Proponents of the value-in-diversity per-spective often make the business case fordiversity (e.g., Cox 1993). These scholars claimthat diversity pays and represents a com-pelling interestan interest that meets cus-tomers needs, enriches ones understanding ofthe pulse of the marketplace, and improves thequality of products and services offered (Cox1993; Cox and Beale 1997; Hubbard 2004;Richard 2000; Smedley, Butler, and Bristow2004). Moreover, diversity enriches the work-place by broadening employee perspectives,strengthening their teams, and offering greaterresources for problem resolution (Cox 2001).The creative conflict that may emerge leads to

closer examination of assumptions, a more com-plex learning environment, and, arguably, bet-ter solutions to workplace problems (Gurin,Nagda, and Lopez 2004). Because of such puta-tive competitive advantages, companies increas-ingly rely on a heterogeneous workforce toincrease their profits and earnings (Florida andGates 2001, 2002; Ryan, Hawdon, and Branick2002; Williams and OReilly 1998).

Critics of the diversity model, however, areskeptical about the extent to which these bene-fits are real (Rothman, Lipset, and Nevitte2003a, 2003b; Skerry 2002; Tsui, Egan, andOReilly 1992; Whitaker 1996). Scholars whosee diversity as process loss argue that diver-sity incurs significant potential costs (Jehn,Northcraft, and Neale 1999; Pelled 1996; Pelled,Eisenhardt, and Xin 1999). Skerry (2002), forinstance, points out that racial and ethnic diver-sity is linked with conflict, especially emotion-al conflict among co-workers. Tsui andcolleagues (1992) concur, suggesting that diver-sity diminishes group cohesiveness and, as aresult, employee absenteeism and turnoverincrease. Greater diversity may also be associ-ated with lower quality because it can lead topositions being filled with unqualified workers(Rothman et al. 2003b; see also Williams andOReilly 1998). For these reasons, skeptics

Does Diversity Pay?: Race, Gender, andthe Business Case for Diversity

Cedric HerringUniversity of Illinois at Chicago

The value-in-diversity perspective argues that a diverse workforce, relative to a

homogeneous one, is generally beneficial for business, including but not limited to

corporate profits and earnings. This is in contrast to other accounts that view diversity

as either nonconsequential to business success or actually detrimental by creating

conflict, undermining cohesion, and thus decreasing productivity. Using data from the

1996 to 1997 National Organizations Survey, a national sample of for-profit business

organizations, this article tests eight hypotheses derived from the value-in-diversity

thesis. The results support seven of these hypotheses: racial diversity is associated with

increased sales revenue, more customers, greater market share, and greater relative

profits. Gender diversity is associated with increased sales revenue, more customers, and

greater relative profits. I discuss the implications of these findings relative to alternative

views of diversity in the workplace.

AMERICAN SOCIOLOGICAL REVIEW, 2009, VOL. 74 (April:208224)

Direct all correspondence to Cedric Herring,Department of Sociology (MC 312), University ofIllinois at Chicago, 1007 W. Harrison, Chicago, IL60607 (Herring@uic.edu). I would like to thankWilliam Bielby, John Sibley Butler, Sharon Collins,Paula England, Tyrone Forman, Hayward DerrickHorton, Loren Henderson, Robert Resek, MosheSemyonov, Kevin Stainback, and the editors of ASRand the anonymous reviewers for their extraordinar-ily helpful comments and suggestions. An earlierversion of this article was presented at the AnnualConference of the American Sociological Associationin Montreal, Canada in August 2006.

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

question the real impact of diversity programson businesses bottom line.

Is it possible that diversity has dual outcomes,some of which are beneficial to organizationsand some of which are costly to group func-tioning? Diversity may be valuable even ifchanges in an organizations composition makeincumbent members uncomfortable. AsDiTomaso, Post, and Parks-Yancy (2007:488)point out, research generally finds that het-erogeneity on most any salient social categorycontributes to increased conflict, reduced com-munication, and lower performance, at the sametime that it can contribute to a broader range ofcontacts, information sources, creativity, andinnovation. This suggests that diversity may beconducive to productivity and counterproduc-tive in work-group processes. Many of theclaims and hypotheses about diversitys impactshave not been examined empirically, so it is notclear what effect, if any, diversity has on theoverall functioning of organizations, especial-ly businesses.

In this article, I empirically examine keyquestions pertaining to organizational diversi-ty and its implications. Does diversity offer themany benefits suggested by the value-in-diver-sity thesis? Or, do costs offset potential bene-f its? Perhaps diversity is simultaneouslyassociated with the twin outcomes of group-level conflict and increased performance at theestablishment level. Although no singularresearch design could adjudicate between allthe claims of proponents and skeptics,1 the ques-tions I raise warrant serious examination giventhe growing heterogeneity of the U.S. work-force. Using data from the 1996 to 1997National Organizations Survey, my analysesexplore the relationship between racial and gen-der workforce diversity and several indicatorsof business performance, such as sales revenue,

number of customers, relative market share,and relative profitability.

VALUE IN DIVERSITY

The definition of diversity is unclear, asreflected in the multiplicity of meanings in theliterature. For some, the term provokes intenseemotional reactions, bringing to mind suchpolitically charged ideas as affirmative actionand quotas. These reactions stem, in part,from a narrow focus on protected groups cov-ered under affirmative action policies, wheredifferences such as race and gender are the focalpoint. Some alternative definitions of diversityextend beyond race and gender to include alltypes of individual differences, such as ethnic-ity, age, religion, disability status, geographiclocation, personality, sexual preferences, and amyriad of other personal, demographic, andorganizational characteristics. Diversity canthus be an all-inclusive term that incorporatespeople from many different classifications.Generally, diversity refers to policies andpractices that seek to include people who areconsidered, in some way, different from tradi-tional members. More centrally, diversity aimsto create an inclusive culture that values and usesthe talents of all would-be members.

The politics surrounding diversity and inclu-sion have shifted dramatically over the past 50years (Berry 2007). Title VII of the 1964 CivilRights Act makes it illegal for organizations toengage in employment practices that discrimi-nate against employees on the basis of race,color, religion, sex, or national origin. This actmandates that employers provide equal employ-ment opportunities to people with similar qual-ifications and accomplishments. Since 1965,Executive Order 11246 has required govern-ment contractors to take affirmative action toovercome past patterns of discrimination. Thesedirectives eradicated policies that formally per-mitted discrimination against certain classes ofworkers. They also increased the costs to organ-izations that fail to implement fair employmentpractices.

By the late 1970s and into the 1980s, therewas growing recognition within the private sec-tor that these legal mandates, although neces-sary, were insufficient to effectively manageorganizational diversity. Many companies andconsulting firms soon began offering training

DOES DIVERSITY PAY?209

1 It is possible to derive propositions from argu-ments against the business case for diversity thesis,but works in this skeptical vein typically provide cri-tiques of more optimistic views of diversity, ratherthan systematic, alternative theoretical formulations.When skeptics do put forth testable hypotheses, theytend to focus on intermediary processes and mech-anisms rather than the diversitybusiness bottomline linkage, per se. I cite such critiques to show thatthere are reasons to be skeptical about the linkbetween diversity and business performance.

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

programs aimed at valuing diversity. In theirstudy of private-sector establishments from1971 to 2002, Kalev, Dobbin, and Kelly (2006)find that programs designed to establish organi-zational responsibility for diversity are moreefficacious in increasing the share of whitewomen, black women, and black men in man-agement than are attempts to reduce manageri-al bias through diversity training or reduce socialisolation through mentoring women and racialand ethnic minorities. They also f ind thatemployers who were subject to federal affir-mative action edicts are likely to have diversi-ty programs with stronger effects.

During the 1990s, diversity rhetoric shiftedto emphasize the business case for workforcediversity (Bell and Hartmann 2007; Berry 2007;Hubbard 1997, 1999; von Eron 1995).Managing diversity became a business neces-sity, not only because of the nature of labormarkets, but because a more diverse workforcewas thought to produce better business results.Exploiting the nations diversity was viewed askey to future prosperity. Ignoring the fact thatdiscrimination limits a societys potentialbecause it leads to underutilization of talentpools was no longer practical nor feasible.Diversity campaigns became part of the attemptto strengthen the United States and movebeyond its history of discrimination by pro-viding previously excluded groups greateraccess to educational institutions and work-places (Alon and Tienda 2007). Today, advo-cates are asked to find evidence to support thebusiness-case argument that diversity expandsthe talent pool and strengthens U.S. institu-tions. Even if the shift from affirmative actionto diversity has tamed what began as a radi-cal fight for equality (Berry 2007), workforcediversity has become an essential business con-cern in the twenty-first century.

THE IMPLICATIONS OF DIVERSITY FORWORKPLACE DYNAMICS AND BUSINESSOUTCOMES

How might diversity affect business outcomes?Page (2007) suggests that groups displaying arange of perspectives outperform groups oflike-minded experts. Diversity yields superioroutcomes over homogeneity because progressand innovation depend less on lone thinkerswith high intelligence than on diverse groups

working together and capitalizing on their indi-viduality. The best group decisions and predic-tions are those that draw on unique qualities. Inthis regard, Bunderson and Sutcliffe (2002)show that teams composed of individuals witha breadth of functional experiences are well-suit-ed to overcoming communication barriers: teammembers can relate to one anothers functionswhile still realizing the performance benefits ofdiverse functional experiences. Williams andOReilly (1998) and DiTomaso and colleagues(2007) concur, arguing that diversity increasesthe opportunity for creativity and the quality ofthe product of group work.

The benefits of diversity may extend beyondteam and workplace functioning and problemsolving. Sen and Bhattacharya (2001), forinstance, propose that diversity influences con-sumers perceptions and purchasing practices.Indeed, Black, Mason, and Cole (1996) findthat consumers have strongly held in-grouppreferences when a transaction involves signif-icant customerworker interaction. Richard(2000) argues that cultural diversity, within theproper context, provides a competitive advan-tage through social complexity at the firm level.Irrespective of the specific processes, diversi-ty may positively influence organizations func-tioning, net of any internal work-groupprocesses that diversity may impede.

The sociological literature on diversity con-tinues to grow (e.g., Alon and Tienda 2007;Bell and Hartmann 2007; Berry 2007;DiTomaso et al. 2007; Embrick forthcoming;Kalev et al. 2006), but, to date, there has beenlittle systematic research on the impact of diver-sity on businesses financial success. Few stud-ies use quantitative data and objectiveperformance measures from real organizationsto assess hypotheses. One exception comparescompanies with exemplary diversity manage-ment practices with those that paid legal dam-ages to settle discrimination lawsuits. The resultsshow that the exemplary firms perform better,as measured by their stock prices (Wright et al.1995).

A second exception is a series of studiesreported by Kochan and colleagues (2003).They find no significant direct effects of eitherracial or gender diversity on business perform-ance. Gender diversity has positive effects ongroup processes while racial diversity has neg-ative effects. The negative relationship between

210AMERICAN SOCIOLOGICAL REVIEW

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

racial diversity and group processes, however,is largely absent in groups with high levels oftraining in career development and diversitymanagement. These scholars also find that racialdiversity is positively associated with growth inbranches business portfolios. Racial diversityis associated with higher overall performance inbranches that enact an integration-and-learn-ing perspective on diversity, but employee par-ticipation in diversity education programs hasa limited impact on performance. Finally, theyfind no support for the idea that diversity thatmatches a firms client base increases sales bysatisfying customers desire to interact withthose who physically resemble them.2

DEMOGRAPHIC DIVERSITY ANDORGANIZATIONAL FUNCTIONING

Researchers studying demographic diversitytypically take one of two approaches in theirtreatment of the subject. One approach treatsdiversity broadly, making statements about het-erogeneity or homogeneity in general, ratherthan about particular groups (e.g., Hambrickand Mason 1984). The alternative treats eachdemographic diversity variable as a distinct the-oretical construct, based on the argument thatdifferent types of diversity produce differentoutcomes (e.g., Hoffman and Maier 1961; Kentand McGrath 1969). Instead of assuming that alltypes of diversity produce similar effects, theseresearchers build their models around specifictypes of demographic diversity (e.g., Zengerand Lawrence 1989). Indeed, Smith and col-leagues (2001) argue that scholars should notlump women and racial minorities together asa standard approach to research. The relativenumber, power, and status of various groupswithin organizations can significantly affectissues such as favoritism and bias, and aggre-gating these groups may mask such variations.

The results of research on heterogeneity ingroups suggest that diversity offers a greatopportunity for organizations and an enormouschallenge. More diverse groups have the poten-tial to consider a greater range of perspectivesand to generate more high-quality solutionsthan do less diverse groups (Cox, Lobel, andMcLeod 1991; Hoffman and Maier 1961;Watson, Kumar, and Michaelsen 1993). Yet, thegreater the amount of diversity in a group or anorganizational subunit, the less integrated thegroup is likely to be (OReilly, Caldwell, andBarnett 1989) and the higher the level of dis-satisfaction and turnover (Jackson et al. 1991;Wagner, Pfeffer, and OReilly 1984). Diversitysimpact thus appears paradoxical: it is a double-edged sword, increasing both the opportunity forcreativity and the likelihood that group memberswill be dissatisfied and will fail to identify withthe organization. Including influential andpotentially confounding demographic and organi-zational factorsas I do in the following analy-seshelps clarify the relationship betweendiversity and organizational functioning, espe-cially in business organizations.

ALTERNATIVE EXPLANATIONS

For-profit workplaces are appropriate sites forexamining questions about diversity, as theyare where employment decisions are made andthe settings in which work is performed (Baronand Bielby 1980; Reskin, McBrier, and Kmec1999). Moreover, it is organizational processesthat perpetuate segregation and influence thecharacter of jobs and workplaces (Tomaskovic-Devey 1993).

In assessing the relationship between diver-sity and business outcomes, the literature sug-gests that there are several organizational factorsthat might be influential. According to the insti-tutional perspective in organizational theory,organizational behavior is a response to pres-sures from the institutional environment(Stainback, Robinson, and Tomaskovic-Devey2005). The institutional environment of anorganization embodies regulative, normative,and cultural-cognitive institutions affecting theorganization, such as law and social attitudes(Scott 2003).

According to this formulation of organiza-tional behavior, adoption of new organization-al practices is often an attempt to gain legitimacy

DOES DIVERSITY PAY?211

2 It would be preferable to model an array of inter-nal processes in organizations to establish how diver-sity affects business performance (e.g., functionalityof work teams, marketing decisions, or innovation inproduction or sales), but this is not possible withdata from the National Organizations Survey becausesuch indicators are not available. These topics dooffer potentially fruitful paths for future research.

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

in the eyes of important constituents, and notnecessarily an attempt to gain greater efficien-cy (DiMaggio and Powell 2003). Based on insti-tutional theory, for-profit businesses that areaccountable to a larger public may be more sen-sitive to public opinion on what constituteslegitimate organizational behavior. Publicly-held, for-profit businesses employment prac-tices should thus be more subject to publicscrutiny. These businesses should employ rela-tively more minorities than private-sectoremployers, to the degree that they are undergreater pressure to achieve racial and ethnicdiversity, as public sentiment views such poli-cies as a necessary element of legitimate organi-zational governance (Edelman 1990).

Organizational size is also important since itis positively related to sophisticated personnelsystems (Pfeffer 1977) that may contribute todiversity in the workplace. Large organizationsconcerned about due process and employmentpractices will institute specific offices and pro-cedures for handling employee complaints(Gwartney-Gibbs and Lach 1993; Welsh,Dawson, and Nierobisz 2002). At the same time,if some organizations, when left to their owndevices, prefer hiring whites over racial minori-ties, their larger size and slack resources givethem more ability to indulge preferences forwhite workers (Cohn 1985; Tolbert andOberfield 1991).

Large establishments also tend to make greaterefforts at prevention and redress because theyhave direct legal obligations. Antidiscriminationlaws make discrimination against minorities andwomen potentially costly, but not all establish-ments are subject to these laws. Federal law ban-ning sex discrimination in employment exemptsfirms with fewer than 15 workers, and enforce-ment efforts often target large firms (Reskin etal. 1999). Moreover, affirmative action regula-tions apply only to firms that do at least $50,000worth of business with the federal governmentand have at least 50 employees (Reskin 1998).Establishment size may thus be related to vul-nerability to equal employment opportunity andaffirmative action regulations, which in turnshould be related to increased racial and ethnicdiversity. Indeed, Holzer (1996) shows that affir-mative action implementation has led to gains inthe representation of African Americans andwhite women in firms required to practice affir-mative action.

Stinchcombe (1965) offers a rationale for whyestablishment age may also be meaningful. Hesuggests a liability of newness, whereby organi-zational mortality rates decrease with organiza-tional age. Younger organizations are more proneto mortality than are older organizations, so theyapproach threats to their existence differently. Itis possible, therefore, that organizations of dif-ferent ages vary in their responses to racial andgender diversity concerns.

The labor pools from which establishmentshire may be important relative to the effects ofdiversity on business outcomes. The sex andracial composition of regional labor markets mayinfluence an establishments composition, as wellas its relative success (Blalock 1957). Regionaldifferences in residential segregation, however,may obscure regional effects of demographiccomposition (Jones and Rosenfeld 1989).

Finally, industrial-sector variations may relatesimultaneously to levels of diversity and busi-ness performance. In particular, organizationsin the service sector will be more proactivewith regard to racial and gender diversity thanwill those that produce tangible goods, as theirperformance depends more on public goodwill.There are also reasons, however, to believe thatthe economic sector in which businesses oper-ate can matter. Compared with manufacturingand public service, service-sector establish-ments are more likely to exclude blacks, espe-cially black men, by using personality traits andappearance as job qualifications (Moss andTilly 1996).

DIVERSITY, BUSINESSPERFORMANCE, ANDEXPECTATIONS

Evidence that directly assesses the businesscase for diversity has, at best, proven elusive.Building on the present discussion, I offer sev-eral predictions. In its most basic form, thebusiness case for diversity perspective predictsa return on investment for diversity (Hubbard2004). It is thus fairly easy to derive somestraightforward expectations:

Hypothesis 1a: As racial workforce diversityincreases, a business organizations salesrevenues will increase.

Hypothesis 1b: As gender workforce diversityincreases, a business organizations salesrevenues will increase.

212AMERICAN SOCIOLOGICAL REVIEW

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

Hypothesis 2a: As racial workforce diversityincreases, a business organizations numberof customers will increase.

Hypothesis 2b: As gender workforce diversityincreases, a business organizations numberof customers will increase.

Hypothesis 3a: As racial workforce diversityincreases, a business organizations marketshare will increase.

Hypothesis 3b: As gender workforce diversityincreases, a business organizations marketshare will increase.

Hypothesis 4a: As racial workforce diversityincreases, a business organizations profitsrelative to its competitors will increase.

Hypothesis 4b: As gender workforce diversityincreases, a business organizations profitsrelative to its competitors will increase.

It is possible that racial and gender diversityin the workforce are related to some outcomesbut not others. My analyses thus incorporate anarray of tangible outcomes and benefits sur-rounding sales revenue, customer base, marketshare, and relative profitability. Following theliterature, I also examine relations betweendiversity and business outcomes net of other fac-tors (e.g., legal form of organization, establish-ment size, company size, organization age,industrial sector, and region).

DATA AND METHODS

The data come from the 1996 to 1997 NationalOrganizations Survey (NOS) (Kalleberg,Knoke, and Marsden 2001), which containsinformation from 1,002 U.S. work establish-ments, drawn from a stratified random sampleof approximately 15 million work establish-ments in Dun and Bradstreets InformationServices data file. I use data from the 506 for-profit business organizations that providedinformation about the racial composition oftheir full-time workforces, their sales revenue,their number of customers, their market share,and their profitability. The NOS concentrates onU.S. work establishments employment con-tracts, staffing methods, work organization, jobtraining programs, and employee benefits andincentives. The data include additional infor-mation about each organizations formal struc-ture, social demography, environmentalsituation, and productivity and performance.

The resulting sample is representative of U.S.profit-making work organizations. For eachorganization sampled, Dun and BradstreetsMarket Identifiers Plus service provides sever-al important pieces of information: companyname, address, and telephone number; size (interms of number of employees and sales vol-ume); year started; and business trends for thepast three years. In addition, historical infor-mation on the sampled organizations is availablefrom the Duns Historical Files. The unit ofanalysis is the workplace.

DIVERSITY

There are two basic approaches to measuringdiversity, either globally or as distinct indicators.Following Skaggs and DiTomaso (2004), I optfor separate but parallel indicators of racial andgender diversity. There are several reasons forthis. Previous research shows that race and gen-der as bases for diversity are extremely impor-tant in understanding human transactions. Formost people, these group identities are not eas-ily changeable. In addition, the base of knowl-edge in the social sciences is more fullydeveloped for these identities than for others thatmay be relevant. On a pragmatic level, indica-tors for these two dimensions are readily avail-able in the data source, while others are not(e.g., sexual preference or age).

The specific indicator draws from the RacialIndex of Diversity (RID) (Bratter and Zuberi2001; Zuberi 2001), which provides an unbiasedestimator of the probability that two individu-als chosen at random and independently fromthe population will belong to different racialgroups. The index ranges from 0 to 1. A scoreof 0 indicates a racially homogenous population;1 indicates a population where, given how raceis distributed, every randomly selected pair iscomposed of persons from two different racialgroups.

RID = 1 ( ni (ni 1))

N (N 1)

1 (1/i)

In this formula, N is the total population, nidenotes that population is separate racial group(i), and i refers to number of racial groups. TheRID is a measure of the concentration of racialclassification in a population if the population

DOES DIVERSITY PAY?213

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

to which each individual belongs is consideredto be one racial group, and n1 .|.|. ni is the num-ber of individuals in the various groups (so that3ni = N for a population with three racialgroups). If all the individuals in a racial groupare concentrated in one category, the measurewould be .0. To constrain the RID between .0and 1.0, I calculate the actual level of diversityas a proportion of the maximum level possiblewith the specified number of races, that is, 1 (1/i). In a population with an even racial distri-bution, the RID would equal 1.00.

One limitation of the RID (as with mostindexes of dissimilarity) is that it treats over-representation and underrepresentation of sub-groups as mathematically equivalent forcomputational purposes. Arguably, an organi-zation with a 20 percent underrepresentation ofminorities (relative to the population) would bevery different from an organization with a 20percent overrepresentation of minorities (rela-tive to the population).

To correct for this limitation, I modify theRID to incorporate the idea that power rela-tions between superordinate and subordinategroups are asymmetrical. The key idea is thatof parity. The Asymmetrical Index of Diversity(AID) is

AID = (1 S) if S > P

AID = (1 P) if P S

where S is the proportion of the organizationcomposed of the superordinate group, and P isthe proportion of the population composed ofthe superordinate group.

Scores can range from 0 (completely homo-geneous organization composed of the super-ordinate group) to parity (i.e., 1 P) (wherethe superordinate and subordinate groups havereached proportional representation in theorganization). For example, if the superordi-nate group constitutes 75 percent of the popu-lation, AID scores can range from 0 to .25 (or25 when multiplied by 100). If the superordinategroup constitutes 50 percent of the population,AID scores can range from 0 to .50 (or 50 whenmultiplied by 100).

The racial diversity index and the genderdiversity index are both operationalized usingtwo alternative premises: (1) without the under-lying notion of parity (i.e., RID) and (2) with anunderlying notion of parity mattering (i.e., AID-

R and AID-G). Both models work well, but theAID-R and AID-G models with the parityassumption give a better fit. Moreover, theyprovide a theoretical rationale for why diversi-ty is related to business outcomes. Given theseresults, I use both the AID-R and AID-G (par-ity) operationalizations (for both race and gen-der).

In this study, the superordinate racial group,whites, make up 75 percent of the population inthe 1996 to 1997 NOS. Scores on the asym-metrical index of diversity for race thus rangefrom a low of 0 (homogenous white) to a highof 25 (racial parity). Males, the superordinategender, make up 54 percent of the populationin the 1996 to 1997 NOS. Scores on the asym-metrical index of diversity for gender thus rangefrom a low of 0 (homogenous male) to a highof 46 (gender parity).

BUSINESS PERFORMANCE

To measure average annual sales revenue,respondents were asked, What was your organi-zations average annual sales revenue for the pasttwo years? Responses range from 0 to 60 bil-lion dollars. For multivariate analysis, I add asmall number (.01) to each observation beforedividing it by 1 million and taking the log ofthese values.

To measure the number of customers, respon-dents were asked, About how many customerspurchased the organizations products or ser-vices over the past two years? Responses rangefrom 0 to 25 million customers.

To measure perceived market share, respon-dents were asked, Compared to other organi-zations that do the same kind of work, howwould you compare your organizations per-formance over the past two years in terms ofmarket share? .|.|. Would you say that it was (1)much worse, (2) somewhat worse, (3) about thesame, (4) somewhat better, or (5) much bet-ter?

To measure relative profitability, respondentswere asked, Compared to other organizationsthat do the same kind of work, how would youcompare your organizations performance overthe past two years in terms of profitability? .|.|.Would you say that it was (1) much worse, (2)somewhat worse, (3) about the same, (4) some-what better, or (5) much better?

214AMERICAN SOCIOLOGICAL REVIEW

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

CONTROLS

Respondents were asked about their establish-ments legal form of organization (if for prof-it). Specifically, What is the legal form oforganization? Is it a sole proprietorship, part-nership or limited partnership, franchise, cor-poration with publicly held stock, corporationwith privately held stock, or something else?Responses are dummy coded for proprietor-ship, partnership, franchise, and corporation. Iexclude not-for-profit organizations becausethe focus of this study is on the business casefor diversity.

To determine organization size, respondentswere asked about the number of full-time andpart-time employees who work at all locationsof the business. Responses range from 15 to300,000 employees. Respondents were alsoasked about the establishment size (i.e., thenumber of full-time and part-time employees onthe payroll at their particular address).Responses are coded from 1 to 42,000.

To determine an organizations age, respon-dents were asked, In what year did the organ-ization start operations? The differencebetween the year of the survey and the year ofestablishment yields the organization age, rang-ing from 1 to 361 years.

To measure industry, respondents were askedabout the main product or service provided attheir establishments. Responses are dummycoded into the following 12 categories: agri-culture; mining; construction; transportationand communications; wholesale; retail; finan-cial services, insurance, and real estate; businessservices; personal services; entertainment; pro-fessional services; and manufacturing.

The region in which the establishment islocated is dummy coded into four categories:Northeast (CT, DE, MA, ME, NH, NJ, NY, PA,RI, and VT), Midwest (IA, IL, IN, KS, MI, MN,MO, ND, NE, OH, OK, SD, and WI), South(AL, AR, DC, FL, GA, KY, LA, MD, MS, NC,SC, TN, TX, VA, and WV), and West and oth-ers (AK, AZ, CA, CO, HI, ID, MT, NM, NV,OR, UT, WA, and WY).

RESULTS

How is diversity related to organizations busi-ness performance? Does a relationship existbetween the racial and gender composition of

an establishment and its sales revenue, numberof customers, market share, or profitability?

Figure 1 illustrates the distribution of estab-lishments with low, medium, and high levels ofracial and gender diversity. Thirty percent ofbusinesses have low levels of racial diversity, 27percent have medium levels, and 43 percenthave high levels. Regarding gender diversity, 28percent of businesses have low levels, 28 per-cent have medium levels, and 44 percent havehigh levels.

Table 1 presents means and percentage dis-tributions of various business outcomes of estab-lishments by their levels of diversity. Averagesales revenues are associated with higher levelsof racial diversity: the mean revenues of organ-izations with low levels of racial diversity areroughly $51.9 million, compared with $383.8million for those with medium levels and $761.3million for those with high levels of diversity.The same pattern holds true for sales revenueby gender diversity: the mean revenues of organ-izations with low levels of gender diversity areroughly $45.2 million, compared with $299.4million for those with medium levels and $644.3million for those with high levels of diversity.

Higher levels of racial and gender diversityare also associated with greater numbers of cus-tomers: the average number of customers fororganizations with low levels of racial diversi-ty is 22,700. This compares with 30,000 forthose with medium levels of racial diversityand 35,000 for those with high levels. The meannumber of customers for organizations withlow levels of gender diversity is 20,500. Thiscompares with 27,100 for those with mediumlevels of gender diversity and 36,100 for thosewith high levels.

Table 1 also shows that businesses with highlevels of racial (60 percent) and gender (62 per-cent) diversity are more likely to report higherthan average percentages of market share thanare those with low (45 percent) or medium lev-els of racial (59 percent) and gender (58 percent)diversity. A similar pattern emerges for organ-izations reporting higher than average prof-itability. Less than half (47 percent) oforganizations with low levels of racial diversi-ty report higher than average profitability. Incontrast, about two thirds of those with medi-um and high levels of racial diversity reporthigher than average profitability. Also, organi-zations with high levels of gender diversity (62

DOES DIVERSITY PAY?215

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

percent) are more likely to report higher thanaverage profitability than are those with low(45 percent) or medium levels (58 percent) ofgender diversity.

Although interesting, the descriptive statisticsdo not provide much information about the netrelationship between diversity and organiza-tions business performance.3 To address this

216AMERICAN SOCIOLOGICAL REVIEW

Figure 1. Percentage Distribution of Racial and Gender Diversity Levels in Establishments

Table 1. Means and Percentage Distributions for Business Outcomes of Establishments by Levelsof Racial and Gender Diversity

Racial Diversity Level Gender Diversity Level

Low Medium High Low Medium High Characteristics (

issue more rigorously, Tables 2 and 3 presentresults from multivariate analysis. Table 2shows the relationship between racial and gen-der diversity in establishments and logged salesrevenue, number of customers, estimates ofrelative market shares, and estimates of relativeprofitability. Model 1 shows that diversity andsales revenues are positively related. Diversityaccounts for roughly 6 percent of the variancein sales revenue. These results are fully con-sistent with Hypotheses 1a and 1b. Model 2shows that racial and gender diversity are alsosignificantly related to the number of cus-tomers. As the racial and gender diversity inestablishments increase, their number of cus-tomers also increases.

Diversity accounts for less than 4 percent ofthe variance in number of customers, but therelationship is statistically significant for bothracial and gender diversity. These results areconsistent with Hypotheses 2a and 2b. Model3 in Table 2 shows a positive relationshipbetween racial and gender diversity and esti-mates of relative market shares. As diversity inestablishments increases, estimates of relativemarket share also increase significantly. Therelationship is statistically significant for racialdiversity and marginally significant for genderdiversity. These results are consistent withHypotheses 3a and 3b. Finally, Model 4 displaysthe relationship between racial and gender diver-sity and estimates of relative profitability. As

diversity increases, estimates of relative prof-itability also increase. The results are statisticallysignificant and consistent with Hypotheses 4aand 4b.

But do these results hold up once alternativeexplanations are taken into account? Table 3presents the same relationships as Table 2, butit includes controls for alternative explanations.Model 1 of Table 3 presents the relationshipbetween racial and gender diversity in estab-lishments and logged sales revenue. In Model1, the relationships between racial and genderdiversity and sales revenues remain significant(p < .05), net of controls for legal form of organ-ization, company size, establishment size, organ-ization age, industrial sector, and region. Model2 shows that a one unit increase in racial diver-sity increases sales revenues by approximately9 percent; a one unit increase in gender diver-sity increases sales revenues by approximately3 percent. Combined, these factors account for16.5 percent of the variance in sales revenue.The results in Table 3 provide support forHypotheses 1a and 1b (i.e., as a business organi-zations racial and gender diversity increase, itssales revenue will also increase).

Model 1 of Table 3 examines alternate expla-nations of sales revenue. Net of all other factors,privately-held corporations have significantlylower sales revenues than do other legal formsof business. No other legal forms depart sig-nificantly from the omitted category. Model 1also shows that sales revenues increase mar-ginally as company size increases, and signifi-cantly as establishment size increases andorganizations age.

The standardized coefficients for this model(not reported in Table 3) show that a one stan-

DOES DIVERSITY PAY?217

Table 2. Regression Equations Predicting Sales, Number of Customers, Market Share, and RelativeProfitability with Racial and Gender Diversity

Model 1 Model 2 Model 3 Model 4

Independent Variables Sales Customers Market Share Profitability

Constant 4.847*** 12111.880*** 3.364*** 3.361***Racial diversity .087*** 509.398*** .009*** .008**Gender diversity .042*** 278.971*** .004* .006**R2 .058*** .038*** .021*** .025***N 506 506 506 470

Notes: Coefficients are unstandardized. For the dummy (binary) variable coefficients, significance levels refer tothe difference between the omitted dummy variable category and the coefficient for the given category.* p < .1; ** p < .05; *** p < .01.

Different researchers can offer several hypothesesabout the mechanisms that produce the observedrelationship between variables (in this case, diversi-ty and business outcomes).

at University of Strathclyde on March 13, 2015asr.sagepub.comDownloaded from

dard deviation increase in racial diversity pro-duces a .188 standard deviation increase in salesrevenue. Furthermore, the relationship betweenracial diversity and sales revenue (Beta = .188)is stronger than the impact of company size(Beta = .067), establishment size (Beta = .087),and organization age (Beta = .077). Genderdiversity is also important to sales revenue. Itmaintains a statistically significant relationshipto sales revenue (Beta = .082), net of controls.Therefore, not only are racial and gender diver-sity significantly related to sales revenue, butthey are among its most important predictors.

Model 2 in Table 3 presents the relationshipbetween racial and gender diversity in estab-lishments and number of customers. This rela-tionship remains statistically significant (p /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Average /GrayImageResolution 150 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Average /MonoImageResolution 300 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /DocumentCMYK /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ]>> setdistillerparams> setpagedevice