best ppractices oor bbest gguesses? assessing tthe ... · diversity lead to the broadest increases...

Lists of “best practices” in diversity man-agement have proliferated recently.

Everyone seems to have a list, from the EqualEmployment Opportunity Commission (1998)

to the Presidential Glass Ceiling Commission(1995), the women’s business advocacy groupCatalyst (1998), and the Society for HumanResources Management (2004). These lists are

Best PPractices oor BBest GGuesses? Assessing tthe EEfficacy oof CCorporate Affirmative AAction aand DDiversity PPolicies

Alexandra Kalev Frank DobbinUniversity of California, Berkeley Harvard University

Erin KellyUniversity of Minnesota

Employers have experimented with three broad approaches to promoting diversity. Someprograms are designed to establish organizational responsibility for diversity, others tomoderate managerial bias through training and feedback, and still others to reduce thesocial isolation of women and minority workers. These approaches find support inacademic theories of how organizations achieve goals, how stereotyping shapes hiringand promotion, and how networks influence careers. This is the first systematic analysisof their efficacy. The analyses rely on federal data describing the workforces of 708private sector establishments from 1971 to 2002, coupled with survey data on theiremployment practices. Efforts to moderate managerial bias through diversity trainingand diversity evaluations are least effective at increasing the share of white women,black women, and black men in management. Efforts to attack social isolation throughmentoring and networking show modest effects. Efforts to establish responsibility fordiversity lead to the broadest increases in managerial diversity. Moreover, organizationsthat establish responsibility see better effects from diversity training and evaluations,networking, and mentoring. Employers subject to federal affirmative action edicts, whotypically assign responsibility for compliance to a manager, also see stronger effectsfrom some programs. This work lays the foundation for an institutional theory of theremediation of workplace inequality.

AMERICAN SOCIOLOGICAL REVIEW, 22006, VVOL. 771 ((August:589–617)

Direct correspondence to Alexandra Kalev, RWJScholars Program, University of California, 140Warren Hall, MC7360, Berkeley, CA 94720([email protected]). The authors thank RonaldEdwards and Bliss Cartwright of the EqualEmployment Opportunity Commission for sharingtheir data and expertise; Nicole Esparza and LeslieHinkson for help with data collection; Kevin Dobbin,John Donohue, Lauren Edelman, Joshua Guetzkow,Heather Haveman, Jer ry A. Jacobs, Seema

Jayachandran, Lawrence Katz, Jordan Matsudaira,John Meyer, Trond Peterson, Daniel Schrage, PaulSegal, Robin Stryker, Donald Tomaskovic-Devey,Bruce Western, Chris Winship, and four anonymousreviewers for suggestions; and Randi Ellingboe fortechnical and editorial assistance. Supported byNational Science Foundation grant 0336642 andRussell Sage Foundation grant 87-02-03 and par-tially supported by the Robert Wood Johnson Scholarsin Health Policy Research Program.

Delivered by Ingenta to :University of California, Berkeley

Mon, 25 Sep 2006 16:51:30

loosely based on academic theories that point tocauses of workplace inequality ranging fromunwitting bias (Lemm and Banaji 1999) todependence on networks for hiring and pro-motion (Reskin and McBrier 2000). Whereasthere has been a great deal of research on thesources of inequality, there has been little on theefficacy of different programs for countering it.At best, “best practices” are best guesses. Weknow a lot about the disease of workplaceinequality, but not much about the cure.

We examine the effects of seven commondiversity programs—affirmative action plans,diversity committees and taskforces, diversitymanagers, diversity training, diversity evalua-tions for managers, networking programs, andmentoring programs—on the representation ofwhite men, white women, black women, andblack men in the management ranks of privatesector firms. Each of these programs may wellincrease diversity. To date, there has been littleevidence one way or the other. This is surpris-ing given the popularity and cost of the pro-grams. Our contribution is to bring to bear richnew data, to theoretically distinguish three typesof diversity programs, and to show that organi-zational structures allocating responsibility forchange may be more effective than programs tar-geting either managerial bias or the social iso-lation of disadvantaged groups.

Previous empirical studies of antidiscrimi-nation and diversity programs have been limit-ed by data constraints. Economists f irstcompared employers who are subject to affir-mative action requirements with those who arenot (Ashenfelter and Heckman 1976; Heckmanand Wolpin 1976; Leonard 1984). They lackeddata on employer programs. Sociologists andeconomists studying employer programs exam-ine data at one or two points in time (but seeBaron, Mittman, and Newman 1991), analyzingthe effects of some programs without account-ing for others. These studies indicate that someprograms may be effective, but their findings areinconsistent (Baron et al. 1991; Edelman andPetterson 1999; Holzer and Neumark 2000;Konrad and Linnehan 1995; Leonard 1990;Naff and Kellough 2003). Gender and racialsegregation has declined remarkably since the1970s, when employers first adopted antidis-crimination programs (Jacobs 1989a; King1992; Tomaskovic-Devey et al. 2006), but there

is no hard evidence that these programs playeda role.

We obtained the federal establishment-leveldata that economists have used (i.e., the annu-al EEO-1 reports that private sector establish-ments submit to the Equal EmploymentOpportunity Commission [EEOC]). We thensurveyed a sample of these establishments on thehistory of their personnel and diversity pro-grams so that we could analyze program effectson diversity.

A strength of the EEO-1 reports is that theydetail annual employment by race, ethnicity,and gender in all medium and large private sec-tor workplaces. A limitation is that they coveronly nine broad job categories, collapsing into“management” all jobs above that of first-linesupervisor (Baron and Bielby 1985; Smith andWelch 1984). We know from previous researchthat women and African Americans are crowd-ed in the lowest ranks of management. Even aswomen moved into management in the 1970sand 1980s, “women managers continued to trailtheir male counterparts in both earnings andauthority” (Jacobs 1992). Thus our analysesindicate which diversity programs help womenand African Americans move at least into thebottom ranks of management and, importantly,which do not. They cannot tell us whether anyof these practices help women and minorities tomove into the executive ranks.

We find a clear pattern in the data. Structuresestablishing responsibility (affirmative actionplans, diversity committees, and diversity staffpositions) are followed by significant increas-es in managerial diversity. Programs that targetmanagerial stereotyping through education andfeedback (diversity training and diversity eval-uations) are not followed by increases in diver-sity. Programs that address social isolationamong women and minorities (networking andmentoring programs) are followed by modestchanges. The effects of these initiatives varyacross groups, with white women benefitingmost, followed by black women. Black menbenefit least. We also find that responsibilitystructures make training, performance evalua-tions, networking, and mentoring programsmore effective. Federal aff irmative actionrequirements, which typically lead to assign-ment of responsibility for compliance, also cat-alyze certain programs.

590—–AMERICAN SSOCIOLOGICAL RREVIEW


Mon, 25 Sep 2006 16:51:30

These findings support an institutional the-ory of inequality remediation that builds on keyprecepts of organizational sociology. As Weber(1978 [1968]) argues, executives must appointspecialists and give them authority to achievespecialized goals. Thus, remedies targeting indi-vidual bias or network isolation may be lesseffective than remedies that establish responsi-ble parties. As neo-institutionalists (Meyer andRowan 1977) note, new programs decoupledfrom everyday practice often have no impact.Therefore, appointing a manager or committeewith responsibility for change is likely to bemore effective than annual diversity training,periodic diversity evaluations, or decentralizednetworking and mentoring programs. As struc-tural theorists of organizational inequality claim(Baron 1984), there is more to segregation thanrogue managers exercising bias. Thus, appoint-ing special staff members and committees torethink hiring and promotion structures may bemore effective than training managers not to asktheir secretaries to make coffee, and not toexclude minorities from football pools.

The argument that organizations should struc-ture responsibility for reducing inequality mayseem commonsensical, but today’s populardiversity programs often focus on changingindividuals. In the academy generally and inmanagement studies particularly, methodolog-ical individualism now holds sway. Theoristsprescribe solutions that change incentives for,and beliefs of, individuals with the idea thatmost problems of management are problemsof motivation rather than structure. Thus themost popular program that is not federally man-dated is diversity training, designed to attackbias. Managerial bias is also the target of diver-sity evaluations that offer feedback to man-agers. Networking and mentoring programsmay appear to operate at the collective level, butthey are designed to “fix” a lack of specifichuman and social capital in individual workers.

Next, we describe the three categories ofdiversity practices, link them to theories ofinequality, and summarize the (scant) evidenceabout the effects of workplace antidiscrimina-tion programs. Then we review the research onthe effects of the Civil Rights Act and presi-dential affirmative action edicts on employ-ment—hitherto the main body of research on theeffectiveness of antidiscrimination measures.After a discussion of data and methods, we

present the results from analyses of white men,white women, black women, and black men inmanagement.

THREE AAPPROACHES TTO IINCREASINGMANAGERIAL DDIVERSITY

Scholars often presume that practices designedto attack known causes of inequality actuallywill reduce it, as Reskin (2003) argues, makinga leap of faith between causes and remedies.Thus, for example, although we know fromexperimental psychology that unconscious biasis endemic, and likely contributes to workplaceinequality, we can only hope that the prevailingtreatments—diversity training and diversityevaluations—diminish inequality. Under-standing the cause of malaria and understand-ing its treatment are two different things.Whether a prescription for inequality is effec-tive is an inherently empirical question. Currentprescriptions are not based in evidence.

Our goal is to take a first step toward devel-oping an empirically based theory of remedia-tion for organizational inequality. We sketchthree mechanisms for remediating workplaceinequality rooted in different social science lit-eratures and discuss the popular humanresources (HR) measures thought to put thesetheories to work. One mechanism, based inarguments from Max Weber and organization-al institutionalists, is the creation of special-ized positions as the way to achieve new goals.Another mechanism, based in theories of stereo-typing and bias, involves training and feedbackas the way to eliminate managerial bias and itsoffspring, inequality. A third mechanism, basedin theories of social networks, involves pro-grams that target the isolation of women andminorities as a way to improve their careerprospects.

ORGANIZATIONAL CHANGE: SSTRUCTURES OF

RESPONSIBILITY

We begin with a canonical insight from orga-nizational theory. Organizational sociologistsand psychologists find that workers ignorenewly announced organizational goals and con-tinue to pursue old goals with old routines. Thedecoupling of formal goals and daily practicemay occur because individuals face informationoverload, and thus stick to the familiar, or

CORPORATE AAFFIRMATIVE AACTION AAND DDIVERSITY PPOLICIES—–591


Mon, 25 Sep 2006 16:51:30

because the old ways of doing things have beenimbued with meaning and value over time(Orton and Weick 1990; Selznick 1949).Institutionalists argue that decoupling is com-mon in programs responsive to regulatorydemands, such as civil rights programs (Dobbinet al. 1988; Edelman and Petterson 1999; Scott2001; Sutton and Dobbin 1996). Thus, forinstance, academic departments have abandonedthe old-boy system of hiring in favor of open jobadvertisement, but department chairs still asktheir pals for leads. Some argue that managersmay simply not perceive it as in their interest topromote gender and racial integration of jobs(Jacobs 1989b). Decoupling is particularly like-ly when there is no office or expert to monitorprogress, as Max Weber (1978 [1968]) hintedwhen he argued that executives should appointspecialists to pursue specialized goals.

If Weber and the institutionalists are correct,where diversity efforts are everyone’s respon-sibility but no one’s primary responsibility, theyare more likely to be decoupled. In organizationsthat do not assign responsibility for diversitygoals to a specific office, person, or group,these goals may fall by the wayside as line man-agers juggle competing demands to meet pro-duction quotas, financial targets, and the like(Edelman 1990; Meyer and Rowan 1977).Scholars (Reskin 2003; Sturm 2001) and con-sultants (Winterle 1992) alike advise ongoingcoordination and monitoring of diversityprogress by dedicated staff members or taskforces. Three common approaches can be usedto establish responsibility for diversity, as dis-cussed in the following sections.

RESPONSIBILITY AND AFFIRMATIVE ACTION

PLANS. Assign responsibility for setting goals,devising means, and evaluating progress; thiswas Weber’s advice to bureaucrats. The agencyLyndon Johnson set up in 1965 to monitor affir-mative action among federal contractors encour-aged this approach. In 1971, the Office ofFederal Contract Compliance (OFCC, whichlater gained a P for “programs” to becomeOFCCP) ordered contractors to write affirma-tive action plans in which they annually evalu-ate their own workforces, specify goals for thefair representation of women and minoritiesbased on labor market analyses, and sketchtimetables for achievement of these goals(Shaeffer 1973:66).

The order also specifies that firms shouldassign responsibility to a staff member: “He orshe must have the authority, resources, supportof and access to top management to ensure theeffective implementation of the affirmativeaction program” (U.S. Department of Labor2005). By collecting and reviewing local infor-mation annually, the affirmative action officercan track “underutilization” of women andminorities and keep managers informed abouttheir departments’ progress (Linnehan andKonrad 1999:410; Reskin 2003:13) or initiate“constructive dialogue” about making furtherprogress (Sturm 2001).

The few studies that examine effects of affir-mative action plans are inconclusive. Baron etal. (1991), studying annual data from 89California state agencies between 1975 and1981, found that, all else being equal, agencieswith affirmative action programs made signif-icantly slower progress in gender desegrega-tion of jobs. Yet those agencies were moreintegrated originally, so it may be that preex-isting affirmative action programs had left lit-tle room for improvement (see also Edelman andPetterson 1999:126; Leonard 1990:65). In astudy of 3,091 federal contractors with affir-mative action plans Jonathan Leonard (1985b)shows that the goals employers set for hiringwhite women, black women, and black men didhave positive effects, although the goals werewildly optimistic. Goals apparently do not actas quotas because virtually no employer everachieves its written goals.

Federal contractors are required to write affir-mative action plans, but contractor status doesnot correspond perfectly with the presence of aplan. Many contractors fail to write plans or toupdate them (Bureau of National Affairs 1986;Leonard 1990:55). Up to one fourth of firmswith affirmative action plans are not contractors.They create plans to bid for contracts or to setdiversity goals (Bureau of National Affairs1986; Reskin 1998). In our sample, 7 percent ofcontractors never had a plan, and 20 percent offirms that had never had a contract wrote plans.

OVERSIGHT VIA STAFF POSITIONS AND DEPART-MENTS. Following the classic bureaucratic dic-tum (Weber 1978 [1968]), some organizationsappoint full-time staff members or create depart-ments to monitor diversity instead of leaving thetask to line managers or assigning it to staffers



Mon, 25 Sep 2006 16:51:30

with other responsibilities. As a newly appoint-ed diversity manager in a high tech companyexplained to us in 2001: “As the organization hasstarted to grow, they realized they needed some-one in there to really pay attention to affirma-tive action and compliance and .|.|. efforts ondiversity.|.|.|. So the position was created at thebeginning of this year.”

Big military contractors were the first to cre-ate special positions, in the wake of Kennedy’sinitial affirmative action order in 1961. Edelmanand Petterson (1999) show that equal opportu-nity departments do not increase gender andracial diversity on their own, but that they doexpand diversity recruitment programs, whichin turn improve diversity. We include a measurefor recruitment programs to isolate the effectsof diversity staff positions.

OVERSIGHT AND ADVOCACY VIA COMMITTEES.From the late 1980s, experts have advisedemployers to appoint diversity committees andtask forces comprising people from differentdepartments, professional backgrounds, andmanagerial levels. Committees typically arecharged with overseeing diversity initiatives,brainstorming to identify remedies, and moni-toring progress. The diversity task force at theaccounting and consulting giant Deloitte &Touche, for instance, created a series of ongo-ing groups responsible for analyzing the gendergap, recommending remedial steps, and estab-lishing systems for monitoring results and ensur-ing accountability (Sturm 2001:492).

These three strategies share a focus onresponsibility. An organization with any one ofthese has assigned responsibility for progress toa person or group—an affirmative action offi-cer, a diversity manager or department, or acommittee or task force. That person or groupmonitors progress regularly. Affirmative actionofficers also write explicit annual goals forprogress, as do some staffers and committees.

BEHAVIORAL CHANGE: RREDUCING BIAS

THROUGH EDUCATION AND FEEDBACK

Social psychologists trace inequality to biasamong managers. Stereotyping is a natural cog-nitive mechanism. It is inevitable given ourinnocent tendency to make associations betweencategories and concepts (Gorman 2005;Heilman 1995; Lemm and Banaji 1999). The

implicit associations we make between race,gender, ethnicity, and social roles can have theeffect of reproducing existing patterns ofinequality (Jost, Banaji, and Nosek 2004).Managers may unwittingly select women forjobs traditionally dominated by women andmen for jobs dominated by men, with the effectof preserving between-group differences.Moreover in-group preference is widespread(Tajfel and Turner 1979) and may likewise con-taminate managerial judgment (Baron andPfeffer 1994; Reskin 2000). Rosabeth MossKanter (1977) sketches the early research on in-group preference to support her theory ofhomosocial reproduction—white men promot-ing their clones. Kanter argues that managersprefer to hire their own for reasons of commu-nication and trust.

Two corporate initiatives are thought to count-er stereotyping and in-group preference.Diversity training is thought to make managersaware of how bias affects their actions and thoseof subordinates. Diversity evaluations arethought to provide managers with feedbackshowing the effects of their decisions on diver-sity.

EDUCATION VIA DIVERSITY TRAINING. Socialpsychological research shows that giving peo-ple information about out-group members andabout stereotyping may reduce bias (Fiske 1998;Nelson, Acker, and Melvin 1996). Diversitytraining provides managers with such informa-tion. It can be traced to the equal opportunity“sensitivity” training programs that a handful ofmajor corporations put together in the mid-1970s in response to the first equal opportuni-ty consent decrees and court orders (Shaeffer1973). By the late 1980s, quite a few corporatetrainers and psychologists had developed train-ing modules designed to familiarize employeeswith antidiscrimination law, to suggest behav-ioral changes that could address bias, and toincrease cultural awareness and cross-culturalcommunication (Bendick, Egan, and Lofhjelm1998).

Employers usually offer training either to allmanagers or to all employees. We look at theeffects of training offered at least to all man-agers. Some studies of diversity training suggestthat it may activate rather than reduce bias(Kidder et al. 2004; Rynes and Rosen 1995;Sidanius, Devereux, and Pratto 2001). Research



Mon, 25 Sep 2006 16:51:30

on diversity training programs has seldomexplored their effects on workforce composition,but one study of federal agencies (Naff andKellough 2003) did show that a broad diversi-ty program had a negative effect on the pro-motion of minorities (Krawiec 2003:514).

FEEDBACK VIA PERFORMANCE EVALUATIONS.Feedback is thought to reduce bias by directingmanagerial attention and motivation (Reskin2003:325). Laboratory experiments show thatwhen subjects know that their decisions willbe reviewed by experimenters, they show lowerlevels of bias in assigning jobs (Salancik andPfeffer 1978; Tetlock 1985). Evaluating man-agers on their diversity performance createsoversight and provides feedback. As early as1973, the Harvard Business Review noted that“as one criterion of a line manager’s perform-ance appraisal, some companies have includedhis success in effectively implementing equalopportunity programs” (Fretz and Hayman1973:137). By the mid-1980s, a study of nineexemplary firms found that managers in eachfirm received regular equal opportunity per-formance evaluations (Vernon-Gerstenfeld andBurke 1985:59–60). To our knowledge, no stud-ies assess the effects of diversity evaluations.

TREATING SOCIAL ISOLATION: NNETWORKING

AND MENTORING

Mark Granovetter (1974) brought insights aboutsocial networks, pioneered by both sociologistsand psychologists, to the study of how peoplefind jobs. Students of inequality have sincespeculated that differential network contactsand differential resources accruing from thesecontacts may explain part of the continuinginequality between whites and blacks, andbetween men and women (Blair-Loy 2001; Burt1998; Ibarra 1992, 1995; McGuire 2000;Petersen, Saporta, and Seidelm 1998). Whitemen are more likely than others to find goodjobs through network ties because their net-works are composed of other white men whodominate the upper tiers of firms (Burt 1998;Reskin and McBrier 2000, but see Fernandezand Fernandez-Mateo 2006; Mouw 2003).Social networks also encourage trust, support,and informal coaching (Baron and Pfeffer 1994;Castilla 2005; Kanter 1977). Networking andmentoring programs designed specifically for

women and minorities are thought to provideuseful contacts and information (Thomas 2001).Both types of programs were pioneered in the1970s and then revived in the 1990s as part ofdiversity management efforts (Wernick 1994:25;Winterle 1992:21).

NETWORKING PROGRAMS. Diversity network-ing programs for women and minorities vary instructure. Some take the form of regular brown-bag lunch meetings, whereas others include lav-ish national conferences (Crow 2003). Theseprograms may be initiated by employees or byHR managers. They provide a place for mem-bers to meet and share information and careeradvice. Some networks also advocate policychanges, such as those involving family policiesand domestic–partner benefits (Briscoe andSafford 2005). Although networking may occurwithout any organizational impetus, we exam-ine formal networking programs that employ-ers support through release time for participants,meeting space, funding, newsletters, and emaillists.

MENTORING PROGRAMS. In 1978, the HarvardBusiness Review published an article titled“Everyone Who Makes It Has a Mentor” thatmade mentors a must-have for aspiring man-agement trainees (Lunding, Clements, andPerkins 1979; see also Roche 1979). Proponentsof formal mentoring programs argue that theycan level the playing field, giving women andminorities the kinds of relationships that whitemen get through the old-boy network.Mentoring programs match aspiring managerswith senior mentors, with the two meeting forcareer counseling and informal advice.Empirical studies, such as Burke and McKeen’s(1997) survey of university graduates, suggesta relationship between mentoring and careersuccess among women, but do not rule out thepossibility that ambitious women seek men-tors. One study of random mentor assignmentwithin a single firm found that, in general,mentees have improved social networks andtactical knowledge, which may help their careers(Moore 2001). Others have found that cross-racementoring relationships often fail (Thomas2001), and that same-sex mentoring does nothave a positive effect on job placement in aca-



Mon, 25 Sep 2006 16:51:30

demic departments of economics (Neumarkand Gardecki 1996).

ADVERSE EFFECTS OF DIVERSITY PRACTICES

Some argue that affirmative action and diver-sity programs can backfire (Bond and Pyle1988; Linnehan and Konrad 1999). First, exec-utives may believe that women and minoritiesbenefit from reverse discrimination and thusmay not deserve their positions (Heilman,Block, and Stathatos 1997; but see Taylor 1995).Second, because of the elusive nature of cogni-tive bias, “conscious attempts at thought regu-lation”—such as diversity training and diversityevaluations—“may even backfire, leading toexaggerated stereotyping under conditions ofdiminished capacity, or when self-regulationefforts are relaxed” (Nelson et al. 1996:31).Indeed, management consultants and researchersfind mixed reactions to diversity managementamong white males, who report that they are“tired of being made to feel guilty in every dis-cussion of diversity .|.|. of being cast as oppres-sors” (Hemphill and Haines 1997). Third,coworkers and executives may have negativereactions when they perceive minorities “asattempting to obtain power by individual andcollective means” (Ragins 1995:106), and exec-utives may fear that networking will lead tounion organizing (Bendick et al. 1998; Carter2003; Friedman and Craig 2004; Miller1994:443; Society for Human ResourcesManagement 2004). Finally, some studies findthat racially diverse work groups communicateless effectively and are less coherent (Baugh andGraen 1997; Townsend and Scott 2001; Vallas2003; Williams and O’Reilly 1998). Takentogether, this research suggests that diversityprograms may inhibit management diversity,particularly for blacks.

THE CCIVIL RRIGHTS AACT, AAFFIRMATIVEACTION EEDICTS, AAND DDIVERSITYPRACTICES

Although there is little research on the effectsof corporate diversity programs, the Civil RightsAct and presidential affirmative action ordershave been shown to increase diversity. The CivilRights Act covers virtually all employers, mak-ing research on its effects difficult (Donohue andHeckman 1991). The effects of presidential

affirmative action orders can be examined bycomparing federal contractors subject to theseorders with noncontractors. Six studies usingEEOC data for periods of 4 to 6 years between1966 and 1980 show that black employmentgrew more quickly among contractors(Ashenfelter and Heckman 1976; Goldstein andSmith 1976; Heckman and Payner 1989;Heckman and Wolpin 1976). Affirmative actionhad negligible effects on white women (Leonard1989:65). Contractor effects on blacks, espe-cially black women, declined from the early1980s (Leonard 1990:58), coincident with theReagan administration’s policy of deregulation.These studies do not look at whether federalcontractors increased black employment byadopting antidiscrimination practices. The twoexceptions are a study by Leonard (1985b)showing that employers who set high recruit-ment goals see more change and a study byHolzer and Neumark (2000) showing thatemployers subject to affirmative action lawexpand recruitment efforts and hire more appli-cants from disadvantaged groups. We examinethe effect of affirmative action orders andexplore the possibility that being subject to suchorders (by being a federal contractor) renders theseven diversity programs more effective.

In summary, we expect the different sorts ofdiversity programs to vary in efficacy. If assign-ing organizational responsibility is more effec-tive than targeting the behavior of individuals,then affirmative action plans, diversity com-mittees, and full-time diversity staff will be fol-lowed by broader increases in diversity thanwill either diversity training and diversity eval-uations, or networking and mentoring programs.By the same logic, the latter four programs maybe more effective when implemented in organ-izations with responsibility structures. Finally,we examine whether affirmative action oversightrenders programs more effective.

ALTERNATIVE SSOURCES OOF CCHANGEIN TTHE MMANAGERIAL WWORKFORCE

We include in the analyses other factors thoughtto affect management diversity. We cannotinclude factors that do not vary with time, suchas industry or location, because our fixed-effectsmodels account for such stable traits.



Mon, 25 Sep 2006 16:51:30

LEGAL ENVIRONMENT

Legal enforcement, through OFCCP compli-ance reviews, lawsuits, and EEOC charges,should increase employers’ hiring and promo-tion of women and minorities (Baron et al.1991:1386; Donohue and Siegelman 1991;Kalev and Dobbin forthcoming; Leonard 1984;Skaggs 2001).

ORGANIZATIONAL STRUCTURES

Organizational size and the availability of man-agerial jobs create new opportunities (Baron etal. 1991), but also more competition. Konradand Linnehan (1995) and Leonard (1990:52)find that increased demand for managers favorswhite women, but not African Americans.Unionization tends to preserve segregation byfavoring old timers through seniority provisions(Blau and Beller 1992; Milkman 1985; but seeKelly 2003; Leonard 1985a). Formalization ofpersonnel systems can reduce favoritism(Dobbin et al. 1993; Reskin and McBrier 2000),although it also can create separate career tra-jectories for different groups (Baldi and McBrier1997; Baron and Bielby 1985; Elvira andZatzick 2002). Legal counsel may sensitizeemployers to diversity in promotion decisions,and recruitment systems targeting women andminorities can increase diversity (Edelman andPetterson 1999; Holzer and Neumark 2000).Finally, work/family policies may remove obsta-cles to the promotion of women (Williams2000).

TOP MANAGEMENT COMPOSITION

The diversity of the top management team mayaffect managerial hires through homosocialreproduction or social closure (Kanter 1977;Tomaskovic-Devey 1993).

LABOR MARKET AND ECONOMIC ENVIRONMENT

Firms can more easily increase managerialdiversity when internal and external labor poolsare diverse (Cohen, Broschak, and Haveman1998; Shenhav and Haberfeld 1992). Demandfor workers from underrepresented groups maybe higher in industries with more federal con-tractors. In hard economic times, black men, andto a lesser extent women, are more vulnerablethan white men to being laid off (Elvira and

Zatzick 2002; Kletzer 1998). Finally, growingindustries can offer more attractive jobs, andboth women and minorities have historicallybeen relegated to less attractive sectors (Reskinand Roos 1990:298).

DATA AAND MMETHODS

We conducted a fixed-effects analysis of lon-gitudinal data on the workforce composition of708 establishments to assess changes in mana-gerial composition after the adoption of each ofseven diversity practices. The data cover theperiod 1971–2002. Fixed-effect models account,implicitly, for organizations’ unobserved char-acteristics that do not vary over time and thatmay affect diversity.

EEOC DDATA

The Civil Rights Act of 1964, as amended,requires private employers with more than 100employees and government contractors withmore than 50 employees and contracts worth$50,000 to file annual EEO-1 reports. Thesereports detail the race, ethnicity, and gender ofemployees in nine broad occupational cate-gories. There are no better data on workforcecomposition (for a methodological discussion onusing EEO-1 reports, see Robinson et al. 2005).We obtained the data from the EEOC throughan Intergovernmental Personnel Act (IPA) agree-ment.

Some argue that employers reclassified jobsin the 1970s, moving women and minoritiesinto management categories to improve theirfederal reports (Smith and Welch 1984).Leonard (1990:53) notes that “pure reclassifi-cation would cause black losses in the loweroccupations [in the EEO data], which is gener-ally not observed.” Jacobs (1992:298) shows adeclining gender earnings gap consistent withreal progress, noting that “the predominant trendhas been toward real, if slow progress into man-agement on the part of women.” In our sample,few firms show sudden increases for women orblacks in management, but we checked resultsfor robustness by eliminating these cases, andthe results did not change. We also eliminatedestablishment-year spells from before 1990, asdiscussed later, and the findings held up.



Mon, 25 Sep 2006 16:51:30

ORGANIZATIONAL SURVEY DATA

We drew a random sample of establishmentsfrom the EEO-1 database for our organization-al survey. For that sample, we constructed adataset comprising all EEO-1 reports for theyears 1971–2002, interpolating for the missingyears of 1974, 1976, and 1977. Establishmentsenter the dataset when they begin filing EEO-1 reports. To ensure that we would be able to fol-low establishments over time, we chose half ofthe sample from establishments that had beenin the dataset since 1980 and half from those thathad been in the dataset since 1992. We alsostratified by size, selecting 35 percent of estab-lishments with fewer than 500 employees in1999, and by industry to represent the manu-facturing, service, and trade sectors. We sampledfrom food, chemicals, computer equipment,transportation equipment, wholesale trade, retailtrade, insurance, business services, and healthservices. Corporate diversity can be influencedby acquisitions, spin-offs, and plant closings, sowe sampled establishments, selecting no morethan one per parent firm.

We conducted a longitudinal survey ofemployment practices at each establishmentcovering the years 1971–2002, in collaborationwith the Princeton Survey Research Center. Wedrew on the experiences of others who had con-ducted organizational surveys of employmentpractices (particularly Kalleberg et al. 1996;Kelly 2000; Osterman 1994, 2000). We com-pleted 833 interviews, for a response rate of 67percent, which compares favorably with therates of those other organizational surveys. Inpreparation, we conducted 41 in-person inter-views with HR managers from randomly sam-pled organizations in four different regions, and20 pilot phone interviews. Data from thoseinterviews are not included in the analysesreported in this discussion.

We began by writing to the HR director ateach establishment. We asked for permission toconduct an interview and for the name of theperson who could best answer questions aboutthe establishment’s history of HR practices. Thetypical interviewee was an HR manager with 11years of tenure. We scheduled phone interviewsat the convenience of the interviewees, andexplained in advance the nature of the infor-mation needed. We asked whether the estab-lishment had ever used each personnel program,when it was adopted, and whether and when it

had been discontinued. Program discontinuationwas rare. When a respondent could not answera question, we sent a copy of that question byemail or fax, asked that she consult records andcolleagues, and called back to fill in the blanks.During our in-person pilot interviews, respon-dents routinely pulled out manuals with copiesof policies and lists of adoption and revisiondates. Nonetheless, because responses aboutevents long past may be inaccurate, we repli-cated the analyses using only establishment-year spells for 1990 to 2002, as discussed later.

We matched survey data for each establish-ment with annual EEO-1 records, creating adataset with annual establishment-year spells.After excluding 10 cases that had EEO-1 dataavailable for fewer than 5 years, 13 cases withexcessive numbers of missing values for EEO-1 or survey data, and 102 cases that were miss-ing the adoption date for at least one keyprogram, our final dataset included 708 casesand 16,265 establishment-year cells, with amedian of 25 years of data per establishment, aminimum of 5 years, and a maximum of 32years. We collected data on national, state, andindustry employment from the Bureau of LaborStatistics.

Because of our stratified sampling designand the response pattern, we were concernedthat respondents might not represent the popu-lation of establishments that file EEO-1 reportsin the sampled industries. We constructedweights based on the inverse probability that anestablishment from each stratum (industry bysize and by time in the EEO-1 dataset) wouldcomplete the survey. We replicated all reportedanalyses using weights, and the results remainedintact. We report unweighted results in the fol-lowing discussion (Winship and Radbill 1994).We also were concerned that employers whorefused to participate might systematically dif-fer, on factors affecting diversity, from thosewho participated. We included in the modelspredicted values from a logistic regression esti-mating the probability of response (Heckman1979). This did not change our results.Covariates in that model were industry, estab-lishment status (headquarters, subunit, stand-alone status), size, contractor status, managerialdiversity, and contact person’s position. The lastvariable was obtained in the initial contact, theothers from the EEO-1 data.



Mon, 25 Sep 2006 16:51:30

DEPENDENT VARIABLES

The dependent variables are the log odds thatmanagers are white men, white women, blackwomen, and black men . For each group, oddsare calculated as the proportion of managersfrom that group divided by the proportion notfrom that group (proportion/(1 – proportion)).Figure 1 presents the trends, in percents, in oursample. Between 1971 and 2002, managementjobs held by white men decline from 81 to 61percent in the average establishment.Management jobs held by white women risefrom 16 to 26 percent, whereas those held byblack women rise from 0.4 to 2 percent, andthose held by black men rise from 1 to 3.1 per-cent. There also is a significant rise in the rep-resentation of other groups, notably Hispanics,during this period, which is why the percentagesdo not sum up to 100 percent.

Black women and men showed dramaticchanges in their proportions in managementrelative to the baseline, quadrupling and tripling,respectively, but saw small changes in percent-age points. Because the absolute changes forblacks are relatively small we log the depend-ent variables. We use log odds, rather than logproportion, because the distribution is close tonormal (Fox 1997:78).1 In a sensitivity analy-sis, log proportion performed very similarly.The dependent variable is measured annually,one year after the independent variables.Changing the lag to 2, 3, or 4 years does not alterthe findings. Our sample is designed to inves-tigate the effects of diversity programs on work-force composition in private sectorestablishments large enough to file EEO-1reports. We do not claim to describe the nation’smanagerial workforce. Nationally representativesamples such as the Current Population Surveyinclude the public and nonprofit sectors, inwhich the gains of women and minorities have

been larger. Furthermore, national figures reflectthe change in women’s representation in man-agement associated with service sector growth(e.g. Jacobs 1992), whereas our data track arelatively stable set of firms.

AFFIRMATIVE ACTION PLANS AND DIVERSITY

PRACTICES

Figure 2 shows the prevalence of all seven diver-sity programs among the 708 employers ana-lyzed later. By 2002, affirmative action planswere used in 63 percent of the workplaces westudy, followed by training in 39 percent, diver-sity committees in 19 percent, networking pro-grams (for women and minorities) in 19 percent,diversity evaluations for managers in 19 percent,diversity staff in 11 percent, and mentoring pro-grams (for women and minorities) in 11 percent.The bivariate correlations and joint frequen-cies of the seven programs are not shown here(see Online Supplement, ASR Web site: http://www2.asanet.org/journals/asr/2006/toc052.html).

In the analyses reported in the following dis-cussion, we use binary variables to represent thepresence of the seven diversity programs. For sixprograms, we asked whether the organizationhad ever had the program, when it was firstadopted, and when (if ever) it was discontinued.For the seventh practice, diversity training, weasked when it was first and last offered. If anemployer had gone for 3 years without training,we treated the program as defunct. We collect-ed additional information about diversity train-ing because our in-person interviews suggestedthat it varied across organizations more thanthe other programs, but we found significantsimilarities in training programs. In 70 percentof the establishments with training for man-agers, training was mandatory. Included in 80percent of the training programs was a discus-sion on the legal aspects of diversity, and 98 per-cent were conducted with live facilitators, asopposed to being offered exclusively via theWeb or video. Although some organizationsoffered training not only to managers, but alsoto all employees, we report effects of trainingfor managers because managers made promo-tion decisions. Training for all employees hadnearly identical effects in the models.

Because the measures are binary, coded 1for all the years the program is in place, programeffects are estimated for the entire period of


1 Because log-odds (logit) is undefined at valuesof 0 and 1, we substituted 0 with 1/2Nj, and 1 with1-1/2Nj, where Nj is the number of managers inestablishment j (Hanushek and Jackson 1977; Reskinand McBrier 2000). The results were robust to dif-ferent substitutions for 0. We chose the one that keptthe distribution unimodal and closest to normal. Toensure that the substitution does not drive the find-ings, we include a binary variable for no group mem-bers in management.


Mon, 25 Sep 2006 16:51:30

http://www2.asanet.org/journals/asr/2006/toc052.html



Figure 1. Percent of Managers: White Men and Women and Black Men and Women, 1971–2002

Note: Based on EEO-1 reports 1971–2002 sampled for Princeton University Human Resources Survey 2002.Varying N. Maximum N = 708; EEO = equal employment opportunity.

Figure 2. Percent of Private-Sector Workplaces with Affirmative Action Plans and Diversity Programs, 1971–2002

Note: Based on Princeton University Human Resources Survey, 2002. Varying N. Maximum N = 708.


Mon, 25 Sep 2006 16:51:30

the program’s existence (not merely for the yearafter initiation).

For six of the programs, between 2 and 4percent of the respondents who reported theprogram’s adoption could not tell us the exactyear. For the seventh practice, affirmative actionplan, the figure was 8 percent. We eliminatedcases with missing data on any of these vari-ables. The results were virtually identical whenwe imputed missing data for variables of inter-est and retained these cases in the analysis.Missing adoption dates for control variableswere imputed using ordinary least squares(OLS) regression, with industry, age of estab-lishment, and type of establishment as covari-ates. Omitting cases with imputed data did notsubstantially alter the findings.

CONTROL VARIABLES

All measures included in the analyses varyannually. Table 1 presents definitions and datasources for key variables as well as means andstandard deviations (based on all organization-al spells). Descriptive statistics for the entire listof control variables are not shown here (seeOnline Supplement, ASR Web site). Becausethe fixed-effects method estimates variationwithin the organization, it captures change overtime. For example, in the models, the variableorganizational size captures the effect of achange in size on change in managerial diver-sity. These models effectively ignore measuresthat do not change, such as industry, but cross-case variation in those measures is captured bythe fixed effects.

LEGAL ENVIRONMENT. We include a binaryvariable based on the EEO-1 reports indicatingwhether the establishment is a federal contrac-tor subject to affirmative action regulation.Legal enforcement is measured using three sur-vey variables that capture the establishment’sexperience with Title VII lawsuits, EEOCcharges, and affirmative action compliancereviews. Each is coded 1 from the year of thefirm’s first enforcement experience. More thanone third of establishment-year spells had pre-viously faced a lawsuit; more than one thirdhad faced an EEOC charge; and nearly 15 per-cent had faced a compliance review (only con-tractors are subject to compliance reviews).

ORGANIZATIONAL STRUCTURES. Organi-zational size and availability of managerial jobsare measured using EEO-1 data on the totalnumber of employees in the establishment andthe number of managerial employees.Unionization is coded 1 when the establish-ment has at least one contract. Substitutingwith a measure of core job unionization doesnot alter the results. Formal HR policies involvea count of hiring, promotion, and dischargeguidelines; job descriptions; promotion lad-ders; performance evaluations; pay grade sys-tem: and internal job posting. Legal counsel ismeasured with a binary variable for the pres-ence of an in-house attorney. Targeted recruit-ment policy is a binary measure of specialdiversity recruitment efforts. Work–family sup-port counts paid maternity leave, paid paterni-ty leave, flextime policies, and top managementsupport for work–family programs as assessedby our respondents.

TOP MANAGEMENT COMPOSITION. Top man-agement team diversity is measured with thepercentage of the top 10 positions held bywomen and/or African Americans, based onsurvey data. We asked about the percentage at10-year intervals and interpolated values forthe intervening years.

LABOR MARKET AND ECONOMIC ENVIRONMENT.The diversity of the establishment’s internallabor pool is measured with two variables basedon the EEO-1 reports: the percent of the focalgroup in nonmanagerial jobs and the percent inthe core job. To determine the EEO-1 categorythat held the core job, we asked respondentsabout the single biggest job in the organiza-tion. We include a variable coded 1 when thereare no members of the focal group in manage-ment. Diversity of the establishment’s externallabor pool is captured by two sets of variableson industry and state labor forces from theCurrent Population Survey. Industry employ-ment variables are logged. We use the industry’spercent of government contractors (based onEEO-1 data) to measure demand for underrep-resented workers in affirmative action sectors.Economic conditions are measured with theyearly state unemployment rate, and industrysize is measured as total annual industry



Mon, 25 Sep 2006 16:51:30

CORPORATE AAFFIRMATIVE AACTION AAND DDIVERSITY PPOLICIES—–601Ta

ble

1.Se

lect

ed V

aria

bles

Use

d in

Ana

lysi

s of

Man

ager

ial W

orkf

orce

Com

posi

tion

Mea

nSt

anda

rd D

evia

tion

Min

imum

Max

imum

Type

Dat

a

Out

com

e V

aria

bles

(per

cent

)—

Man

ager

s w

ho a

re w

hite

men

70.0

23.6

010

0C

ontin

uous

EE

O-1

—M

anag

ers

who

are

whi

te w

omen

22.2

21.2

010

0C

ontin

uous

EE

O-1

—M

anag

ers

who

are

bla

ck w

omen

1.4

4.2

066

.7C

ontin

uous

EE

O-1

—M

anag

ers

who

are

bla

ck m

en2.

45.

90

100

Con

tinuo

usE

EO

-1A

ffir

mat

ive

Act

ion

and

Div

ersi

ty M

easu

res

—A

ffir

mat

ive

actio

n pl

an.4

22.4

940

1B

inar

ySu

rvey

—Fu

ll tim

e E

EO

/div

ersi

ty s

taff

.045

.206

01

Bin

ary

Surv

ey—

Div

ersi

ty c

omm

ittee

.052

.222

01

Bin

ary

Surv

ey—

Div

ersi

ty tr

aini

ng.0

64.2

440

1B

inar

ySu

rvey

—D

iver

sity

eva

luat

ions

of m

anag

ers

.102

.303

01

Bin

ary

Surv

ey—

Net

wor

king

pro

gram

s.0

64.2

440

1B

inar

ySu

rvey

—M

ento

ring

pro

gram

s.0

33.1

790

1B

inar

ySu

rvey

Leg

al E

nvir

onm

ent

—A

ffir

mat

ive

actio

n st

atus

(gov

ernm

ent c

ontr

act)

.455

.498

01

Bin

ary

EE

O-1

—C

ompl

ianc

e re

view

.149

.356

01

Bin

ary

Surv

ey—

Dis

crim

inat

ion

law

suits

.341

.474

01

Bin

ary

Surv

ey—

EE

OC

cha

rges

.314

.464

01

Bin

ary

Surv

eyO

rgan

izat

iona

l Str

uctu

res

—Pe

rcen

t man

ager

s in

est

ablis

hmen

t.1

24.0

90.0

02.7

89C

ontin

uous

EE

O-1

—E

stab

lishm

ent s

ize

702

827

1012

,866

Con

tinuo

usE

EO

-1—

Uni

on a

gree

men

t.2

54.4

360

1B

inar

ySu

rvey

—Fo

rmal

HR

pol

icie

s4.

917

2.51

60

9C

ount

Surv

ey—

In-h

ouse

atto

rney

.277

.448

01

Cou

ntSu

rvey

—Sp

ecia

l rec

ruitm

ent f

or w

omen

and

min

oriti

es.1

56.3

630

1B

inar

ySu

rvey

—W

ork-

fam

ily a

ccom

mod

atio

ns.9

12.9

780

4C

ount

Surv

eyTo

p M

anag

emen

t Com

posi

tion

(per

cent

)—

Top

man

ager

s w

ho a

re m

inor

ities

3.47

110

.239

010

0C

ontin

uous

Surv

ey—

Top

man

ager

s w

ho a

re w

omen

16.4

4523

.575

010

0C

ontin

uous

Surv

ey

Not

e:N

= 1

6,26

5. L

abor

mar

ket a

nd e

cono

mic

env

iron

men

t var

iabl

es a

re in

clud

ed in

the

anal

yses

but

not

sho

wn

here

. See

not

e to

Tab

le 2

for a

det

aile

d lis

t of v

aria

bles

not

sho

wn

here

(see

ent

ire

list o

f con

trol

var

iabl

es o

n O

nlin

e Su

pple

men

t, A

SRW

eb s

ite: h

ttp://

ww

w2.

asan

et.o

rg/jo

urna

ls/a

sr/2

006/

toc0

52.h

tml)

. EE

O =

equ

al e

mpl

oym

ent o

ppor

tuni

ty;

HR

= hu

man

reso

urce

s.


Mon, 25 Sep 2006 16:51:30


employment, both from the Current PopulationSurvey.

METHODS

We use pooled cross-sectional time-series mod-els, with fixed effects for both establishment andyear (Hicks 1994; Hsiao 1986). We use fixedeffects for establishments to account for unmea-sured, time-invariant characteristics that mightaffect outcome variables (for recent empiricalexamples of these methods applied to individ-uals, see Budig and England 2001; Western2002). This specification, achieved by sub-tracting the values of each observation fromthe establishment mean (Hsiao 1986:31),strengthens our causal inferences about theeffects of affirmative action plans and diversi-ty practices by ruling out the possibility thatorganizations that adopted those practices hadstable unobserved preferences for diversity. Tocapture environmental changes, such as legaland cultural shifts, we use a binary variable foreach year, omitting 1971. The large number ofparameters involved in estimating fixed-effectsmodels renders them less efficient than otherestimators. However, we prefer these to alter-native models because they provide the moststringent tests of our hypotheses. The estab-lishment and year fixed effects also offer anefficient means of dealing with nonconstantvariance of the errors (heteroskedasticity) stem-ming from the cross-sectional and temporalaspects of the pooled data.

Because our dependent variables are meas-ured as parts of the same whole (the wholebeing management jobs), we expect their errorterms to be correlated. Ordinary least squareswould thus produce unbiased and consistent, butinefficient, estimators. We use seemingly unre-lated regression, which takes into accountcovariance between the errors and producesunbiased, efficient estimators (Felmlee andHargens 1988; Greene 1997; Zellner 1962).Simultaneous estimation also allows us to com-pare the effect of each diversity practice acrossgroups with formal chi-square tests (Kallebergand Mastekaasa 2001; Zellner 1962).

FINDINGS

The analysis shows substantial variation in theeffectiveness of diversity programs. Someincrease managerial diversity across the board,

whereas others have meager effects, or posi-tive effects for some groups and negative effectsfor others. The most effective practices are thosethat establish organizational responsibility: affir-mative action plans, diversity staff, and diver-sity task forces. Attempts to reduce socialisolation among women and African Americansthrough networking and mentoring programsare less promising. Least effective are programsfor taming managerial bias through educationand feedback.

DIVERSITY PROGRAMS AT WORK

In Table 2, we report models of managerialdiversity. (Selected control variables are pre-sented; the remaining coefficients can be seenon the Online Supplement, ASR Web site). Eachdependent variable is the (natural) log odds ofmanagers being from a certain group. To trans-form the coefficient ! from representing changein log odds to representing percentage changein odds, it should be exponentiated: [exp(!) –1]*100. Once exponentiated in this way thecoefficient represents the average percentagechange in the odds that managers are from a cer-tain group, associated with a change in the inde-pendent variable. In the discussion below we use‘odds for [group]’ as a shorthand. We also pro-vide an illustrative summary of the results inproportion terms.

The R2 figures for these fixed-effects mod-els represent the percentage of the varianceexplained by the predictors when the uniqueeffects of each establishment are excluded. A loglikelihood ratio test shows that the variablesreported in Table 2 significantly improve themodel fit (chi(28) = 405.66; p < .001), as com-pared with the baseline models that have novariables representing diversity programs (avail-able on request).

ORGANIZATIONAL RESPONSIBILITY. Coeffi-cients for the diversity programs represent thechange in the log odds that managers are froma certain group that is attributable to the pres-ence of a practice, averaged across all years ofthe program’s existence. After employers set upaffirmative action plans, the odds for white menin management decline by 8 percent; the oddsfor white women rise by 9 percent; and the oddsfor black men rise by 4 percent. These numbersrepresent the estimated average difference



Mon, 25 Sep 2006 16:51:30


Table 2. Fixed Effects Estimates of the Log Odds of White Men and Women and Black Women and Men inManagement, 1971–2002

White Men White Women Black Women Black Men

Organizational Responsibility—Affirmative action plan –.078** .086** .005 .039*— (.017) (.017) (.014) (.015)—Diversity committee –.081** .175** .242** .114**— (.028) (.029) (.024) (.026)—Diversity staff –.055 .104** .123** .128**— (.033) (.034) (.028) (.030)Managerial Bias—Diversity training –.038 –.001 –.066** .031— (.021) (.022) (.018) (.019)—Diversity evaluations .028 .061* –.027 –.081**— (.027) (.028) (.023) (.025)Social Isolation—Networking programs –.083** .080** .012 –.096**— (.027) (.028) (.023) (.024)—Mentoring programs –.011 –.004 .213** .037— (.033) (.035) (.029) (.031)Legal Environment—Government contract .032 .006 –.039* –.027— (.019) (.019) (.016) (.017)—Compliance review –.083** .077** .020 .081**— (.020) (.020) (.017) (.018)—Title VII lawsuit –.107** .141** .044** .029*— (.015) (.016) (.013) (.014)—EEOC charge –.007 .014 .019 .034*— (.016) (.017) (.014) (.015)Organizational Structures—Proportion managers in establishment –.896** .309** –4.499** –3.989**— (.108) (.112) (.092) (.099)—Establishment size (log) –.021 –.023* –.661** –.515**— (.012) (.012) (.010) (.011)—Union agreement –.053 –.068* –.007 –.029— (.033) (.034) (.028) (.030)—Formal personnel policies –.002 –.003 –.016** –.015**— (.004) (.004) (.003) (.003)—In-house attorney –.100** .126** –.040* .021— (.023) (.024) (.020) (.021)—Targeted recruitment policy –.071** .108** .131** .099**— (.021) (.021) (.018) (.019)—Work-family accommodations –.078** .065** .026** .004— (.008) (.009) (.007) (.008)Top Management Composition—Proportion minorities in top management –.002 –.002 .007** .012**— (.001) (.001) (.001) (.001)—Proportion women in top management –.002** .004** .002** –.002*— (.001) (.001) (.001) (.001)——R2 (64 parameters) .3335 .3146 .3636 .2799

Note: Log likelihood ratio test; "2 (28) = 405.66; p < .001. Data shown are coefficients from seemingly unrelatedregression with standard errors in parentheses. Variables included in the analyses but not shown here are 8 vari-ables for proportion of each group in non-managerial jobs and in core job in each establishment; 4 binary vari-ables for no workers from a group in management; 8 variables for proportion of each group in state and industrylabor forces; proportion of contractor firms in industry; industry employment; and state unemployment rate (fullresults on Online Supplement, ASR Web site: http://www2.asanet.org/journals/asr/2006/toc052.html). Analysesalso include establishment and year fixed effects. All independent variables are lagged by 1 year, excludingproportion of managerial jobs. N (organization-year) = 16,265; N (organizations) = 708. EEOC = EqualEmployment Opportunity Commission. * p < .05; ** p < .01 (two tailed test).


Mon, 25 Sep 2006 16:51:30


between having a plan and the counterfactualcondition of not having a plan for the entireperiod of the plan’s existence. These results areconsistent with Leonard’s (1990) finding thataffirmative action plan goals are effective. Notethat the coefficient for black women is not sig-nificant here. When we introduced industryinteractions, we discovered that in manufactur-ing (computers, electronics, transportation),affirmative action plans had negative effectson black women, whereas in service (retail,insurance, business services), affirmative actionplans had positive effects (results available uponrequest). Creating a diversity committee increas-es the odds for white women, across the periodof the committee’s existence, by 19 percent.The odds for black women rise 27 percent, andthe odds for black men rise 12 percent.Employers who appoint full-time diversity staffalso see significant increases in the odds forwhite women (11 percent), black women (13percent), and black men (14 percent) in man-agement.

As noted, the coefficients in Table 2 representthe average changes in log odds that managersare from a certain group. The effect of eachprogram on the percent of women and minori-ties in management will vary depending onwhere organizations begin (Fox 1997:78). Forexample, an 8 percent decrease in the odds ofmanagers being white men resulting from adop-tion of affirmative action plan would translateto a decline of 2.6 percent in the percent ofwhite men in management if they constituted 70percent before adoption, but it would mean alarger decline of 4.3 percent if they made uponly 50 percent at the baseline (Petersen1985:311).

PROGRAMS FOR REDUCING MANAGERIAL BIAS.Programs designed to reduce managerial biasthrough education (diversity training) and feed-back (diversity evaluations) show one modestpositive effect and two negative effects acrossthe three disadvantaged groups. Diversity train-ing is followed by a 7 percent decline in the oddsfor black women. Diversity evaluations are fol-lowed by a 6 percent rise in the odds for whitewomen, but an 8 percent decline in the odds forblack men. These mixed effects are anticipatedin the literature. As noted, laboratory studies andsurveys often show adverse reactions to train-ing (Bendick et al. 1998; Nelson et al. 1996).

Moreover, critics argue that trainers definediversity broadly to include groups not coveredby federal civil rights law (parents, smokers),and thereby draw attention away from protect-ed groups (Edelman, Fuller, and Mara-Drita2001; Kochan et al. 2003; Konrad and Linnehan1995).

PROGRAMS FOR REDUCING SOCIAL ISOLATION.Networking and mentoring programs, designedto counter social isolation, show modest effectson managerial diversity. Networking is followedby a rise in the odds for white women and adecline in the odds for white men and blackmen. The negative coefficient for black men isanticipated by qualitative research (Carter 2003;Friedman and Craig 2004) showing that whitescan develop negative attitudes toward African-American organizing. In contrast, mentoringprograms show a strong positive effect on theodds for black women. These findings suggestthat having personal guidance and support atwork can facilitate career development (Castilla2005) for black women, whereas networking ismore effective for white women.

GENDER AND RACIAL PATTERNS. Overall, itappears that diversity programs do most forwhite women and more for black women thanfor black men. Black men gain significantlyless from affirmative action than do whitewomen (chi-sq(1) = 4.15, p < .05), and signif-icantly less from diversity committees than doblack women (chi-sq(1) = 22.47, p.< .01). Threeprograms show negative effects on AfricanAmericans, whereas no program shows a neg-ative effect on white women. We hesitate tooverinterpret this pattern, but note that there issomething of a trade-off among groups.

Table 3 evaluates the magnitude of the effectsof programs on the proportion of each group inmanagement based on the coefficients in Table2. “Proportion in year of adoption” is the meanproportion of each group in management,among adopters, in their actual years of programadoption (i.e., just before treatment). “Estimatedproportion with practice” shows the predictedmean proportion after the practice is in place.Thus, for example, the proportion of whitewomen among managers in the average estab-lishment adopting an affirmative action pro-gram was 0.132, and the net effect of the



Mon, 25 Sep 2006 16:51:30

program, with control for other factors, is toraise white women proportion to 0.142.Similarly, the proportion of black women amongmanagers was 0.014 in the average firm adopt-ing a diversity committee, and adoption bringsblack women to 0.018, an increase of almost30%. The third row, based on the first two rows,reports the percentage change over the baselineresulting from program adoption.

Tables 2 and 3 support our contention thatprograms establishing organizational responsi-bility are more broadly effective than those thataddress managerial bias or social isolationamong women and African Americans.Organizations that structure responsibility seeconsistent positive effects for white women,black women, and black men.

Coefficients for control variables are con-sistent with expectations, with one possibleexception. The negative effect of formal per-

sonnel policies is not consistent with the ideathat bureaucracy impedes cronyism or bias inpromotion decisions (Reskin and McBrier2000), but is consistent with the argument thatformalization leads to the needless inflation ofeducational prerequisites (Collins 1979), andwith findings that the determinants of promo-tion differ systematically for whites and blackseven when formal personnel systems exist(Baldi and McBrier 1997). Other coefficients ofcontrol variables show that although growthand unionization have not improved diversity,and although legal staff had only limited effects,targeted recruitment programs, work/familyaccommodations, and top management teamdiversity show positive effects on managerialdiversity. Coefficients for the labor market andeconomic environment measures, not shownhere, are in the expected direction as well (seeOnline Supplement, ASR Web site).


Table 3. Estimated Average Differences in Managerial Composition Due to Adoption of Affirmative Action andDiversity Practices


Affirmative Action Plan—Proportion in year of adoption .783 .132 .017 .024—Estimated proportion with practice .769 .142 .017 .025—Percent difference due to adoption –1.8%** 7.6%** .0% 4.2%**Diversity Committee—Proportion in year of adoption .630 .230 .014 .020—Estimated proportion with practice .611 .262 .018 .022—Percent difference due to adoption –3.0%** 13.9%** 29.8%** 10.0%**Diversity Staff—Proportion in year of adoption .724 .157 .014 .021—Estimated proportion with practice .713 .171 .016 .024—Percent difference due to adoption –1.5% 8.9%** 14.3%** 14.3%**Diversity Training—Proportion in year of adoption .687 .194 .017 .022—Estimated proportion with practice .679 .194 .016 .023—Percent difference due to adoption –1.2% .0% –5.9%** 4.5%Diversity Evaluations—Proportion in year of adoption .720 .160 .017 .024—Estimated proportion with practice .726 .168 .017 .022—Percent difference due to adoption .8% 5.0% .0% –8.3%**Networking Programs—Proportion in year of adoption .702 .193 .014 .020—Estimated proportion with practice .684 .206 .014 .018—Percent difference due to adoption –2.6%** 6.7%** .0% –10.0%**Mentoring Programs—Proportion in year of adoption .690 .216 .017 .021—Estimated proportion with practice .688 .215 .021 .022—Percent difference due to adoption –.3% –.5% 23.5%** 4.8%

Note: Estimates based on coefficients presented in Table 2. * p < .05; ** p < .01 (two tailed test).


Mon, 25 Sep 2006 16:51:30

DOES ORGANIZATIONAL RESPONSIBILITY

IMPROVE PROGRAM EFFECTIVENESS?

It is possible that some programs work best incombination with others (MacDuffie 1995;Perry-Smith and Blum 2000). Our finding thatorganizational responsibility structures havebroader effects than other programs suggeststhat perhaps training, evaluation, mentoring,and networking would be more successful incombination with responsibility structures. Weundertake several analyses of program combi-nations.

First, we explore the possibility that the sim-ple number of programs matters. Perhaps ourmeasures capture not the effects of discrete pro-grams so much as an orientation toward chang-ing workplace demography. We introduce threebinary variables representing the presence of anyone, two, and three or more programs. Acrossthe 16,265 organization-year spells of data, 49percent had no programs, 34 percent had oneprogram, 10 percent had two programs, and 7percent had three or more programs. In the toppanel of Table 4, we report the effects of the


Table 4. Fixed-Effects Estimates of the Log Odds of White Men and Women and Black Women and Men inManagement with Bundles of Programs, 1971–2002

White White Black BlackMen Women Women Men

Adoption of One or More AA Plans & Diversity Programs—Only one program –.043** .056** –.009 .026— (.016) (.016) (.013) (.014)—Two programs –.091** .121** .020 .024— (.023) (.023) (.019) (.021)—Three or more programs –.158** .232** .127** .046— (.029) (.030) (.025) (.027)——R2 (60 parameters) .3323 .3124 .3569 .2767—Interaction with Responsibility Structures—Responsibility structures –.063** .081** .007 .042**— (.017) (.017) (.014) (.015)—Diversity training –.026 –.064 –.046 .026— (.036) (.038) (.031) (.033)——! Responsibility structure –.026 .132** .044 .040— (.042) (.043) (.036) (.038)—Diversity evaluations .294** –.042 –.065 –.077— (.057) (.059) (.049) (.052)——! Responsibility structure –.326** .136* .057 .009— (.061) (.063) (.053) (.057)—Networking programs –.090 .163** –.026 –.172*— (.050) (.052) (.043) (.046)——! Responsibility structure –.003 –.088 .073 .118*— (.056) (.058) (.048) (.051)—Mentoring programs .140** –.101 –.042 .127*— (.066) (.068) (.057) (.061)——! Responsibility structure –.183* .133 .344** –.108— (.074) (.076) (.063) (.068)

—R2 (66 parameters) .3347 .3136 .3602 .2785

Note: Data shown are coefficients from 2 seemingly unrelated regression analyses with standard errors in paren-theses. Responsibility Structures include affirmative action plans, diversity committees and diversity staff. Theanalyses include establishment and year fixed effects and all the control variables included in the models present-ed in Table 2 (for coefficients of control variables, see Online Supplement, ASR Web site: http://www2.asanet.org/journals/asr/2006/toc052.html). N (organization-year) = 16,265; N (organizations) = 708.* p < .05; ** p < .01 (two tailed test).


Mon, 25 Sep 2006 16:51:30



number of programs in models parallel to thosepresented in Table 2 (results for the control vari-ables are available on the Online Supplement,ASR Web site). We compared coefficients for thebinary count variables using t tests. For whitewomen, the sheer number of programs matters;one is better than zero, two better than one, andthree or more are better than two. For whitemen, we find the opposite pattern, suggestingthat each additional program reduces the oddsfor white men. For black women, having one ortwo programs is not significantly different fromhaving none. Having three is significantly dif-ferent. For black men, none of the count vari-ables show an effect significantly different fromhaving no programs. Hence, for white women,the more programs the better. For blacks, thenumber of programs matters less than the con-tent of the programs. This is not surprising giventhat some practices in Table 2 show no effects,or even negative effects, on blacks.

Although each additional program, regardlessof content, does not always translate into greaterdiversity, particular bundles of programs mightoperate well together. To test this idea, we ran(in models otherwise identical to those in Table2) all two-way interactions between affirma-tive action plan, diversity committee, diversitystaff, training, evaluation, networking, and men-toring. (The bivariate correlations and joint fre-quencies of the seven programs are presented onthe Online Supplement, ASR Web site.) Thetwo-way interactions among training, evalua-tion, networking, and mentoring did not indicatethat any pairs operated better than individualprograms. But two-way interactions withresponsibility structures did render training,evaluation, networking, and mentoring moreeffective. For ease of presentation, we collapsethe three responsibility structures into a singlevariable, interacting it with the four other pro-gram variables. The second panel in Table 4includes estimates from models with these inter-actions (results for the control variables are pre-sented on the Online Supplement, ASR Website).

Diversity training, evaluation, networking,and mentoring programs are more effective infirms with responsibility structures. With diver-sity training and evaluations, the responsibilitystructure interaction positively affects whitewomen. With networking, the responsibilitystructure interaction positively affects black

men, and with mentoring, it positively affectsblack women. Note that the noninteracted vari-able, responsibility structure, continues to showthe expected effects for white men, whitewomen, and black men. The overall pattern isstriking and suggests that these authority struc-tures render the other programs more effective.Yet even with responsibility structures in place,none of these programs show the sort of con-sistent pattern across outcomes that we findfor, say, diversity committee.

DO AFFIRMATIVE ACTION ORDERS MEDIATE

PROGRAM EFFICACY?

In Table 2, we also examine whether affirma-tive action enforcement shows direct effects.Employers who sign a government contract,and thereby become subject to affirmative actionregulation, do not see increases in managerialdiversity as a direct result. When we interactedcontractor status with the period 1971–1980, theresults did not support early researchers’ find-ings that contractors experienced faster growthin black employment in the 1970s. Of course,effects found in earlier studies were quite small,and it may be that they were concentrated inindustries we do not sample. For the entire peri-od, we find a decline in the odds for blackwomen after the approval of a government con-tract. This may be because employers who striveto improve their numbers before seeking gov-ernment work, improve more slowly afterreceiving contracts (Baron et al. 1991:1389;Leonard 1990:65). Government contractor sta-tus does not show positive effects even when weexclude programs that may be associated withcontractor status: the seven diversity measures,formal HR policies, work–family policies, andcompliance reviews (results available onrequest).

Unlike contractor status, antidiscriminationenforcement shows effects. Federal compliancereviews, which 32 percent of the contractors inour data faced, increased representation of whitewomen and black men. Leonard (1985b) alsofound effects of compliance reviews in his studyof the 1970s. When we interacted compliancereview with the period 1971–1980, our results(available upon request) replicated his findingfrom the 1970s as well (see also Kalev andDobbin forthcoming). Discrimination lawsuitsincrease the odds for all three groups in man-



Mon, 25 Sep 2006 16:51:30

agement (Skaggs 2001), and EEOC chargesincrease the odds for black men.

The natural follow-up question is whetheraffirmative action oversight mediates the effi-cacy of the seven affirmative action and diver-sity measures. Theory suggests that programimplementation may be taken more seriously infirms subject to regulatory scrutiny. Those firmstypically assign responsibility for compliance toan office or person. In Table 5, we add interac-tion terms between programs and contractorstatus to the model presented in Table 2.Coefficients for control variables are availableon the Online Supplement, ASR Web site. A

log-likelihood test shows a significant improve-ment in fit over that of the model presented inTable 2. The interaction coeff icients showwhether effects are significantly different amongcontractors and noncontractors. We also exam-ine the linear combination of the interactioncomponents (using Lincom in Stata) to assesswhether programs have signif icant effectsamong contractors.

Diversity training shows the greatest differ-ence in effects on all four groups. Whereasamong noncontractors training decreases therepresentation of white and black women inmanagement, among contractors it is followed


Table 5. Fixed-Effects Estimates of the Log Odds of White Men and Women and Black Women and Men inManagement with Government Contractor Interactions, 1971–2002


Affirmative Action Plan –.050* .086** .000 .007(.023) (.023) (.019) (.021)

—! Government contract –.050 .003 .000 .053*(.028) (.029) (.024) (.026)

Diversity Committee –.096* .173** .270** .076*(.038) (.040) (.033) (.035)

—! Government contract .029 –.006 –.050 .074(.053) (.055) (.046) (.049)

Diversity Staff –.076 .018 .205** .240**(.058) (.060) (.050) (.053)

—! Government contract .024 .120 –.127* –.145*(.066) (.068) (.056) (.060)

Diversity Training .005 –.094** –.116** –.016(.027) (.028) (.023) (.025)

—! Government contract –.092* .197** .107** .100**(.038) (.040) (.033) (.035)

Diversity Evaluations .049 .090* –.097** –.063(.039) (.041) (.034) (.036)

—! Government contract –.041 –.035 .118** –.027(.050) (.051) (.042) (.045)

Networking Programs –.133** .171** –.034 –.035(.038) (.039) (.033) (.035)

—! Government contract .111* –.195** .069 –.113*(.051) (.052) (.043) (.046)

Mentoring Programs .028 –.053 .179** .070(.046) (.047) (.039) (.042)

—! Government contract –.081 .086 .057 –.056(.063) (.065) (.054) (.058)

R2 (71 parameters) .3341 .3165 .3650 .2811

Note: Log likelihood ratio test; "2 (28) = 135.86; p < .001; Data shown are coefficients from seemingly unrelatedregression with standard errors in parentheses. The analyses include establishment and year fixed effects and allthe control variables included in the models presented in Table 2 (for coefficients of control variables, see OnlineSupplement, ASR Web site: http://www2.asanet.org/journals/asr/2006/toc052.html). N (organization-year) =16,265; N (organizations) = 708. * p < .05; ** p < .01 (two tailed test).


Mon, 25 Sep 2006 16:51:30


by a significant decline in the odds for whitemen (! = .086; SE = .004) and significantincreases among white women (! = .103; SE =.030) and black men (! = .083; SE = .027).Diversity evaluations also are less likely to back-fire among contractors, where the effect onblack women is now zero.

Affirmative action plans show significantlylarger effects for black men among contractors,further supporting Leonard’s (1990) findings.The coefficients for diversity staff in the mod-els for black women and men, although signif-icantly smaller among contractors, are stillpositive and significant (B = .078; SE = .032 andB = .095; SE = .034, respectively). Networkingprograms help white women in noncontractorestablishments, at the expense of white men, butthis effect disappears among contractors, andblack men see negative effects for reasons thatare not clear.

FURTHER ANALYSES

A key challenge in analysis of nonexperimen-tal data is to account for heterogeneity thatstems from nonrandom selection into the “treat-ment” (in our case, adopting a program).Heterogeneity may bias casual inference. Ourmodel specification, with fixed effects for eachyear and each establishment and with controlvariables measuring organizational structures,labor pool composition, and economic and legalenvironment, is designed to minimize this pos-sibility.

We conducted three additional robustnesstests (results available on request). First, weadded binary variables as proxies for unspeci-fied, unobserved events (impending lawsuit,local news coverage) that may have causedemployers both to implement new antidiscrim-ination programs and to hire more women andAfrican Americans. We created proxies for eachof the seven programs. We re-ran the analysis14 times, with proxies measured 2 and 3 yearsbefore program adoption in models parallel tothose presented in Table 2. These proxy variablesdid not substantially alter the coefficients orstandard errors for affirmative action and diver-sity programs, and most did not show signifi-cant effects. This adds to our confidence that theobserved relationships between diversity pro-grams and managerial diversity are not spurious

(Rossi, Lipsey, and Freeman 2004; Snyder2003).

Second, program adopters may be differentfrom nonadopters in ways that are not absorbedby the establishment fixed effects. Perhapsadopters change faster than nonadopters interms of management fads and demographics.We therefore re-ran the analyses in Table 2seven times, each time only with establishmentsthat ever adopted a particular program (once foraffirmative action plan adopters, then for diver-sity committee, etc.). If the effects in Table 2 areattributable to differences between adopters andnonadopters, then program effects should dis-appear when we exclude nonadopters. Theresults of our “adopters only” analyses are sub-stantively similar to those in Table 2.

Third, we were concerned that because thedataset is not rectangular (some establishmentsenter the data after 1971), unobserved hetero-geneity might distort the results if establish-ments are missing in early years for reasons(e.g. organizational size or age) associated withthe outcome variables. We thus replicated theanalysis using a rectangular subsample of estab-lishments. The results were substantially simi-lar to those reported in this discussion.

To examine the robustness of the results towithin-unit serial correlation, we corrected forthe possibility that each error is partially depend-ent on the error of the previous year (AR[1])with the Cochrane–Orcutt method (available inState using xtregar, not the seemingly unrelat-ed regression). This transforms the data by sub-tracting from the equation for time t the equationfor time t-1 multiplied by the autocorrelationcoefficient. The AR(1) results are substantial-ly similar to those reported in Table 2 (availableon the Online Supplement, ASR Web site). Theone exception is that affirmative action plan issignificant for whites only at the p < 0.1 level.We report seemingly unrelated regression mod-els in Table 2 because they account for related-ness of outcome variables and are thus moreefficient, and because they allow us to comparecoefficients for different groups.

Because our analyses cover more than threedecades, we also explored two theories of tim-ing and program efficacy (results available onrequest) to rule out the possibility that some pro-grams showing no effects in the aggregate actu-ally were effective at certain points in time.One theory is that employer practices are more



Mon, 25 Sep 2006 16:51:30

effective under active regulatory regimes. Wethus added to the model reported in Table 2interaction terms between each of the practicesand the Reagan and first Bush era (1981–1992)as well as the Bill Clinton and George W. Bushera (1993–2002). The comparison period, 1971to 1980, encompassed the activist Nixon admin-istration, the brief Ford administration, and theactivist Carter administration (Skrentny 1996).A finding that programs were more effectiveduring the 1970s might help to explain whyresearch on the period (e.g., Leonard 1990)found the greatest increases in black employ-ment among contractors. We find no evidencethat programs operated differently across peri-ods.

The second timing argument is that earlyprogram adopters are those most committed tochange (Tolbert and Zucker 1983). We lookedat whether the effects of each practice werestronger among the first 15, 25, and 40 percentof eventual adopters. Our analyses showed thatpractices are no more effective among earlyadopters.

We also explored whether some programsshowed weak effects in the models because theyhad differential effects by establishment sizeor industry. With regard to size interactions,some negative program effects were neutral-ized in very large establishments, but the pro-grams that proved ineffective in general were noteffective among large or small organizations. Inindustry interactions, most program effects werestable in direction if not in magnitude acrossindustries. One notable pattern was that theeffect of aff irmative action plans on blackwomen was negative in manufacturing and pos-itive in service, as discussed earlier.

Finally, we were concerned that surveyrespondent reports of early program dates mightbe inaccurate, which could cause us to under-estimate program effects by including post-treatment values (i.e., that reflect changesattributable to a program) as pretreatment data.We were particularly concerned about resultsshowing weak effects for training, evaluations,networking, and mentoring. Correlationsbetween respondent tenure and adoption yearswere small and not significant, the one excep-tion being for networking (correlation of –0.20;p < 0.05). To evaluate the effects of measure-ment error, we re-ran Table 2 models, elimi-nating establishment-year spells before 1990,

thus excluding from the analysis possibly erro-neous information on early years of adoption.Using fixed-effects models to analyze only datafor 1990–2002 would prevent us from evaluat-ing the effects of programs adopted any timebefore 1990, so we first replicated the full analy-sis (for the entire period) without fixed estab-lishment effects, replacing differenced variableswith undifferenced variables. The results weresimilar to those presented in Table 2. Then usingthe undifferenced variables, we re-ran the mod-els eliminating all establishment-year spellsbefore 1990. We lost many spells, but the sub-stantive results held up (for results, see OnlineSupplement on ASR Web site). This increasesour confidence in the models, and particularlyin the weak effects of training, evaluations, net-working, and mentoring.

CONCLUSION

The antidiscrimination measures we study havebecome popular among employers, HR man-agers, lawyers, and advocacy groups, despite theabsence of hard evidence that they work (Bisom-Rapp 1999; Krawiec 2003). Employers use thesepractices to defend themselves in court, andthe courts, in many cases, accept them as goodfaith efforts to stamp out discrimination(Edelman et al. 2005). There are reasons tobelieve that employers adopt antidiscrimina-tion measures as window dressing, to inoculatethemselves against liability, or to improvemorale rather than to increase managerial diver-sity. In the final analysis, however, the measureof these programs—for scholars, practitioners,and the courts—should be whether they do any-thing to increase diversity. Using EEO-1 reports,we cannot examine whether these programshelp women and African Americans to move upfrom the bottom rungs of management. But wecan show that some popular diversity programsat least help women and African Americans toclimb into the ranks of management. Other pop-ular programs do not do even that.

There is a rich tradition of theory and researchon the causes of workplace inequality. We con-tend that this work may not always hold clearimplications for remedies. The question of howto reduce inequality is just as deserving of atten-tion. Our conceptualization of different types ofdiversity programs and our analyses of theireffects lay the groundwork for research and the-



Mon, 25 Sep 2006 16:51:30

ory on the remediation of inequality in work-places.

Broadly speaking, our findings suggest thatalthough inequality in attainment at work maybe rooted in managerial bias and the social iso-lation of women and minorities, the best hopefor remedying it may lie in practices that assignorganizational responsibility for change. Ourown theory of the remediation of inequalitybuilds on classical organizational sociologyrather than on theories of cognitive bias or socialnetworks (see also Blum, Fields, and Goodman1994).

Structures that embed accountability, author-ity, and expertise (affirmative action plans,diversity committees and taskforces, diversitymanagers and departments) are the most effec-tive means of increasing the proportions ofwhite women, black women, and black men inprivate sector management. Moreover, theyshow effects even in the presence of controls forthe specific initiatives that specialists oftenimplement, from formal hiring and promotionrules to work–family programs. Responsibilitystructures also catalyze the other diversity pro-grams, rendering each a bit more effective forone group. Some programs also prove moreeffective among federal contractors, likelybecause legal requirements encourage employ-ers to assign responsibility for compliance.

Practices that target managerial bias throughfeedback (diversity evaluations) and education(diversity training) show virtually no effect inthe aggregate. They show modest positiveeffects when responsibility structures are alsoin place and among federal contractors. Butthey sometimes show negative effects other-wise. Research to date from HR experts and psy-chologists suggests that interactive trainingworkshops, of the kind we examine, often gen-erate backlash. Finally, programs designed tocounter the social isolation of women andminorities through mentoring and networkingare disappointing, although mentoring doesappear to help black women.

The poor performance of practices thataddress social–psychological and social–rela-tional sources of inequality should not be takenas evidence that these forces do not producesocial inequality. A preponderance of empiricalresearch shows that bias and poor network con-nections contribute to inequality. Further

research is needed to determine why these pro-grams do not live up to their promise.

Much management theorizing from law andeconomics scholars (Becker 1968; Gray andShadbegian 2005; Posner 1992; see alsoSimpson 2002) and psychologists (e.g. Tetlock1985) suggests that corporate behavior is bestcontrolled by doling out incentives to individ-ual managers and shaping their attitudes. Thisapproach is rooted in a sort of methodologicalindividualism that is prominent in managementresearch and practice. However, when it comesto addressing corporate inequality we find thatthe strategies designed to change individualsare less effective than the conventional man-agement solution of setting goals and assigningresponsibility for moving toward these goals.

That said, the three programs we found to bemost effective likely operate in somewhat dif-ferent ways. Whereas affirmative action plansand diversity staff both centralize authority overand accountability for workforce composition,diversity committees locate authority andaccountability in an interdepartmental task forceand may work by causing people from differentparts of the organization to take responsibilityfor pursuing the goal of integration.

In this study, we examine managers alone. Itis important for both theory and practice toextend this research to other occupationalgroups. Yet for employers seeking solutions tothe problem of gender and racial segregation,our analyses offer hope. Most employers dosomething to promote diversity—76 percenthad adopted one of these seven programs by2002—but do they do what is most effective?Diversity committees have been quite effective,requiring neither additional staff nor expensiveconsultants. Less than 20 percent of the estab-lishments we studied had them by 2002.Diversity staff are also quite effective, but only11 percent of establishments had them. On theother hand, diversity training, which 39 percentof establishments had adopted, and which canbe quite costly, was not very effective andshowed adverse effects among noncontractors.

Even the programs that work best have mod-est effects, particularly for African Americans,who are poorly represented to begin with.Diversity committees raise the proportion ofblack women in management by a remarkable30 percent on average, but from a baseline ofonly 1.4 percent. Appointing full-time diversi-



Mon, 25 Sep 2006 16:51:30

ty staffer raises the proportion of black men bya healthy 14 percent, but from a baseline ofonly 2.1 percent. These programs alone willnot soon change the look of management. Note,however, that our sample of large, private firmshas changed less quickly than the economy asa whole. In young start-up firms and in the pub-lic sector, these practices may be even moreeffective than they are in our sample.

The effects of these programs should not beconflated with the effects of antidiscriminationlegislation. First, as we demonstrate, federalaffirmative action regulations clearly mediatethe efficacy of diversity evaluations and train-ing. Our findings thus go against the popularclaim that antidiscrimination regulation is nolonger needed because diversity programs havegained a life of their own (Fisher 1985;Liberman 2003). Moreover, it was federal reg-ulations that led employers to first establishaffirmative action plans, the most commonintervention and one of the most effective.

Second, enforcement has been effectiveregardless of corporate policies. As researchhas shown, and as our findings support, Title VIIlawsuits and affirmative action compliancereviews led to increases in women’s and minori-ties’ share of management jobs, especially inperiods and judicial circuits wherein civil rightsenforcement was strong (Kalev and Dobbinforthcoming; Leonard 1989; 1990; Skaggs2001).

Finally, to assess the impact of antidiscrimi-nation legislation on employment inequality,one needs to consider broader political, social,and cultural changes associated with the CivilRights Act, affirmative action, and related laws(Burstein 2000). Yet if the effects of governmentantidiscrimination measures have slowed, assome observers suggest, then we should wasteno time sorting out which corporate programsare effective.

Alexandra Kalev received her Ph.D. from Princetonin 2005. Her dissertation examines how workplacerestructuring (“high performance” systems anddownsizing) affects the careers of women and minori-ties. Kalev is a postdoctoral fellow in the RobertWood Johnson Scholars in Health Policy ResearchProgram at UC Berkeley studying gender and racialdisparities in work related injuries and illnesses.Kalev has published with Frank Dobbin on civilrights law enforcement in the face of deregulation(Law and Social Inquiry), and with Erin Kelly on how

companies manage flexible schedules (Socio-Economic Review).

Frank Dobbin is Professor of Sociology at Harvard.He edited The New Economic Sociology: A Reader(Princeton University Press) and The Sociology ofthe Economy (Russell Sage Foundation), both pub-lished in 2004. He is continuing work with Kalev andKelly on the effects of employer policies on workforcediversity, and is spending the 2006–2007 academicyear at the Radcliffe Institute for Advanced Study, withfellowships from Radcliffe and from the John SimonGuggenheim Foundation.

Erin L. Kelly is Assistant Professor of Sociology atthe University of Minnesota. Her research on thedevelopment, diffusion, and implementation of fam-ily-supportive policies has appeared in the AmericanJournal of Sociology and the SocioEconomic Review.She and Phyllis Moen are conducting a multimethodstudy of whether and how flexible work initiativesaffect organizational cultures, the experiences ofworkers on the job, and the health and well-being ofworkers and their families. That project is part of theNational Institutes of Health’s research network onwork, family, health, and well-being.

REFERENCES

Ashenfelter, Orley and James J. Heckman. 1976.“Measuring the Effect of an AntidiscriminationProgram.” Pp. 46–89 in Evaluating the Labor-Market Effects of Social Programs, edited by O.Ashenfelter and J. Blum. Princeton, NJ: PrincetonUniversity Press.

Baldi, Stephane and Debra Branch McBrier. 1997.“Do the Determinants of Promotion Differ forBlacks and Whites? Evidence from the U.S. LaborMarket.” Work and Occupations 24:478–97.

Baron, James N. 1984. “Organizational Perspectiveson Stratification.” Annual Review of Sociology10:37–69.

Baron, James N. and William T. Bielby. 1985.“Organizational Barriers to Gender Equality: SexSegregation of Jobs and Opportunities.” Pp.233–251 in Gender and the Life Course, edited byA. S. Rossi. New York: Aldine de Gruyter.

Baron, James N., Brian S. Mittman, and Andrew E.Newman. 1991 . “Targets of Opportunity:Organizational and Environmental Determinantsof Gender Integration within the California CivilServices, 1976–1985.” American Journal ofSociology 96:1362–401.

Baron, James N. and Jeffrey Pfeffer. 1994. “TheSocial Psychology of Organizations andInequality.” Social Psychology Quarterly57:190–209.

Baugh, Gayle S. and George B. Graen. 1997. “Effectsof Team Gender and Racial Composition onPerceptions of Team Performance in Cross-



Mon, 25 Sep 2006 16:51:30

http://www.ingentaconnect.com/content/external-references?article=0190-2725()57L.190%5Baid=7468255%5D






Functional Teams.” Group and OrganizationManagement 22:366.

Becker, Gary. 1968. “Crime and Punishment: AnEconomic Approach.” Journal of PoliticalEconomy 76: 169–217.

Bendick, Mark, Jr., Mary Lou Egan, and Suzanne M.Lofhjelm. 1998. The Documentation andEvaluation of Antidiscrimination Training in theUnited States. Washington, DC: Bendick and EganEconomic Consultants.

Bisom-Rapp, Susan. 1999. “Bulletproofing theWorkplace: Symbol and Substance in EmploymentDiscrimination Law Practice.” Florida StateUniversity Law Review 26:959–1049.

Blair-Loy, Mary. 2001. “It’s Not Just What You Know,It’s Who You Know: Technical Knowledge,Rainmaking, and Gender among FinanceExecutives.” Research in the Sociology of Work10:51–83.

Blau, Francine D. and Andrea Beller. 1992. “BlackWhite Earnings over the 1970s and 1980s: GenderDifferences in Trends.” The Review of Economicsand Statistics 74:276–86.

Blum, Terry C., Dail L. Fields, and Jodi S. Goodman.1994. “Organization-Level Determinants ofWomen in Management.” Academy ofManagement Journal 37:241–68.

Bond, Megan A. and Jean L. Pyle. 1988. “DiversityDilemmas at Work.” Journal of ManagementInquiry 7:252–69.

Briscoe, Forrest and Sean Safford. 2005. “Agency inDiffusion: Activism, Imitation, and the Adoptionof Domestic Partner Benefits among the Fortune500.” Unpublished paper. Pennsylvania StateUniversity, University Park, PA.

Budig, Michelle J. and Paula England. 2001. “TheWage Penalty for Motherhood.” AmericanSociological Review 66:204–25.

Bureau of National Affairs. 1986. Work and Family:A Changing Agenda. Washington, DC: Bureau ofNational Affairs.

Burke, Ronald J. and Carol A. McKeen. 1997. “NotEvery Managerial Woman Who Makes It Has aMentor.” Women in Management Review 12:136–9.

Burstein, Paul. 2000. “The Impact of EEO Law: ASocial Movement Perspective.” Pp. 129–155 inLegacies of the 1964 Civil Rights Act, edited by B.Grofman. Charlottesville, VA: University ofVirginia Press.

Burt, Ronald S. 1998. “The Gender of Social Capital.”Rationality and Society 10:5–46.

Carter, John. 2003. Ethnicity, Exclusion, and theWorkplace. London: Palgrave Macmillan Press.

Castilla, Emilio J. 2005. “Social Networks andEmployee Performance in a Call Center.” AmericanJournal of Sociology 110:1243–84.

Catalyst. 1998. Advancing Women in Business - theCatalyst Guide: Best Practices from the CorporateLeaders. San Francisco: Jossey-Bass.

Cohen, Lisa E. , Joseph P. Broschak, and Heather A.Haveman. 1998. “And Then There Were More? TheEffect of Organizational Sex Composition on theHiring and Promoting of Managers.” AmericanSociological Review 63:711–27.

Collins, Randall. 1979. The Credential Society: AHistorical Sociology of Education andStratification. New York: Academic Press.

Crow, Kelly. 2003. “Reaching Out: Staying Focusedon Diversity Goals.” New York Times, October 28,p. G2.

Dobbin, Frank, Lauren B. Edelman, John W. Meyer,W. Richard Scott, and Ann Swidler. 1988. “TheExpansion of Due Process in Organizations.” Pp.71–100 in Institutional Patterns andOrganizations: Culture and Environment, editedby L. G. Zucker. Cambridge, MA: Ballinger.

Dobbin, Frank, John R. Sutton, John W. Meyer, andW. Richard Scott. 1993. “Equal Opportunity Lawand the Construction of Internal Labor Markets.”American Journal of Sociology 99:396–427.

Donohue, John J. and James Heckman. 1991.“Continuous Versus Episodic Change: The Impactof Federal Civil Rights Policy on the EconomicStatus of Blacks.” Journal of Economic Literature29:1603–43.

Donohue, John J. III and Peter Siegelman. 1991.“The Changing Nature of EmploymentDiscrimination Litigation.” Stanford Law Review43:983–1033.

Edelman, Lauren B. 1990. “Legal Environments andOrganizational Governance: The Expansion ofDue Process in the American Workplace.”American Journal of Sociology 95:1401–40.

Edelman, Lauren B., Sally Riggs Fuller, and IonaMara-Drita. 2001. “Diversity Rhetoric and theManagerialization of the Law.” American Journalof Sociology 106:1589–641.

Edelman, Lauren B., Linda Hamilton Krieger, ScottEliason, and K. T. Albiston. 2005. “JudicialDeference to Institutionalized ComplianceStrategies.” Working Paper Thesis, Boalt HallSchool of Law, University of California, Berkeley,CA.

Edelman, Lauren B. and Stephen M. Petterson. 1999.“Symbols and Substance in Organizations’Response to Civil Rights Law.” Research in SocialStratification and Mobility 17:107–35.

Elvira, Marta M. and Christopher D. Zatzick. 2002.“Who’s Displaced First? The Role of Race inLayoff Decisions.” Industrial Relations 41:329–61.

Equal Employment Opportunity Commission(EEOC). 1998. “Best Practices of Private SectorEmployers.” Washington, DC: Equal EmploymentOpportunity Commission. Retrieved April 19,2004 (http://www.eeoc.gov/abouteeoc/task_reports/practice.html).

Felmlee, D. H. and L. L. Hargens. 1988. “Estimationand Hypothesis Testing for Seemingly Unrelated



Mon, 25 Sep 2006 16:51:30

http://www.eeoc.gov/abouteeoc/task_reports/practice.html
















http://www.eeoc.gov/abouteeoc/task_reports/practice.html

Regression: A Sociological Application.” SocialScience Research 17:384–99.

Fernandez Roberto M. and Isabel Fernandez-Mateo.2006. “Network, Race and Hiring.” AmericanSociological Review 71:42–71.

Fisher, Anne B. 1985. “Businessmen Like to Hire bythe Numbers.” Fortune, September 16, pp. 26–30.

Fiske, Susan T. 1998. “Stereotyping, Prejudice, andDiscrimination.” Pp. 357–411 in Stereotyping,Prejudice and Discrimination, edited by D. T.Gilbert, S. T. Fiske, and G. Lindzey. New York:McGraw-Hill.

Fox, John. 1997. Applied Regression Analysis, LinearModels, and Related Methods. Thousand Oaks,CA: Sage.

Fretz, C. F. and Joanne Hayman 1973. “Progress forWomen? Men Are Still More Equal.” HarvardBusiness Review, September–October, pp.133–142.

Friedman, Raymond A. and Kellina M. Craig. 2004.“Predicting Joining, and Participating in MinorityEmployee Network Groups.” Industrial Relations43:793–816.

Goldstein, Morris and Robert S. Smith. 1976.“Changes in Labor Market for Black Americans,1948–1970.” Industrial and Labor RelationsReview 29:523–43.

Gorman, Elizabeth. 2005. “Gender Stereotypes,Same-Gender Preferences, and OrganizationalVariation in the Hiring of Women: Evidence fromLaw Firms.” American Sociological Review70:702–28.

Granovetter, Mark. 1974. Getting a Job: A Study ofContracts and Careers. Chicago: University ofChicago Press.

Gray, Wayne B. and Ronald J. Shadbegian. 2005.“When and Why Do Plants Comply? Paper Millsin the 1980s.” Law and Policy 27: 238–60.

Greene, W. H. 1997. Econometric Analysis. UpperSaddle River, NJ: Prentice-Hall.

Hanushek, Erin A. and John E. Jackson. 1977.Statistical Methods for Social Scientists. NewYork, NY: Academic Press.

Heckman, James and Kenneth Wolpin. 1976. “Doesthe Contract Compliance Program Work? AnAnalysis of Chicago Data.” Industrial and LaborRelations Review 29:544–64.

Heckman, James J. 1979. “Sample Selection Bias asa Specification Error.” Econometrica 45:153–61.

Heckman, James J. and Brook S. Payner. 1989.“Determining the Impact of FederalAntidiscrimination Policy on the Economic Statusof Blacks: A Study of South Carolina.” AmericanEconomic Review 79:138–77.

Heilman, Madeline E. 1995. “Sex Stereotypes andTheir Effects in the Workplace: What We Knowand What We Don’t Know.” Journal of SocialBehavior and Personality 10:3–26.

Heilman, Madeline E., Caryn J. Block, and Peter

Stathatos. 1997. “The Affirmative Action Stigmaof Incompetence: Effects of PerformanceInformation Ambiguity.” Academy of ManagementJournal 40:603–25.

Hemphill, Hellen and Ray Haines. 1997.Discrimination, Harassment, and the Failure ofDiversity Training: What to Do Now. Westport,CT: Quorum Books.

Hicks, Alexander M. 1994. “Introduction to Pooling.”Pp. 169–188 in The Comparative PoliticalEconomy of the Welfare State, edited by T. Janoskiand A. M. Hicks. New York: Cambridge UniversityPress.

Holzer, Harry J. and David Neumark. 2000. “WhatDoes Affirmative Action Do?” Industrial andLabor Relations Review 53:240–71.

Hsiao, Cheng. 1986. Analysis of Panel Data, vol.11. Cambridge, England: Cambridge UniversityPress.

Ibarra, Herminia. 1992. “Homophily and DifferentialReturns: Sex Differences in Network Structureand Access in an Advertising Firm.” AdministrativeScience Quarterly 34:422–47.

———. 1995. “Race, Opportunity and Diversityof Social Circles in Managerial Networks.”Academy of Management Journal 38:673–703.

Jacobs, Jerry A. 1989a. “Long-Term Trends inOccupational Segregation by Sex.” AmericanJournal of Sociology 95:160–73.

———. 1989b. Revolving Doors: Sex Segregationand Women’s Careers. Stanford, CA: StanfordUniversity Press.

———. 1992. “Women’s Entry into Management:Trends in Earnings, Authority, and Values amongSalaried Managers.” Administrative ScienceQuarterly 37:282–301.

Jost, John T. , Mahzarin Banaji, and Brian A. Nosek.2004. “A Decade of System-Justification Theory:Accumulated Evidence of Conscious andUnconscious Bolstering of the Status Quo.”Political Psychology 25:881–919.

Kalev, Alexandra and Frank Dobbin. Forthcoming.“Enforcement of Civil Rights Law in PrivateWorkplaces: The Effects of Compliance Reviewsand Lawsuits over Time.” Law and Social Inquiry.

Kalleberg, Arne L., David Knoke, Peter V. Marsden,and Joe L. Spaeth. 1996. Organizations inAmerica: Analyzing Their Structures and HumanResource Practices. Thousand Oaks, CA: Sage.

Kalleberg, Arne L and Arne Mastekaasa. 2001.“Satisfied Movers, Committed Stayers: The Impactof Job Mobility on Work Attitudes in Norway.”Work and Occupations 28:183–209.

Kanter, Rosabeth Moss. 1977. Men and Women of theCorporation. New York, NY: Basic Books.

Kelly, Erin. 2000. “Corporate Family Policies in U.S.Organizations, 1965–1997.” Ph.D. dissertation,Department of Sociology, Princeton University,Princeton, NJ.



Mon, 25 Sep 2006 16:51:30




















http://www.ingentaconnect.com/content/external-references?article=0162-895X()25L.881%5Baid=7468263%5D

———. 2003. “The Strange History of Employer-Sponsored Childcare: Interested Actors,Uncertainty, and the Transformation of Law inOrganizational Fields.” American Journal ofSociology 109:606–49.

Kidder, Deborah L., Melenie J. Lankau, DonnaChrobot-Mason, Kelly A. Mollica, and RaymondA. Friedman. 2004. “Backlash toward DiversityInitiatives: Examining the Impact of DiversityProgram Justif ication, Personal and GroupOutcomes.” International Journal of ConflictManagement 15:77–104.

King, Mary C. 1992. “Occupational Segregation byRace and Sex, 1940–1988.” Monthly Labor Review115:30.

Kletzer, Lori. 1998. “Job Displacement.” Journal ofEconomic Perspective 12:115–37.

Kochan, Thomas A., Katerina Bezrukova, Robin J.Ely, Susan Jackson, Aparna Joshi, Karen Jehn,Jonathan S. Leonard, David I. Levine, and DavidA. Thomas. 2003. “The Effect of Diversity onBusiness Performance: Report of the DiversityResearch Network.” Human ResourcesManagement 42:3–21.

Konrad, Alison. M. and Frank. Linnehan. 1995.“Formalized HRM Structures: Coordinating Equal-Employment Opportunity or ConcealingOrganizational Practices.” Academy ofManagement Journal 38:787–820.

Krawiec, Kimberly D. 2003. “Cosmetic Complianceand the Failure of Negotiated Governance.”Washington University Law Quarterly 81:487–544.

Lemm, Kristi and Mahzarin R. Banaji. 1999.“Unconscious Attitudes and Beliefs about Menand Women.” Pp. 215–35 in Perceiving andPerforming Gender, edited by U. Paseor and F.Braun. Opladen, Germany: Westdutscher Verlag.

Leonard, Jonathan S. 1984. “The Impact ofAffirmative Action on Employment.” Journal ofLabor Economics 2:439–63.

———. 1985a. “Unions and the Employment ofBlacks, Hispanics, and Women.” Industrial andLabor Relations Review 39:115–32.

———. 1985b. “What Promises are Worth: TheImpact of Affirmative Action Goals.” The Journalof Human Resources 20:3–20.

———. 1989. “Women and Affirmative Action.”The Journal of Economic Perspectives 3:61–75.

———. 1990. “The Impact of Affirmative ActionRegulation and Equal Employment OpportunityLaw on Black Employment.” The Journal ofEconomic Perspectives 4:47–63.

Liberman, Vadim. 2003. “Workplace Diversity: It’sAll in the Mix.” Across the Board XL:51–2.

Linnehan, Frank and Alison Konrad. 1999. “DilutingDiversity: Implications for Intergroup inOrganizations.” Journal of Management Inquiry8:399–413.

Lunding, F. S., C. E. Clements, and D. S. Perkins.

1979. “Everyone Who Makes It Has a Mentor.”Harvard Business Review 56:89–101.

MacDuffie, John Paul. 1995. “Human ResourceBundles and Manufacturing Performance:Organizational Logic and Flexible ProductionSystems in the World Auto Industry.” Industrialand Labor Relations Review 48:2.

McGuire, G. M. 2000. “Gender, Race, Ethnicity, andNetworks: The Factors Affecting the Status ofEmployees’ Network Members.” Work andOccupations 27:500–23.

Meyer, John W. and Brian Rowan. 1977.“Institutionalized Organizations: Formal Structureas Myth and Ceremony.” American Journal ofSociology 83:340–63.

Milkman, Ruth. 1985. “Women Workers, Feminism,and the Labor Movement since the 1960s.” inWomen, Work, and Protest: A Century of Women’sLabor History, edited by R. Milkman. Boston,MA: Routledge and Kegan Paul.

Miller, Joanne. 1994. Corporate Responses toDiversity. New York: Center for the New AmericanWorkplace at Queens College.

Moore, Paul Cameron. 2001. “The Transfer of Humanand Social Capital: Employee DevelopmentThrough Assigned Peer Mentoring.” Ph.D. disser-tation, Stanford Graduate School of Business,Stanford, CA.

Mouw, Ted. 2003. “Social Capital and Finding a Job:Do Contacts Matter?” American SociologicalReview 68:868–98.

Naff, Katherine C. and J. Edward Kellough. 2003.“Ensuring Employment Equity: Are FederalDiversity Programs Making a Difference?”International Journal of Public Administration26:1307–36.

Nelson, Thomas E., Michele Acker, and ManisMelvin. 1996. “Irrepressible Stereotypes.” Journalof Experimental Social Psychology 32:13–38.

Neumark, David and Rosella Gardecki. 1996.“Women Helping Women? Role Model andMentoring Effects on Female Ph.D. Student inEconomics.” Working Paper. National Bureau ofEconomic Research, Cambridge, MA.

Orton, Douglas J. and Karl E. Weick. 1990. “LooselyCoupled Systems: A Reconceptualization.”Academy of Management Review 15:203–23.

Osterman, Paul. 1994. “How Common is WorkplaceTransformation and Who Adopts It?” Industrialand Labor Relations Review 47:173–88.

———. 2000. “Work Reorganization in an Era ofRestructuring: Trends in Diffusion and Effects onEmployee Welfare.” Industrial and Labor RelationsReview 53:179–96.

Perry-Smith, Jill E. and Terry C. Blum. 2000.“Work–Family Human Resource Bundles andPerceived Organizational Performance.” Academyof Management Journal 43:1107–17.

Petersen, Trond. 1985. “A Comment on Presenting



Mon, 25 Sep 2006 16:51:30






















Results from Logit and Probit Models.” AmericanSociological Review 50:130–131.

Petersen, Trond, Ishak Saporta, and Marc DavidSeidelm. 1998. “Offering a Job: Meritocracy andSocial Networks.” American Journal of Sociology106:763–816.

Posner, Richard A. 1992. Economic Analysis of Law,4th edition. Boston: Little, Brown and Company.

Presidential Glass Ceiling Commission. 1995. ASolid Investment: Making a Full Use of ourNations’ Human Capital. Washington, DC: U.S.Department of Labor Glass Ceiling Commission.

Ragins Belle Rose. 1995. “Diversity, Power andMentorship in Organizations: A Cultural,Structural and Behavioral Perspective.” Pp 91–132in Diversity in Organizations: New Perspectives fora Changing Workplace, edited by M. M. Chemers,S. Oskamp, and M. A. Costanzo. Thousand Oaks,CA: Sage.

Reskin, Barbara F. 1998. The Realities of AffirmativeAction in Employment. Washington, DC: AmericanSociological Association.

———. 2000. “The Proximate Causes ofEmployment Discrimination.” ContemporarySociology 29:319–28.

———. 2003. “Including Mechanisms in OurModels of Ascriptive Inequality.” AmericanSociological Review 68:1–21.

Reskin, Barbara F. and Debra B. McBrier. 2000.“Why Not Ascription? Organizations’Employmentof Male and Female Managers.” AmericanSociological Review 65:210–33.

Reskin, Barbara F. and Patricia Roos. 1990. JobQueues, Gender Queues: Explaining Women’sInroads into Male Occupations. Philadelphia, PA:Temple University Press.

Robinson, Corre, Tiffany Taylor, Donald Tomaskovic-Devey, Catherine Zimmer, and Matthew W. Irvine,Jr. 2005. “Studying Race/Ethnic and SexSegregation at the Establishment-Level:Methodological Issues and SubstantiveOpportunities Using EEO-1 Reports.” Work andOccupations 32:5–38.

Roche, Gerard R. 1979. “Much Ado about Mentors.”Harvard Business Review 57:14.

Rossi Peter H., Mark W. Lipsey, and Howard E.Freeman. 2004. Evaluation: A SystematicApproach. Thousand Oaks, CA: Sage.

Rynes, Sara and Benson Rosen. 1995. “A FieldSurvey of Factors Affecting the Adoption andPerceived Success of Diversity Training.”Personnel Psychology 48:247–70.

Salancik, Gerald R. and Jeffrey Pfeffer. 1978.“Uncertainty, Secrecy, and the Choice of SimilarOthers.” Social Psychology 41:246–55.

Scott, W. Richard. 2001. Institutions andOrganizations. Thousand Oaks, CA: Sage.

Selznick, Philip. 1949. TVA and the Grass Roots.Berkeley, CA: University of California Press.

Shaeffer, Ruth G. 1973. “Nondiscrimination inEmployment: Changing Perspectives, 1963–1972.”New York: The Conference Board.

Shenhav, Yehouda A. and Yitchak Haberfeld. 1992.“Organizational Demography and Inequality.”Social Forces 71:123–43.

Sidanius, Jim, Erik Devereux, and Felicia Pratto.2001. “A Comparison of Symbolic Racism Theoryand Social Dominance Theory as Explanationsfor Racial Policy Attitudes.” Journal of SocialPsychology 132:377–95.

Simpson, Sally S. 2002. Corporate Crime, Law, andSocial Control. New York: Cambridge UniversityPress.

Skaggs, Sheryl. 2001. “Discrimination Litigation:Implications for Women and Minorities in RetailSupermarket Management.” Ph.D. dissertation,Department of Sociology, North Carolina StateUniversity, Raleigh, NC.

Skrentny, John D. 1996. The Ironies of AffirmativeAction: Politics Culture and Justice in America.Chicago: University of Chicago Press.

Smith, James P. and Finis Welch. 1984. “AffirmativeAction and the Labor Markets.” Journal of LaborEconomics 2:269–301.

Snyder, Lori D. 2003. “Are Management-BasedRegulations Effective?: Evidence from StatePollution Prevention Programs.” Regulatory PolicyProgram Working Paper RPP-2003-21. Center forBusiness and Government, John F. KennedySchool of Government, Harvard University,Cambridge, MA.

Society for Human Resources Management . 2004.“What Are The Components of a SuccessfulDiversity Initiative?” Alexandria, VA: Society forHuman Resources Management. Retrieved April18, 2004 (http://www.shrm.org/diversity/components.asp).

Sturm, Susan. 2001. “Second-GenerationEmployment Discrimination: A StructuralApproach.” Columbia Law Review 101:459–568.

Sutton, John R. and Frank Dobbin. 1996. “The TwoFaces of Governance: Responses to LegalUncertainty in American Firms, 1955–1985.”American Sociological Review 61:794–811.

Tajfel, Henri and John C. Turner. 1979. “AnIntegrative Theory of Intergroup Conflict.” Pp.33–47 in The Social Psychology of IntergroupRelations, edited by W. G. Austin and S. Worchel.Monterey, CA: Brooks/Cole.

Taylor, Marylee. 1995. “White Backlash toWorkplace Affirmative Action: Peril or Myth.”Social Forces 73:1385–414.

Tetlock, E. Philip. 1985. “Accountability: A SocialCheck on the Fundamental Attribution Error.”Social Psychology Quarterly 48:227–36.

Thomas, David A. 2001. “The Truth about MentoringMinorities: Race Matters.” Harvard BusinessReview, April, pp. 99–107.



Mon, 25 Sep 2006 16:51:30

http://www.shrm.org/diversity/components.asp










http://www.shrm.org/diversity/components.asp

Tolbert, Pamela S. and Lynne G. Zucker. 1983.“Institutional Sources of Change in the FormalStructure of Organizations: The Diffusion of CivilService Reform 1880–1935.” AdministrativeScience Quarterly 28:22–39.

Tomaskovic-Devey, Donald. Catheine R. Zimmer,Kevin Stainback, Corre Robinson, Tiffany Taylor,and Tricia McTague. 2006 “DocumentingDesegregation: Segregation in AmericanWorkplaces by Race, Ethnicity, and Sex,1966–2003.” American Sociological Review71:565–588

Tomaskovic-Devey, Donald D. 1993. Gender andRacial Inequality at Work: The Sources andConsequences of Job Segregation. Ithaca, NY:Industrial and Labor Relations Press.

Townsend, Anthony M. and Dow K. Scott. 2001.“Team Racial Composition, Member Attitudes,and Performance: A Field Study.” IndustrialRelations 40:317–37.

U.S. Department of Labor. 2005. “Placement Goals.Code of Federal Regulations Pertaining toEmployment Standards Administration, Title 41,Part 60-2 subpart B.” Washington, DC: U.S.Department of Labor.

Vallas, Steven P. 2003. “Rediscovering the ColorLine within Work Organizations: The `Knittingof Racial Groups’ Revisited.’ Work andOccupations 30:379–400.

Vernon-Gerstenfeld, Susan and Edmund Burke. 1985.“Affirmative Action in Nine Large Companies: AField Study.” Personnel 62:54–60.

Weber, Max. [1968] 1978. “Sociological Categoriesof Economic Action.” Pp. 63–211 in Economyand Society, vol. 1, edited by R. Guenther and C.Wittich. Berkeley, CA: University of CaliforniaPress.

Wernick, Ellen D. 1994. Preparedness, CareerAdvancement, and the Glass Ceiling. Washington,DC: U.S. Department of Labor Glass CeilingCommission.

Western, Bruce. 2002. “The Impact of Incarcerationon Wage Mobility and Inequality.” AmericanSociological Review 67:477–98.

Williams, Joan. 2000. Unbending Gender: WhyFamily Work Conflict and What to Do about It?New York: Oxford University Press.

Williams, Katherine Y. and Charles A. O’Reilly.1998. “Demography and Diversity inOrganizations.” Pp. 77–140 in Research inOrganizational Behavior, vol. 20, edited by B. M.Staw and L. L. Cummings. Greenwich, CT: JAI.

Winship, Christopher and Larry Radbill. 1994.“Sampling Weights and Regression Analysis.”Sociological Methods and Research 23:230–57.

Winterle, Mary J. 1992. Work Force Diversity:Corporate Challenges, Corporate Responses. NewYork: Conference Board.

Zellner, Alfred. 1962. “An Efficient Method ofEstimating Seemingly Unrelated Regressions andTests for Aggregation Bias.” Journal of theAmerican Statistical Association 57:348–68.



Mon, 25 Sep 2006 16:51:30










best ppractices oor bbest gguesses? assessing tthe ... · diversity lead to the broadest increases...

Documents