had limited im-pact on performance. Branches inwhich a higher proportion of em-ployees had participated in at leastone diversity-education program out-performed branches with lower em-ployee participation in these pro-grams on only one measure ofperformance (sales productivity).Participation in diversity-educationprograms did not foster a positive re-lationship between racial and genderdiversity and performance. It had noimpact on the racial diversity-perfor-mance link, and, unexpectedly, anegative impact on the gender diver-sity-performance link for one mea-sure of performance, such that inbranches with high employee partic-ipation, greater gender diversity wasassociated with lower performance.

In summary, this organization has manyof the features that should be conducive toleveraging the potential benefits of diversityor, at the very least, mitigating its potentialcosts. This may explain the direct positive ef-fect that racial diversity had on one measureof performance and the lack of any directnegative effects. Nevertheless, there wasroom for improvement. Racial diversity had apositive effect on overall performance inbranches that used that diversity as a re-source for innovation and learning and a neg-ative effect otherwise, suggesting that enact-ing a learning perspective on diversity can payoff, even in groups embedded within organi-zations that are already highly committed toand relatively sophisticated about diversity.We expected that employee participation indiversity-education programs might enhanceperformance, especially in the more diversebranches, but found little evidence to supportthis. One possible explanation for our findingthat, overall, these programs had little impacton performance is that they are serving as aneffective ameliorative to problems encoun-tered in more diverse branches, thereby help-ing to create parity in performance between

branches with more and less diversity. The in-verse relationship between gender diversityand performance in branches with higher em-ployee participation in diversity-educationprograms is consistent with this interpreta-tion. It may be that branches that are moresex-balanced (i.e., more gender-diverse) aremore likely to encounter problems in the firstplace, relative to their more typical, female-dominated counterparts, thus sending ahigher proportion of their employees to theseprograms. In this case, program participationmay more likely signal performance-inhibit-ing problems than provide performance-en-hancing training.

Study 3: An Information-Processing Firm

Study 3 was conducted by Susan Jackson atthe School of Management and Labor Rela-tions of Rutgers University and Aparna Joshiat the Institute of Labor and IndustrialRelations, University of Illinois at Urbana-Champaign. The data were drawn from an-other Fortune 500 firm in the information-processing industry. Of all the organizationsstudied, this company's diversity-managementefforts varied least across its different busi-ness units or locations.

Over the past several decades, the com-pany consistently promoted workforce diver-sity and proactively worked to increase theproportions of women and minority employ-ees. It has a long history of supporting em-ployee support/identity groups that providementoring and other supports to their mem-bers. The company's initiatives for managingdiversity are reflected in staffing procedures,performance appraisals, and training. Sev-eral national awards have recognized thecompany's excellent programs for creatingand managing diversity.

All divisions of the company must meetannual targets for the representation of ma-jority and minority males and females ineach employee grade level. Availability of mi-nority and majority males and females is de-termined by examining the internal laborpools as well as U.S. Census data. Of the en-tire work force in the U.S. division, approxi-mately 33% were women, 17% were AfricanAmericans, less than 10% were Hispanics,

Effects of Diversity on Business Performance

and Asian and other ethnic groups comprisedabout 5%. These numbers attest to the suc-cess the company has achieved in recruitingand promoting a diverse workforce.

In order to enforce the annual targets de-scribed above, performance appraisals forline managers included measures of the man-agers' abilities to achieve the targets. The per-formance appraisals were used for makingpromotion and compensation-related deci-sions. Training practices included intensivediversity training. Trainers used behavioralmodeling techniques to help develop mana-gerial capabilities for interacting with subor-dinates and colleagues irrespective of demo-graphic differences. Thus the training effortsfocused more on skill-building than on build-ing awareness or modifying attitudes.

The quantitative portion of this study fo-cused on the company's U.S. employees inthe sales (n = 3970 employees) and service(n = 8636 employees) divisions. The demo-graphic characteristics of employees in salesand service are substantially different. Gen-der and ethnic diversity was greater in thesales division than in the service division. Inservice, only 6% of employees were female,while 35% of sales personnel were female.Within sales, the ethnic distribution was thesame statistically for males and females.Within service, however, the ethnic distribu-tion was statistically different for males andfemales. Although the difference was small,the pattern shows that females were slightlymore likely to be White. The individual par-ticipants in this study were organized into578 sales teams and 1820 service teams.Analyses were conducted separately for thesetwo occupational groups.

The performance measures used in theanalyses for the sales teams were sales-teamgoal achievement and sales-based bonuses.The company recorded performance for indi-vidual sales personnel, and from this infor-mation we created team-performance mea-sures by averaging across all members of ateam. Sales-team goal achievement was theaverage value of the team sales representa-tives' performance against their individualgoals for generating sales revenue. This mea-sure was a percentage value that reflectedthe actual revenue generated compared to

13

The company'sinitiatives formanagingdiversity arereflected instaffingprocedures,performanceappraisals, andtraining.

14 •

HUMAN RESOURCE MANAGEMENT, Spring 2003

Genderdiversity withinregions waspositivelyrelated to goalachievement aswell as speed ofresponse tocustomers.

the targeted revenue. The team sales-basedbonus measure was the average dollaramount of monthly bonuses awarded to theteam sales representatives. Bonuses werebased on individual sales performance rela-tive to revenue goals and were considered bythe firm to be a direct indicator of perfor-mance for these employees.

For the service teams the performancemeasures used were the teams achievementof goals related to machine performance andservice response times. All service teams wereresponsible for maintaining reliable machineswhile utilizing resources effectively. Thus,one measure of team performance used inthis company was success in meeting targetsfor the performance of the machines servicedby the team. Performance targets were set forthe team using historical data that took intoaccount the type of machine and the typicalperformance of such machines. The resultingvalue was expressed as a percentage. A scoregreater than 100% indicated that the qualityof repair service was better than the standardand machines assigned to these teams weremore reliable. Response time was anothermeasure of team performance used in thiscompany. Response time refers to the amountof time that elapsed from the time a call wasreceived by the team until a technician ar-rived at a customer site. Response time tar-gets were based on the products that teamsserviced and the locations of clients. Re-sponse time was expressed as a percentage,with a value over 100% indicating that theteam responded faster than its target.

A summary of results for service teams (seeJoshi [2002] for more details) are as follows:

1. Team-level gender and ethnic diver-sity was not significantly related toprocess outcomes such as team co-operation.

2. Team gender diversity was not signif-icantly related to team goal achieve-ment. However, there was a signifi-cant negative relationship betweenteam ethnic diversity and team goalachievement.

3. In a second round of analyses, we ex-amined the effects of diversity withinlarger organizational units using hi-

erarchical linear modeling tech-niques. In these analyses, we consid-ered whether diversity had differenteffects when regions (not smallerwork teams) were the focus. Withinregions, employees were interde-pendent with each other, but noteveryone was in close personal con-tact. When we studied regions, ourfindings changed. Specifically, gen-der diversity within service regionswas positively related to cooperationwithin the region. Additional analy-ses revealed that, in service regionswith a greater proportion of femaleservice technicians and female man-agers, teams were more cooperative(regardless of their diversity). Wefound a similar pattern for perfor-mance outcomes. Gender diversitywithin regions was positively relatedto goal achievement as well as speedof response to customers. The storydoes not end here, however.

4. In a third set of analyses, we askedwhether the regional demographiccontext in which teams worked influ-enced whether the teams benefitedfrom their diversity. We reasonedthat diverse teams would be best ableto leverage their resources when themembers of the team could formbeneficial relationships with otheremployees in the region. Our analy-ses examined the influence of serviceregion peer and managerial diversityon the relationship between team di-versity and team process and perfor-mance. Our results indicated thatthe effects of team diversity were in-fluenced by diversity within the re-gion. For regions with greater genderdiversity, there was a stronger posi-tive relationship between team gen-der diversity and team cooperation.That is, gender diverse teams weremore likely to be cooperative in re-gions that also were gender diverse.Similar analyses improved our un-derstanding of the relationship be-tween team ethnic diversity andteam goal achievement. In regions

with little ethnic diversity, the nega-tive relationship between ethnic di-versity and goal achievement wasmoderately strong. However, in re-gions with more ethnic diversity, thenegative effects of ethnic diversitywere reduced. Furthermore, wefound that ethnically diverse teamsperformed better when they wereembedded in ethnically diverse orga-nizational units.

A summary of results for sales teams (seeJackson & Joshi [2002] for more details)show the following:

1. For sales teams, we again examinedthe direct relationship between eachindicator of team diversity (e.g., gen-der diversity and ethnic diversity)and both team processes and teamperformance. We found no evidencefor a direct relationship betweenteam gender or ethnic diversity andthe team processes or performance.However, further analyses revealedthat tenure diversity moderated theeffects of gender and ethnic diversity.For example, in teams with littletenure diversity, ethnic diversity wasnegatively related to team perfor-mance, but in teams with greatertenure diversity, there was a positiverelationship between ethnic diversityand team performance.

2. A second set of analyses consideredthe role of the demographic charac-teristics of team managers. We foundthat team performance was influ-enced by the combination of man-ager demographics and team diver-sity. For example, for teams led bymale managers there was no rela-tionship between team gender diver-sity and team performance. How-ever, for teams led by femalemanagers greater gender diversitywas associated with poorer perfor-mance. In contract, for teams led bymanagers of color, we found a posi-tive relationship between team eth-nic diversity and team performance.

Effects of Diversity on Business Performance • 1 5

3. As we did for service teams (see #4,above), we also examined the effectsof regional diversity. For sales teams,we again found that regional diversityinfluences the effects of team diver-sity. However, the findings for salesteams did not mirror the findings forservice teams. For the sales teams, wefound a negative relationship be-tween team ethnic diversity and teamperformance for teams located in re-gions with greater ethnic diversity.This finding suggests that in a salessetting the ethnic diversity may havesome problematic outcomes unlessmanaged effectively. In analyses thatconsidered the combined effects ofgender, tenure, and ethnic diversity,we found a similar pattern. That is, inregions with more total diversity(gender, ethnic, and tenure), greaterteam diversity was associated withpoorer team performance.

In summary, this study found no signifi-cant direct effects of racial or gender diversityon team performance, but the results do re-inforce the argument that organizational con-text matters. The demographic makeup of thelarger organization within which teams func-tion has important consequences for effectsof diversity within the team. Although thisgeneral conclusion was supported by ouranalyses for both sales and service teams, wecannot draw sweeping generalizations aboutthe relationships we studied because differ-ent patterns of results were found in thesetwo occupational groups.

Study 4: A Large Retail Company

David Levine and Jonathan Leonard at theUniversity of California-Berkeley HassSchool of Business and Aparna Joshi from atthe University of Illinois conducted Study 4.The data are drawn from a large retailer withlocations across the country. This studyasked whether workforces that reflect theracial makeup of the communities in whichthey are located perform better than thosethat do not reflect the makeup of the com-munity. It also examined whether diversity

We found noevidence for adirect relation-ship betweenteam gender orethnic diversityand the teamprocesses orperformance.

1 6

HUMAN RESOURCE MANAGEMENT, Spring 2003

These diversityproponentsfought againstemployers whoclaimed their(mostly White)customers caredabout the raceand gender ofthe employeeswho servedthem.

within the workplace had an effect on work-place performance. No data about teamprocesses or interactions among employeeswere available for this study.

Like most national chains, this firm or-ganizes a subset of these outlets into a singleentity that invests heavily in establishingbrand image. Most of the nonmanagerial em-ployees at work at any time are visible to thepublic, literally by looking through a window.These workplaces are located in every U.S.market of any significance. Each workplacetypically employs 15 to 40 part-time employ-ees with several full-time managers and as-sistant managers. The employees work scat-tered shifts through the week. Thus,employees work with a changing mix of theother employees on the payroll that month.In general, most frontline employees rotatethrough the several tasks in the store, spend-ing some of their time dealing with cus-tomers and other time in support tasks.

This study combined employee-level dataon demographics, store-level data on sales,and data from the 1990 U.S. Census oncommunity characteristics. The employeedata was the complete personnel recordsfrom February 1996 to October 1998. Datawere analyzed on frontline workplace em-ployees, dropping workplaces with fewerthan ten employees. The performance mea-sure was average sales at a store. A numberof employee, store, and community charac-teristics were included as controls.

The results can be summarized as follows:

1. Contrary to theories of customer dis-crimination, communities with moreWhites, Blacks, Hispanics, or Asiansdid not buy more from stores withsimilar employees.

2. The effects of diversity within a storeare more complex. Gender diversityhad no meaningful effect. At thesame time, stores with highly femaleworkforces sold less than moremixed stores.

3. Racial diversity has little effect onsales due to two offsetting effects.On the one hand, the index of racialdiversity (the odds that two employ-ees picked at random are the same

race) predicts higher sales. On theother hand, stores with more Whitessell more in this chain, and the pri-mary means of increasing diversity isto hire fewer Whites. These two ef-fects roughly cancel each other out.

In summary, this study finds no consis-tent evidence that most customers carewhether the salespeople who serve them areof the same race or gender. These results donot support some recent proponents of diver-sity who advocate diversity so as to satisfycustomers' desires to be served by those whophysically resemble them. Such argumentsmay still hold in other sectors, where rela-tionships last longer and involve a higherlevel of trust and communication betweencustomers and service providers. The resultsalso did not show any harm or benefits fromracial or gender diversity within the work-place. Again, we might see larger beneficialeffects in settings when employees have morediscretion and autonomy-so that they havemore scope to act on their group-specific in-formation. Conversely, the harm might belarger in settings where communicationamong employees is more important. At thesame time, these results are heartening forold-fashioned proponents of workplace inte-gration. These diversity proponents foughtagainst employers who claimed their (mostlyWhite) customers cared about the race andgender of the employees who served them.

Summary and Implications

The studies reported here were conducted inlarge firms that have well-deserved reputa-tions for their longstanding commitments tobuilding a diverse workforce and managing di-versity effectively. Each of these firms hastaken steps to ensure that its formal policiessupport and reinforce its diversity objectives.While their specific practices vary, our investi-gation clearly documents the importance andvalue of firm-wide, diversity-sensitive manage-rial strategies, human resource policies, andorganizational cultures. Despite the variabilityin industry contexts, specific practices, andthe performance measures we examined, ourquantitative results are strikingly similar.

We found that racial and gender diversitydo not have the positive effect on perfor-mance proposed by those with a more opti-mistic view of the role diversity can play inorganizations-at least not consistently orunder all conditions-but neither does itnecessarily have the negative effect on groupprocesses warned by those with a more pes-simistic view. Most analyses yielded no nega-tive effects on team processes at all, butwhen racial diversity was shown to have anegative effect, it was mitigated by trainingand development-focused initiatives. Genderdiversity had either no effects or positive ef-fects on team processes. This is consistentwith research that has shown that sex bal-anced groups have more positive interactionsthan either predominantly male or predomi-nantly female groups (Hoffman & Maier,1961; Wood, 1987).

There were few direct effects of diversityon performance-either positive or negative.Our findings suggest that, as we expected,this is likely because context is crucial in de-termining the nature of diversity's impact onperformance. Conditions that exacerbatedracial diversity's negative effects on perfor-mance included a highly competitive contextamong teams. Finally, there was some prom-ising evidence to suggest that, under certainconditions, racial diversity may even enhanceperformance, namely when organizationsfoster an environment that promotes learn-ing from diversity.

In general, we also found that gender di-versity was less problematic than racial di-versity. We expect that this may be, at least inpart, because in the companies in these stud-ies, women-typically White women-tended to be better represented than eithermen or women of color.

If these studies are representative ofother leading companies with similarly strongcommitments to diversity, our results maysuggest that efforts to create and manage di-verse workforces have generally paid off byeliminating many of the potentially negativeeffects of diversity on group processes andperformance documented previously in theliterature. Moreover, there appear to be someconditions under which diversity, if managedwell, may even enhance performance.

Effects of Diversity on Business Performance

An important goal of this research was toexplore the feasibility of conducting researchon diversity in organizational settings. Ourexperience demonstrated how difficult it is toconduct this type of field research and howlittle analytical attention practitioners pay tothese issues in organizations today. Few com-panies are equipped to assess the impact oftheir diversity efforts on their performance.One clear implication of our work is that or-ganizations need to do a better job of track-ing and evaluating the impact of their strate-gies for managing a diverse workforce.

Managerial Implications

What implications do we draw from thiswork for managers? Given the limited natureof our sample and our findings, it would beinappropriate to propose broad or sweepingimplications for managerial action. In thecourse of this project, however, we discussedthe state of practice with managers frommore than 20 large, well-known, and highlyregarded firms as we sought their involve-ment in our research. Through these discus-sions we obtained what we believe is a validpicture of the state of practice in managingdiversity in large organizations today. More-over, while our empirical research is limitedto four cases, to our knowledge, this researchrepresents the first effort to test a model re-lating diversity to performance in multiplefirms. Thus, with appropriate caution, weoffer the following implications for practice.

Modify the business case. The simplistic busi-ness case of the past is simply not supportedin our research. Our experience and findingsin these companies suggest that those whowant to invoke a business case to advancethe cause of diversity need to modify the waythey frame the argument. They should startby recognizing that there is virtually no evi-dence to support the simple assertion that di-versity is inevitably either good or bad forbusiness. Based on our findings, we proposea more nuanced view, which focuses on theconditions that can leverage benefits fromdiversity or, at the very least, mitigate its neg-ative effects. Our proposed reframing of thebusiness case for diversity follows.

1 7

There were fewdirect effects ofdiversity onperformance-either positiveor negative.

1 8 • HUMAN RESOURCE MANAGEMENT, Spring 2003

Human re-source man-agers and otherprofessionals incharge of diver-sity effortsshould take amore analyticalapproach inperformingtheir roles.

Diversity is a reality in labor markets andcustomer markets today. To be successfulin working with and gaining value fromthis diversity requires a sustained, systemicapproach and long-term commitment.Success is facilitated by a perspective thatconsiders diversity to be an opportunity foreveryone in an organization to learn fromeach other how better to accomplish theirwork and an occasion that requires a sup-portive and cooperative organizational cul-ture as well as group leadership andprocess skills that can facilitate effectivegroup functioning. Organizations that in-vest their resources in taking advantage ofthe opportunities that diversity offersshould outperform those that fail to makesuch investments.

Look beyond the business case. We believethis restatement of the business case accu-rately reflects both our results and the re-sults from prior laboratory studies. How-ever, our results may offer an even strongerimplication. It may be that the business-case rhetoric has run its course. Diversityprofessionals, industry leaders, and re-searchers might do better to recognize thatwhile there is no reason to believe diversitywill naturally translate into better or worseresults, diversity is both a labor-market im-perative and societal expectation and value.Therefore managers might do better tofocus on building an organizational cul-ture, human resource practices, and themanagerial and group process skills neededto translate diversity into positive organiza-tional, group, and individual results. Ourmore specific recommendations for doingso follow.

Adopt a more analytical approach. Despitethe widespread availability and use of humanresource information systems, we found thatbasic HR data about individuals or groupscould not be readily linked to business-levelperformance data. Unable to link HR prac-tices to business performance, HR practi-tioners will be limited in what they can learnabout how to manage diversity effectively,and their claims for diversity as a strategicimperative warranting financial investments

weakened accordingly. Human resourcemanagers and other professionals in chargeof diversity efforts should take a more ana-lytical approach in performing their roles.Sophisticated data collection and analysesare needed to understand the consequencesof diversity within organizations, and to mon-itor an organization's progress in managingdiversity effectively. Currently, organizationstypically assess their diversity efforts by sim-ply comparing attitudes, performance, ad-vancement, pay, and so on, among differentgroups of employees. These comparisons canbe useful, but they are only a first step.Equally important but very different ques-tions are: Under what conditions do workunits that are diverse with respect to genderor race outperform or underperform workunits that are more homogeneous? Whatconditions mitigate or exacerbate diversity'spotential negative or positive effects?

Support experimentation and evaluation.More work is needed to design and evaluatespecific interventions or experiments aimedat creating a positive link between diversityand performance. Of necessity, we relied onassessing this relationship by examining nat-ural covariations in diversity and perfor-mance across groups, but there were manyother factors that we could neither measurenor control, which may have influenced ourfindings and no doubt attenuated the size oftrue effects. Researchers who are better ableto isolate effects by studying them in thecontrolled setting of the laboratory tend tofind larger effects than we observed in thefield research on which we reported here.Studies that can better replicate these exper-imental conditions in real organizational set-tings would increase control without the ar-tificiality of the laboratory. To conduct suchresearch, however, will require executives tocommit to this type of experimentation andlearning within their own organizations.

Train for group-process skills. Our results sug-gest that training programs must help man-agers to develop the leadership and groupprocess skills needed to facilitate construc-tive conflict and effective communication.These are challenges that will inevitably arise

s

for managers who attempt to make diversitya resource for learning, change, and renewal.

In summary, we believe that progress inboth research and organizational practicewill come through continuing collaborativeefforts between researchers and managers asthey design and evaluate new approaches toleveraging workforce diversity. Training pro-grams that improve the skills of managersand team members may be particularly use-ful, but training alone is not likely to be suf-

ficient. Organizations must also implementmanagement and human resource policiesand practices that inculcate cultures of mu-tual learning and cooperation.

This research was conducted in partner-ship with the BOLD Initiative. Support wasprovided by the Alfred P. Sloan Foundationand the Society for Human Resource Man-agement Foundation. All views expressedhere are solely those of the authors.

THOMAS KOCHAN is the George M. Bunker Professor of Management at the MITSloan School of Management. His research focuses on the changing nature of workand employment relationships and their implications for organizational governanceand public policies. As Codirector of the MIT Workplace Center he is active in bring-ing together business, community, labor, and policy making groups to coordinate ef-forts to better integrate work and family life. His most recent book is Working inAmerica: A Blueprint for the New Labor Market.

KATERINA BEZxuxovA is an affiliated faculty member at the Wharton School of theUniversity of Pennsylvania. Her research focuses on workplace diversity, organiza-tional conflict, and group productivity. Her most recent work investigates the effectsof demographic faultlines in diverse workgroups.

ROBIN ELY is an associate professor of organizational behavior at the Harvard Busi-ness School. Her research focuses on how organizations can better manage their raceand gender relations while at the same time increasing their effectiveness. She is cur-rently exploring learning-the process of improving performance through interactionsthat advance knowledge and understanding-as a key mechanism through which di-versity enhances work group functioning.

SUSAN JACKSON is professor of human resource management and director of the doc-toral program in Industrial Relations and Human Resources at the School of Manage-ment and Labor Relations, Rutgers University. Her research interests include human re-source management systems, work team functioning, and workforce diversity. For moreinformation about this work, visit her web site at www.rci.rutgers.edu/~sjacksox.

APARNA JOSHI is assistant professor at the Institute of Labor and Industrial Relations,University of Illinois at Urbana-Champaign. She conducts research in the area of diver-sity in teams and organizations and in the area of management of global teams. She cur-rently teaches courses on Managing Workforce Diversity and Training and Development.

KAREN JEHN is professor of management at the Wharton School of the University ofPennsylvania. Her research focuses on the effects of diversity on group processes andon the effects of conflict on group performance.

JONATHAN LEONARD is professor of economics at the Haas School of Business at theUniversity of California, Berkeley. His research focuses on the effects of affirmativeaction on employment and on the effects of industrial relations practices and activi-ties on performance.

Effects of Diversity on Business Performance • 1 9

20 • HUMAN RESOURCE MANAGEMENT, Spring 2003

DAVID LEVINE is a professor at the Haas School of Business at the University of Cal-ifornia, Berkeley, where he is also editor of the journal Industrial Relations and Asso-ciate Director of the Institute of Industrial Relations. His research focuses on labormarkets and workplaces. His recent books are Reinventing the Workplace (1995); TheAmerican Workplace: Skills, Pay and Employee Involvement (1999); and Changes inCareers and Wage Structures at Large American Employers (2002).

DAVID THOMAS is the H. Naylor Fitzhugh Professor of Business Administration at theHarvard Business School. His research focuses on the influence of cultural diversityon organizational and individual effectiveness, minority executive development, andstrategic human resource management. His most recent book (with John Gabarro),Breaking Through: The Making of Minority Executives in Corporate America, was the2001 Academy of Management George R. Terry Book Award winner for outstandingcontribution to management theory and practice.

REFERENCES

Bantel, K.A., & Jackson, S.E. (1989). Top manage-ment and innovations in banking: Does thecomposition of the top team make a difference?Strategic Management Journal, 10 (SpecialIssue), 107-124.

Bezrukova, K., & Jehn, K.A. (2001). The effects ofdiversity training programs. Unpublished man-uscript. Philadelphia, PA: The Wharton School,University of Pennsylvania.

Cohen, H. (1998). Studying industries and theirpeople. Perspectives on Work, 2, 13-17.

Corporate Leadership Council. (1997). The state ofcorporate diversity initiatives. Washington, DC:The Advisory Board.

Doty, D.H., Glick, WH, & Huber, G.P. (1993). Fit,equifinality, and organizational effectiveness: Atest of two configurational theories. Academy ofManagement Journal, 36, 1196-1250.

Ely, R.J., & Thomas, D.A. (2001). Cultural diversityat work: The effects of diversity perspectives onwork group process and outcomes. Administra-tive Science Quarterly, 46, 229-73.

Hoffman, L., & Maier, N. (1961). Quality and ac-ceptance of problem solutions by members ofhomogeneous and heterogeneous groups. Jour-nal of Abnormal and Social Psychology, 62,401-407.

Ichniowski, C., Kochan, T.A., & Olson, C. (1998).Assessing the effects of HR on firms, individu-als and society: The role of the HR Network.Perspectives on Work, 2, 18-22.

Jackson, S. (1992). Team composition in organiza-tions. In S. Worchel, W Wood, & J. Simpson

(Eds.), Group process and productivity (pp.1-12). London: Sage.

Jackson, S.E., Brett, J.F., Sessa, V.1., Cooper, D.M.,Julin, J.A., & Peyronnin, K. (1991). Some dif-ferences make a difference: Individual dissimi-larity and group heterogeneity as correlates ofrecruitment, promotions, and turnover. Journalof Applied Psychology, 76, 675-689.

Jackson, S.E., & Joshi, A. (2001). Research on do-mestic and international diversity in organiza-tions. In N. Anderson, D.S. Ones, H.K. Sinangil,& C. Vishwesvaran (Eds.), Handbook of indus-trial, work and organizational psychology, Volume2 (pp. 206-231). Thousand Oaks, CA: Sage.

Jackson, S.E., & Joshi, A. (2002). Diversity in socialcontext: A multi-attribute, multi-level analysisof team diversity and performance in a sales or-ganization. Working paper, School of Manage-ment and Labor Relations, Rutgers University,New Brunswick, NJ.

Jackson, S.E., May, K.A., & Whitney, K. (1995). Un-derstanding the dynamics of diversity in deci-sion making teams. In R.A. Guzzo and E. Salas(Eds.), Team decision making effectiveness inorganizations (pp. 204-261). San Francisco:Jossey-Bass.

Jehn, K.A. (1995). A multimethod examination ofthe benefits and detriments of intragroup con-flict. Administrative Science Quarterly, 40,256-282.

Jehn, K.A., Neale, M., & Northcraft, G. (1999).Why differences make a difference: A fieldstudy of diversity, conflict, and performance inworkgroups. Administrative Science Quarterly,44, 741-763.

J

Jehn, K.A., & Werner, O. (1993). Hapax LegomenonII: Theory, a thesaurus, and word frequency.Cultural Anthropology Method, 5, 8-10.

Joshi, A. (2002). How does context matter? Examin-ing the process and performance outcomes ofwork team diversity. Dissertation, Rutgers Uni-versity, New Brunswick, NJ.

Millikin, F.J., & Martins, L.L. (1996). Searching{from?} common threads: Understanding themultiple effects of diversity in organizationalgroups. Academy of Management Review 21(2),402-433.

O'Reilly, C.A., & Flatt, S. (1989). Executive teamdemography: Organizational innovation andfirm performance. Working paper. Berkeley, CA:School of Business, University of California atBerkeley.

Richard, O.C. (2000). Racial diversity, businessstrategy, and firm performance: A resource-based view. Academy of Management Journal,43, 164-177.

Richard, O.C., & Johnson, N.B. (1999). Makingthe connection between formal human re-source diversity practices and organizationaleffectiveness: Beyond management fashion.Performance Improvement Quarterly, 12,77-96.

Effects of Diversity on Business Performance • 21

Richard, O.C., Kochan, T.A., & Mcmillan-Capehart,A. (2002). The impact of visible diversity on or-ganizational effectiveness: Disclosing the con-tents in Pandora's black box. Journal of Busi-ness and Management, 8, 1-26.

Steiner I.D. (1972). Group process and productivity.San Diego: Academic Press.

Thomas, D., & Ely, R. 1996. Making differences matter:A new paradigm for managing diversity. HarvardBusiness Review, September-October, 79-91.

Williams, K., & O'Reilly, C.A. (1998). Demographyand diversity: A review of 40 years of research.In B. Stave and R. Sutton (Eds.), Research inorganizational behavior, Volume 20 (pp.77-140). Greenwich, CT: JAI Press.

Wood, W (1987). Meta-analytic review of sex differ-ences in group performance. Psychological Bul-letin, 102, 53-71.

Wright, P., Ferris, S.P, Hiller, J.S., & Kroll, M.(1995). Competitiveness through managementof diversity: Effects on stock price valuation.Academy of Management Journal, 38, 272-287.

ENDNOTES

1. The details of our analyses and results are avail-able upon request.