AN EXAMINATION OF WHETHER AND HOW RACIAL …leeds-faculty.colorado.edu/dahe7472/Hekman AMJ Best Paper Finalist...GENDER BIASES INFLUENCE CUSTOMER SATISFACTION ... We examined whether

AN EXAMINATION OF WHETHER AND HOW RACIAL ANDGENDER BIASES INFLUENCE CUSTOMER SATISFACTION

DAVID R. HEKMANUniversity of Wisconsin–Milwaukee

KARL AQUINOUniversity of British Columbia

BRADLEY P. OWENSUniversity of Michigan

TERENCE R. MITCHELLUniversity of Washington

PAULINE SCHILPZANDKEITH LEAVITT

U.S. Military Academy at West Point

We examined whether and how various biases may influence customers’ satisfactionevaluations and produce discriminatory judgments for minority and female serviceemployees. We argue that customer satisfaction evaluations are biased because theyare anonymous judgments by untrained raters that usually lack an evaluation stan-dard. Laboratory and field samples provide disturbing evidence generally confirmingour arguments and suggesting that the presence of nonwhite and women serviceemployees may produce lower aggregated customer satisfaction evaluations that mayultimately hurt individuals and organizations financially.

Customer satisfaction surveys have become acommon source of performance feedback for em-ployees and organizations (Hagan, Konopaske, Ber-nardin, & Tyler, 2006). The Mercer ConsultingGroup (2007) reported that in 2006 customer satis-faction surveys were of primary importance forstrategic decision making, and over two-thirds oforganizations used such surveys to determine someaspect of employee compensation. Moreover, cus-tomer satisfaction is an important predictor of awide range of financial measures (see Gupta andZeithaml [2006] for a review), so it is not surprisingthat some companies tie some portion of employeecompensation directly to customer satisfaction. Forexample, a 1 percent change in customer satisfac-tion for an average Fortune 500 firm has beenshown to lead to a 1.02 percent change in Tobin’s q,which equates to a change of $275 million in firmvalue (Anderson, Fornell, & Mazvancheryl, 2004), a

$55 million gain or loss in cash flow in the nextyear (Gruca & Rego, 2005), and a 5.03 percentchange in return on investment (Anderson & Mittal,2000). Satisfying customers is also increasinglyimportant to organizations as the global economybecomes more service-oriented. Macroeconomictrends indicate 76 percent of U.S. employees workin service industries, and by 2016 the number ofemployees working in these industries is expectedto increase by over 17 million (Figueroa & Woods,2007). The expanding service sector is perhaps onereason why 65 percent of 681 senior executivessurveyed by the Economist Intelligence Unit fromOctober through December 2002 reported custom-ers as their main focus over the next three years;only 18 percent of these executives reported share-holders as their main focus.

Manybusiness leaders (Bracken,Church,&Timm-reck, 2001) and researchers (Salam, Cox, & Sims,1997) have applauded the use of customer satisfac-tion surveys because they believe that aggregatedevaluations are highly reliable measures of em-ployee performance quality. However, a potentialdisadvantage of using customer surveys, particu-larly in making compensation or promotion deci-sions, is that they are ultimately subjective judg-

Parts of this research were supported by the Businessand Economic Development Center, Foster School ofBusiness, University of Washington. We would like tothank our AMJ action editor, Peter Bamberger, as well asthree anonymous reviewers for their help in the reviewprocess.

� Academy of Management Journal2010, Vol. 53, No. 2, 238–264.

238

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

ments. As a result, they are vulnerable to biases,including those based on the “bandwagon effect,”confirmation of preexisting beliefs, education orcognitive ability, and stereotypes based on the raceor gender of the person being rated (Gilovich, Grif-fin, & Kahneman, 2002). Some researchers suspectthat biases are unavoidable when gathering subjec-tive evaluations of performance, especially whensuch judgments come from naı̈ve and inexperi-enced raters who are not held accountable for theaccuracy of their ratings (Pulakos, White, Oppler,& Borman, 1989; Wilkinson & Fontaine, 2002;Woehr & Roch, 1996). To date, though, surpris-ingly little research has examined how and ifdifferent biases influence the judgments of cus-tomers about whether an organization’s employ-ees, organizational attributes, or services meet orexceed their expectations.

The purpose of our research is to examinewhether and how customer satisfaction ratingsare potentially influenced by gender and racialbias. We extend the existing literature on biasesin supervisory ratings of employee performanceby focusing on customer satisfaction ratings,which have been mostly excluded from organiza-tional behavior research (see Moshavi [2004] for arare exception). We conceptualize these satisfac-tion ratings as judgments and examine judgmentsmade not only about individual employees, butalso about their organizational context (e.g., itsperceived cleanness or appearance) and their or-ganizational unit as a whole. To our knowledge,no prior research has examined bias in customerjudgments of organizational context or overallwork unit. Drawing on the literature on modernforms of racial and gender bias (Crandall & Eshel-man, 2003; Greenwald & Banaji, 1995), we con-tend that customer satisfaction with organization-al contexts or units may be vulnerable to suchbiases. Finding empirical evidence of racial orgender biases in customer satisfaction would sug-gest that organizations have a financial incentiveto favor white or male employees. Such a findingmay help explain the persistent inequality be-tween demographic groups in the workplace.

Two methodological attributes of our researchdistinguished it from previous studies of bias incustomer judgments of employee performance andallowed us to conduct a stronger test of the validityof our predictions. First, we took into account em-ployees’ objectively measured performance behav-iors when examining customer judgments. Theproblem with relying solely on customer satisfac-tion scores to assess customer bias is that suchscores can be interpreted as capturing both true

performance and biases (Landy, Shankster, &Kohler, 1994; Latham & Wexley, 1977). Rotundoand Sackett summarized this state of affairs: “Thereis no definitive way of determining whether therated criterion used in a validity study is biased.Thus, there is no current method of establishingwhether there is bias in performance ratings”(1999: 816). Our study design allowed us to teaseapart differences in satisfaction judgments thatwere attributable to customer bias arising from em-ployee demographic characteristics from those thatwere due to objective employee performance(Greenhaus, Parasuraman, & Wormley, 1990; Pula-kos et al., 1989; Wilkinson & Fontaine, 2002; Woehr& Roch, 1996).

Second, we heeded recent calls for researchers toengage in “full-cycle” research in which initialfield-based findings are tested in the laboratory andthen revalidated in a different field setting (Chat-man & Flynn, 2005; Cialdini, 1995). Full-cycle re-search allows researchers to compensate for theweaknesses of one context or study design with thestrengths of another. It also allows researchers toinvestigate a broad initial question in a field setting(e.g., Is there bias in customer judgments?) and thenconduct a laboratory study that can utilize morecontrol and enable examination of more specificquestions in detail (e.g., What is a potential causeand consequence of bias in customer judgments?).Finally, the investigators can move back to a fieldsetting to confirm findings from the first two stud-ies. The interplay of field and lab designs pre-scribed by the full-cycle approach fosters greatertheoretical insight into the causality and generaliz-ability of study findings.

Following the full-cycle research model, wefirst tested for bias in customer judgments regard-ing a sample of professional employees (doctors).Next, in a carefully controlled laboratory setting,we tested for customer bias again, but this time ina bookstore. We also identified and measured aspecific mechanism that might explain the ob-served effect. Finally, we tested for customer biasin judgments of an organizational unit with asample consisting of country clubs belonging to alarge hospitality company. Since the focal unitrated shifted from the individual in the first sam-ple to the organizational in the third sample, wewere able to provide some initial confirmation ofthe generalizability of our theory. In the follow-ing section, we present the theoretical rationalefor our predictions regarding the possible effectsof customers’ diversity-related biases on their sat-isfaction judgments.

2010 239Hekman, Aquino, Owens, Mitchell, Schilpzand, and Leavitt

THEORETICAL BACKGROUNDAND HYPOTHESES

Customers are often asked to assess an individualservice provider, such as a salesperson, teller,teacher, or physician (Davis & Davis, 1999; Haas,Cook, Puopolo, Burstin, Cleary, & Brennan, 2000;Sixma, Spreeuwenberg, & van der Pasch, 1998); thequality of an environment in which they wereserved (e.g., the merchandise available, the new-ness or cleanness of the setting, or the efficiency ofthe technology) (Pellegrin, Stuart, Maree, Frueh, &Ballenger, 2001; Simonet, 2005); or the unit orgroup providing a service (e.g., a bank, school,country club, or law firm (Anderson et al., 2004;Ittner & Larcker, 1998). This last type of judgment islikely to include opinions about both the server(s)and the context in which the economic transactionoccurred and is therefore a more global judgmentthan the other two types previously mentioned. Inthis study, we investigated the possibility of sys-tematic bias in all three types of satisfaction judg-ments, thereby providing a strong test of the poten-tial generality of such biases. An assumption of ourstudy was that, like anyone else who makes a socialjudgment, customers are not immune to informa-tion-processing biases. One potential source of thisbias is the demographic characteristics of the ser-vice providers who are being rated.

Racial and Gender Biases in CustomerSatisfaction Judgments

U.S. society has made considerable progress inreducing overt expressions of prejudice since theCivil Rights Movement of the 1960s (Bobo, 1998).Yet despite these gains, there is abundant social-psychological evidence that biases against womenand minorities persist in a more covert and non-conscious form. Researchers have used terms suchas “modern racism” (McConahay, 1983), “aversiveracism and sexism” (Dovidio & Gaertner, 1981;Gaertner & Dovidio, 1986), and “implicit genderand racial stereotypes” (Greenwald & Banaji, 1995)to describe these types of biases, and many studieshave demonstrated their influence on informationprocessing and judgment in a variety of social do-mains (see Brief, Dietz, Cohen, Pugh, and Vaslow[2000] for an overview). For example, Bertrand andMullainathan (2004) showed that when job appli-cant resume quality was ambiguous, applicantswith African American–sounding names (e.g.,Aisha, Rasheed) were much less likely to be calledfor job interviews than applicants with white-sounding names (e.g., Kristin, Brad). Similarly,when evaluators of orchestra position applicants

could see an applicant’s gender, they were morelikely to select men. When the applicant’s gendercould not be observed, the number of women hiredsignificantly increased (Goldin & Rouse, 2000). Inanother study, Dovidio and Gaertner (2000) foundthat although raters were not biased against blacksin a simulated hiring decision when the applicantswere clearly qualified or unqualified for a job, rat-ers were biased when the applicant’s qualificationswere ambiguous. Dovidio and Gaertner (2000) in-terpreted this finding as supporting an aversiveracism framework in which prejudice occurs in amore subtle form in ambiguous or uncertain con-texts. Considerable evidence demonstrating the op-eration of covert and unconscious racial and genderbiases in a variety of social domains provides rea-son to suspect that such biases can also influencecustomer satisfaction judgments.

Our theoretical arguments supporting the in-fluence of racial and gender biases in customersatisfaction judgments are based on the idea thatobservers (e.g., customers) have preconceived ex-pectations about others depending on whether theperson being observed belongs to a high- or low-status demographic group (Berger, Conner, & Fisek,1974; Berger, Fisek, Norman, & Zelditch, 1977). Inthe United States, men and whites are consideredby most people to be members of a high-statussocial group relative to women and ethnic minori-ties (see Ridgeway [1991] for a review). One of thebenefits of belonging to a high-status social group isthat observers are more likely to make favorableinferences about one’s competence, normality,and legitimacy (Aquino & Bommer, 2003; Gian-nopoulos, Conway, & Mendelson, 2005; Sidanius& Pratto, 1999). In contrast, members of low-statusgroups are subject to negative stereotypes and attri-butions concerning their work-related competen-cies (Fernandez, 1981; O’Leary & Ickovics, 1992).

In view of rating theory (Wherry & Bartlett, 1982),we believe there are at least three reasons to expectcustomer satisfaction ratings to be susceptible tonegative stereotypes and racial and gender biases.First, an important difference between performanceratings made by supervisors and judgments madeby customers is that usually customers are affordedthe luxury of anonymity. Although supervisors’and customers’ ratings are both viewed by organi-zational administrators, in most cases only super-visors are identifiable. Customer anonymity de-creases accountability and the desire to engage inthe effortful cognitive processing required to con-ceal or overcome any biases (Lerner & Tetlock,1999; Richman, Kiesler, Weisband, & Drasgow,1999). Moreover, supervisors, but not customers,know that their ratings are part of the employee

240 AprilAcademy of Management Journal

record and are used for employee training, feed-back, and advancement decisions (Murphy, 1991).Supervisors can not only be identified, but alsomust justify their ratings, and as such they are evenmore motivated to engage in effortful informationprocessing to help them reduce the influence ofracial or gender bias and appear, at least superfi-cially, to be objective.

Customer anonymity does not motivate raters toreduce bias, and customer satisfaction question-naire instructions and items may even facilitate theexpression of such biases. Supervisors completinga performance appraisal are frequently reminded ofthe importance of rating accuracy, and they ratespecific behavioral items (Judge & Ferris, 1993;Kane, Bernhardin, Villanova, & Peyrefitte, 1995).Customers are typically asked only for their “opin-ions” or “attitudes” about employees or organiza-tions, which suggests that they “make a judgment”(Schneider, Ehrhart, Mayer, Saltz, & Niles-Jolly,2005). Common customer satisfaction items such as“I would recommend this organization to others”and “This organization/employee meets my expec-tations” do not solicit recall of specific employee ororganizational behaviors, and so they may providemore information about the state of mind of therater than about the actual performance of the rateeor the organization. Such items are also problem-atic because customers may have higher expecta-tions when rating women and nonwhites (Biernat &Kobrynowicz, 1997; Yarkin, Town, & Wallston,1982) and so provide lower customer satisfactionjudgments for such workers, even if their perfor-mance is objectively equivalent to that of their maleor white counterparts.

A variety of techniques has been shown to effec-tively reduce bias in performance appraisal. Forexample, Roch and O’Sullivan (2003) found that acombination of frame-of-reference (i.e., having rat-ers establish a prototype of good performance) andbehavioral observation training (i.e., focusing raterson specific behaviors) increased accuracy in ap-praisal. Baltes and Parker (2000) found that haloerror training (i.e., knowing what factors should notinfluence ratings) and structured recall memory in-tervention (i.e., memory enhancement techniques)reduced bias in performance ratings. DeNisi, Rob-bins, and Cafferty (1989) argued that behavioraldiaries aided in the recall and categorization ofbehavioral events. Customers are not trained in orexpected to use these techniques when formingsatisfaction judgments.

In sum, customer satisfaction judgments arelikely to be highly susceptible to racial and genderbiases because customers are usually anonymous,are asked to make summary judgments rather than

to accurately recollect performance-related behav-iors, and are untrained in techniques that mighthelp them overcome unconscious biases. Our argu-ments led us to expect that, in general, employeesbelonging to low-status demographic groups(women, racial minorities) will receive lower cus-tomer satisfaction scores than employees belongingto high-status demographic groups. But even if wefound evidence for this difference, it would notdemonstrate the operation of bias in customer judg-ments, because the lower customer satisfactionjudgments received by members of lower-status de-mographic groups might in fact indicate lower lev-els of true performance. Logically, customer satis-faction judgments should be at least partlyinfluenced by employee objective performance(Wherry & Bartlett, 1982). Consequently, more di-rect evidence of bias would be demonstrated ifbehaviors performed by a high-status group mem-ber are viewed more favorably by customers thanthe same behaviors performed by someone from alow-status group. Evidence for the plausibility ofthis hypothesis comes from studies showing thatwomen in leadership roles are rated lower thanmen in similar roles (Eagly, Makhijani, & Klonsky,1992) and that ethnic minorities and women arerewarded less than whites and males for exhibitingthe same advice-giving or ingratiatory behaviors(Westphal & Stern, 2007). There is also evidencethat racial minorities and women who achieve per-formance equivalent to that of whites and men arejudged as having less underlying ability (Biernat &Kobrynowicz, 1997; Yarkin et al., 1982). To exam-ine whether prior findings of biases in judgments ofcompetence based on a target person’s group mem-bership generalize to customer judgments, wetested the following hypothesis:

Hypothesis 1. The relationship between em-ployee objective performance and customersatisfaction judgments is attenuated for em-ployees belonging to low-status demographicgroups compared to employees belonging tohigh-status demographic groups.

Hypothesis 1 deals with customer judgments thatask for evaluations about an individual service pro-vider. But we also believe that racial and genderbiases can influence evaluations of the service pro-vider’s organizational context (e.g., its cleanness orappearance). Our prediction is drawn from the no-tion that the positive or negative properties of anitem or person can “spill over” onto the nearbycontext or surrounding targets (Rozin, Millman, &Nemeroff, 1986). Our logic is based on the simpleidea that evaluations of different aspects of serviceexperience (e.g., the employee, the context) are


connected in a rater’s unconscious and consciousbelief system (Argo, Dahl, & Morales, 2009; Morales& Fitzsimons, 2007). Theoretical and empiricalwork on the cognitive structure of attitudes (Ander-son, 1981; Wyer & Schrull, 1989) has suggested thatevaluation of any person, object, or idea is partlybased on evaluations of other persons, objects, orideas with which the target object is linked. Recentwork in marketing has elaborated on this idea(Keller, 2003). Research shows that the evaluationof a product, service, or brand is partly related tothe evaluation of the persons who are associatedwith the product, service, or brand (e.g., the personin an advertisement; the service provider) (Folkes &Patrick, 2003; Morales & Fitzsimons, 2007; Simo-nin & Ruth, 1998). Accordingly, we expect thatcustomers’ evaluations of organizational contextswill be unconsciously connected to their evalua-tions of highly visible employees. Although suchan unconscious connection may be unwarranted,we expect customers’ attitudes about low-statusemployees to be reflected in less favorable evalua-tions of an organizational context where such em-ployees are highly visible.

It may even be possible for customers to make aconscious connection between employee statusand the quality of an organizational context (e.g.,Rynes, Heneman, & Schwab, 1980; Rynes & Miller,1983; Spence, 1973). In line with this thinking, jobapplicants believe that recruiters’ competence andthoroughness signal their organization’s overallquality (Rynes, Bretz, & Gerhart, 1991). Customersmay judge an organizational context’s employmentof a low-status person as signaling physical andsocial inferiority to one employing a high-statusperson. Combining our spillover and signaling ar-guments, we expected the mere presence of mem-bers of lower-status groups in an organizationalenvironment to lead to less positive customer judg-ments of the environment:

Hypothesis 2. Individuals report lower cus-tomer satisfaction judgments of an environ-ment context when a highly visible employeein that environment belongs to a low-status(i.e., female, African American) rather than ahigh-status (i.e., male, white) demographicgroup.

All customers are not so susceptible to racial orgender bias that they will evaluate an employeeand environment more negatively simply becausethe employee belongs to a low-status demographicgroup. We expected the customers most prone tomaking these types of judgments to be those whohold more negative preexisting attitudes towardfemales or racial minorities. Individuals with neg-

ative preexisting attitudes toward members of low-status groups quickly associate negative words(e.g., “terrible,” “nasty,” “evil”) with pictures ofnonwhite or female faces, and they quickly associ-ate positive words (e.g., “laughter,” “glorious,”“joy”) with pictures of white or male faces (Green-wald & Banaji, 1995). We tested this possibility byexamining whether the degree to which individu-als have unconscious, negative attitudes towardmembers of low-status groups moderated the effectproposed in our first two hypotheses:

Hypothesis 3a. Low-status employees (i.e.,women, African Americans) receive even lowercustomer satisfaction judgments than equallyperforming high-status employees (i.e., men,whites) when the judges have negative atti-tudes toward members of low-status groups.

Hypothesis 3b. Contexts employing low-statusemployees receive even lower customer satis-faction judgments than contexts employingequally performing high-status employees,when the judges have negative attitudes to-ward members of low-status groups.

Thus far we have argued that customer satisfac-tion judgments can be influenced by perceiving asingle employee belonging to a low-status group ina service environment. But in many cases custom-ers interact with a variety of employees during agiven customer service encounter. For example,when doing something as simple as buying grocer-ies, customers may observe and interact with deliworkers, produce employees, cashiers, and baggers.Employees in each of these positions likely havedifferent demographic characteristics. For this rea-son, another way to examine the possible influenceof bias in the customer judgment process is to seewhether the demographic composition of an organ-izational unit influences customer judgments ofthat unit. Extending our theoretical argument to theorganizational level, we expected to observe thatthe degree to which an organizational unit’s em-ployees are members of low-status demographicgroups will influence customer satisfaction judg-ments of that organizational unit in such a way thatit will be judged as being of poorer quality than anorganizational unit whose employees are mostlyfrom high-status demographic groups. But as wenoted when predicting the influence of rater biason judgments of individual service providers, sucha finding might reflect true differences in unit per-formance. Hence, we make this more precise pre-diction that parallels Hypothesis 1:

Hypothesis 4. The relationship between an or-ganizational unit’s objective performance and


customer satisfaction is attenuated as a func-tion of the percentage of unit employees be-longing to low-status demographic groups (i.e.,women and minorities).

Our theoretical arguments and hypotheses aresummarized in the conceptual model shown inFigure 1.

We tested the hypotheses in our model in threestudies using different samples and methods. Wetested Hypothesis 1 by looking at customers’ judg-ments of their physicians (the Medcorp study); wetested Hypotheses 1, 2, and 3 by examining cus-tomer judgments of a bookstore and one of its em-ployees (the Bookcorp study); finally, we testedHypothesis 4 by examining customers’ judgmentsof their golf club (the Golfcorp study). In eachstudy, a large number of customers rated each ofthe targets.

MEDCORP STUDY

Our first sample was drawn from all 113 primarycare physicians (i.e., family practitioners) em-ployed by a large health maintenance organization(HMO), hereafter referred to as Medcorp (a pseud-onym). Medcorp provides coverage and health carefor about 350,000 people in the Pacific Northwestof the United States. Within our sample, 38.4 per-cent were women, 11.5 percent were ethnic minor-ities, and all had medical degrees. The 2006 Diver-sity Report by the Association of American Medical

Colleges reported that 24.5 percent of practicingphysicians were women and 12.1 percent werenonwhites.1

Measures

Medcorp routinely collected patient satisfactionratings as well as objective behavioral indicators ofphysician performance that were assumed to have adirect, positive impact on patient health and well-being. This feature of our data represents a meth-odological improvement over studies that haveonly measured employee performance with a singlesubjective rating and have therefore been unable todetermine if the rating was biased (Rotundo & Sack-ett, 1999). The dependent variable in our study waspatient satisfaction with a physician. The indepen-dent variables were physician demographic charac-teristics (race and gender) and three types of pa-tient-centered behaviors.

Customer satisfaction. Medcorp mailed a post-card survey to a percentage of each physician’spatients, following doctor visits, selecting the pa-tients so as to avoid a bias toward those patientswith frequent appointments. Patients completed

1 However, the Association of American Medical Col-leges also reports that these numbers are changing dra-matically as 44 percent of American medical school grad-uates in 2006 were women, and 34 percent werenonwhite.

FIGURE 1Conceptual Model of How Bias Influences Customer Satisfaction Ratingsa

H3b H2

Anonymity

Attributes of the Rating

Rating of Employee

Rating of Organizational Unit (Employee and Context)

Service Provider Performance

CustomerRacial/Gender Bias

Rating of Context

Organizational Unit Performance

H1

H4

Race/Sex of Employee

H3a

NoEvaluation Standard

Lack of Training

a Dotted lines indicate untested relationships. We expected main effects, but our contribution lies with tests of the interactions.


and returned a total of 12,091 surveys, which con-stituted a response rate of 52 percent, with eachphysician rated by an average of 107 patients. Eachpatient rated only one physician, so the individualratings were independent. Patients were asked:“How would you rate . . . ” (1) “the attention theprovider paid to you,” (2) “this provider’s thor-oughness and competence,” and (3) “your opportu-nity to ask questions of this provider” (1 � “verypoor,” to 5 � “excellent”). The three items werehighly correlated (the average correlation was .93),so the organization combined them to create a com-posite patient satisfaction variable. These itemscapture general rather than very specific behaviors(e.g., minutes spent with provider, the number ofquestions the doctor asked). The organization didnot provide us with access to raw patient-levelsurveys. Instead, it provided us with data indicat-ing what percentage of each physician’s patientsrated the physician as “excellent.” Thus, the rangeon this measure was from 0 to 100 percent. Physi-cians who were rated above average on this mea-sure received a 4 percent pay bonus. This measurewas collected in the same quarter as all othervariables.

Physician race. Medcorp identified each physi-cian’s race, and we coded whites as 0 and ethnicminorities as 1. Of the 113 physicians in the sam-ple, 10 were Asian or Pacific Islanders, 2 wereblack, and 1 was Native American. The percentageof ethnic minority physicians in our sample is con-sistent with the national average of 12.1 percent.

Physician gender. We coded males as 0 andfemales as 1. Forty-three of the physicians werefemale, which is slightly higher than the nationalaverage of 24.5 percent.

Objectively measured employee performance.With the growing prevalence of HMOs and theincreasing corporatization of medicine (Feinglass &Salmon, 1990), patients are increasingly beingviewed by organizational administrators and phy-sicians as customers. Therefore, physicians are in-creasingly being rewarded for behaviors that bene-fit their patients and the organization’s customers(Laine & Davidoff, 1996; Stewart et al., 2000). Weused the customer-benefiting behaviors identifiedby Medcorp as our indicators of objective physicianperformance.

Medcorp measures customer-benefiting behav-iors along three dimensions. The first is physicianproductivity, which is the number of health proce-dures performed and issues discussed in a giventime period. The second is the physician’s accessi-bility to customers, measured by the number ofsecure e-mails that doctors send to customers. Thethird is the physician’s level of quality, measured

by the standardized prescription rates of particularmedications for customers that possess precise dis-ease criteria. All three dimensions reflect behaviorsthat benefit customers by reducing the amount oftime and money they spend receiving medical care.For all these metrics, physicians are shown howthey compare to both the organizational goal andthe organizational average. More productive physi-cians are able to treat more customer problems pervisit, thereby saving customers’ time and trips tothe doctor. More accessible physicians providegreater convenience to customers, who can simplye-mail their medical questions. Higher-quality phy-sicians are better at preventing costly and deadlyhealth events such as strokes and heart attacks.Physician compensation is tied to each of thesecustomer-benefiting behaviors. Physicians who ex-ceed the 40th percentile are given a bonus.

Physician productivity. The average number ofpatients seen, medical issues discussed, and med-ical procedures performed by each doctor in a stan-dardized eight-hour day was recorded by the organ-ization’s scheduling software. Medcorp physicianshad a great deal of control over the amount of workthat they did in a day as they could control theintensity of each visit (e.g., the number of proce-dures performed and patient health issues ad-dressed). Organizational administrators controlledthe number of patients physicians saw each day.

The objective performance assessment we usedwas the composite of average face-to-face visits andphone visits adjusted by the average intensity ofeach visit. Intensity was measured by relative valueunits (RVUs), which physicians coded at the end ofeach visit using national guidelines. RVUs cap-tured the amount of time involved, the requiredphysical and mental effort, the required judgmentand technical skill, and the psychological stressentailed (Hsaio, Braun, Becker, & Thomas, 1988;Hsaio, Braun, Dunn, & Becker, 1988). Physicianschecked one of three RVU boxes after seeing eachpatient. If the patient appointment was a quickcheck-back or follow-up appointment, physicianschecked the first box, which was worth .5 RVUs. Ifthe patient appointment involved at least two pa-tient issues or concerns, but fewer than four, thenthe physician checked the middle box, indicating1.0 RVUs for that visit. If the patient appointmentinvolved five or more patient issues, then the phy-sician checked the third box, which indicated 1.5RVUs. According to quarterly audits by administra-tors, Medcorp physicians accurately recordedRVUs in 90 percent of patient visits. Coding errorsresulting from physicians coding too many or toofew RVUs were normally and equally distributed.The raw measure of productivity was standardized


on the basis of physician full-time status and thenmultiplied by each physician’s average visit inten-sity to obtain the quarterly average RVU-adjustedpatient encounters per day.

Physician accessibility. The average daily num-ber of secure e-mails that physicians sent to pa-tients for the quarter was used to measure anothertype of customer-benefiting behavior. Patientshighly value the ability to easily contact their phy-sician. Indeed, a Harris poll has shown that 90percent of Americans who are online want the abil-ity to e-mail their physicians, and 37 percent areeven willing to pay for it (Taylor, 2002). Medcorppatients and physicians could communicate elec-tronically regarding health-related issues through aportal designed exclusively for patient-doctor com-munication. To use the system, patients loggedonto a secure website that provided access to theirpersonal health records, their lab results, and a hostof health-related information. Patients could sendunlimited e-mails through the portal to any physi-cian they had visited in the prior two years at nocost, and Medcorp physicians were expected toreply to each message within 24 hours. Patientswere encouraged to contact their physicians via thesystem to ask basic health-related questions, to re-quest prescription refills, and to schedule fol-low-up appointments. Medcorp administratorsassigned an equal number of patients to each phy-sician (taking into account patient sickness, age,and gender) and thought that the system savedpatients doctor visits, thereby saving them time andmoney. In general, physicians do not think thate-mail improves the quality of patient care, butrather that it increases convenience for patients(Kleiner, Akers, Burke, & Werner, 2002).

The Medcorp computer server automatically re-corded the number of e-mails that each medicalprofessional sent to his or her patients. Medicalprofessionals had a great deal of control over howmany e-mails they sent for two reasons: (1) theycould try to discourage patients from using thesystem, and (2) they could choose whether to per-sonally respond to their patients’ e-mails. We cal-culated the number of e-mails physicians sent perday, taking into account the number of full workingdays that they were in clinic during the study pe-riod. To enhance the normality of the variable, weused an inverse transformation and then reflectedthese values so that higher values representedgreater use (Tabachnick & Fidell, 2003).

Physician quality. Every Medcorp primary carephysician was responsible for a panel of member-patients. Of the thousands of possible treatments,prescriptions, and procedures that physicianscould perform to benefit patients, one of the most

important was prescribing statins and angiotensin-converting-enzyme (ACE) inhibitors to patientswith cardiovascular disease. Treatment of cardio-vascular events such as strokes, clots, and heartattacks is the biggest healthcare cost in the UnitedStates (Willerson & Cohn, 2000), and these drugsprevent cardiovascular events over patients’ life-times (Gerstein et al., 2000).2

According to Medcorp guidelines, all patientswith cardiovascular disease should be regularlytaking ACE inhibitors and some form of a statin.ACE inhibitors lower blood pressure, and statinslower cholesterol. Nationally, only 50 percent of allcardiovascular disease patients who should betreated with statins and ACE inhibitors are cur-rently taking such medication (Dubois et al., 2002).These drugs significantly lower the immediate riskof a cardiovascular event (e.g., stroke, heart attack)for all individuals, regardless of sex or previoushistory of cardiovascular disease (LaRosa, He, &Vupputuri, 1999; Yusuf, Sleight, Pogue, Bosch, Da-vies, & Dagenais, 2000). To promote a higherprescription rate, Medcorp administrators send e-mails to physicians reminding them to prescribesuch treatment. Although these medications bene-fit patients by helping them avoid death and reducehealthcare expenses, physicians often forget to pre-scribe them (Isles, 2002).

Our quality variable was a composite of the per-centage of cardiovascular disease patients 18 yearsand older who had been dispensed the equivalentof a 90-day supply of ACE inhibitors and statins atany time within the quarterly reporting period. Thecomponent variables approached normality, werestandardized, and were added together. The result-ing variable was each physician’s overall prescrip-tion rate of statins and ACE inhibitors for cardio-vascular disease patients. The average prescriptionrate at Medcorp was 50 percent, which was similarto the national average.

Control Variables

We controlled for several variables that were notof direct interest for testing our hypotheses butcould be theoretically related to the dependentvariable and might provide plausible alternativeexplanations for our findings.

Average practice busyness. Patients who had towait a long time to see their physicians might be

2 We call this variable “quality” because statin andACE inhibitor prescription rate accuracy are measures ofphysician quality according to the most influential qual-ity assurance organizations (e.g., HEDIS, NCQA, and IHI).


less satisfied, so we controlled for the busyness ofeach physician’s practice. At the close of businesseach day, the Medcorp computer counted howmany days into the future each physician’s thirdavailable appointment was. According to the Na-tional Quality Measures Clearinghouse, countingthe days until the third next available appointmentis the healthcare industry’s standard measure ofaccess to care and indicates how long a patientwaits to be seen. Doctors who are not very busytypically have three available appointments thenext day, whereas busy doctors often do not havethree available appointments for several days. Thefinal variable was the quarterly average number ofdays until each physician’s third open appoint-ment slot.

Physician full-time status. We included thenumber of hours a physician worked in our modelbecause patients might be more satisfied if theirphysician worked more hours. Physicians rangedfrom working 30 to 100 percent of full time.

Number of patients in panel. Medcorp assignedphysicians to care for a particular group (i.e., apanel) of patients. Patients in larger panels may beless satisfied, and so we controlled for the totalnumber of patients in each physician’s panel stan-dardized by the full-time status of the physician.

Average patient age. Older patients might havedifferent expectations about doctor demographiccharacteristics, so we included the average patientage for each physician’s panel in our model.

Average chronic sickness of panel. Sicker pa-tients might be less satisfied, so we controlled forthe panel chronic sickness variable calculated byMedcorp, which captured, for example, the per-

centage of patients with diabetes and cardiovascu-lar disease.

Physician age and tenure. Physicians who wereolder or who had been employed by Medcorp formore years might have more loyal, satisfiedpatients.

Physician tenure by objective performance. Be-cause women and nonwhite physicians tended tobe more recently hired than male and white physi-cians, any influence of physician race and genderon customer satisfaction might be masked by phy-sician tenure. We therefore included the interac-tions of tenure by objective performance in ourmodels so that we could more clearly determine theinteractive influence of physician race by objectiveperformance and physician gender by objectiveperformance on customer satisfaction.

Results

Table 1 reports the means, standard deviations,and correlation coefficients for the dependent, in-dependent, and control variables in the Medcorpstudy. We found no significant differences in ourobjective measures of performance based on em-ployee race and gender. Our first hypothesis statesthat the relationship between employee objectiveperformance and customer satisfaction judgmentsis less positive for employees belonging to low-status demographic groups than for employees be-longing to high-status demographic groups. To testthis statement, we examined the interactions of theobjective measures of employee performance (i.e.,quality, productivity, and accessibility) by em-ployee race and gender. We used hierarchical mod-

TABLE 1Descriptive Statistics, and Correlations, Medcorp Samplea

Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 12

1. Patient satisfaction 0.51 0.112. Practice busyness 0.66 0.47 �.303. Full-time equivalent 0.80 0.20 �.07 .114. Number of patients in

panel1,749.77 550.63 �.10 .26 .59

5. Panel age 45.84 4.89 .07 �.07 .05 �.036. Chronic sickness of

panel1.04 0.12 .13 �.12 �.15 �.13 .55

7. Tenure with Medcorp 14.81 8.51 .20 �.14 .14 .08 .33 �.208. Age 50.34 6.58 .09 �.09 .16 .21 .29 �.05 .699. Nonwhite 0.12 0.32 �.15 .02 .01 �.04 �.14 �.05 �.12 �.03

10. Female 0.38 0.49 �.06 �.06 �.63 �.44 �.21 .04 �.25 �.31 .1211. Productivity 23.00 1.97 .05 .12 .22 .30 �.06 .22 �.25 �.01 �.04 �.1512. Quality �0.01 1.55 .11 .03 .07 .08 .21 .05 .14 .11 .02 .04 �.0313. Accessibility to patients 0.16 0.15 .13 �.11 �.18 �.23 .05 �.06 �.04 �.16 �.11 .21 .05 .23

a n � 113; all correlations larger than .15 are significant at p � .05.


erated regression models to do so, centering allvariables involved in the interaction terms to min-imize multicollinearity between these terms andtheir individual components (Aiken & West, 1991).We entered all of the control variables in model 1.In model 2 we entered the control variables plusthe interactions involving physician gender. Inmodel 3 we entered all the control variables as wellas the interactions involving physician race. Fi-nally, in model 4, we entered all control variablesand all interaction effects. Table 2 presents theresults of this analysis.

The two-way gender by objective performance

interactions as a set explained a significant amountof incremental variance in the dependent variable(R2 � .07, p � .01), providing preliminary supportfor Hypothesis 1. Inspection of the individual re-gression weights showed that the physician acces-sibility by gender and physician quality by genderinteractions were significant (p � .05). We probedthe pattern of the interaction by examining thesimple slope of the objective performance measuresfor male and female physicians (Aiken & West,1991). Figure 2 graphically shows the results of thisanalysis.

The figure shows a stronger positive relationshipbetween physician customer-centered behaviorsand performance ratings for men than for women.We calculated the significance of the simple slopesfor interactions. The coefficient of the simple slopeof quality behaviors on customer satisfaction wassignificant and positive for male physicians (b �.32, p � .01), but it was not significant for femalephysicians (b � �.01, n.s.). The coefficient of thesimple slope for accessibility behaviors was posi-tive for male physicians (b � .13, n.s.) but negativefor female physicians (b � �.17, n.s.). Althoughneither of the simple slopes for accessibility behav-iors were significantly different from zero, theywere significantly different from each other (p �.05). By looking at the plots, one can also see thatthe interaction is a crossover, which shows that thedirection of the relationship is the opposite formembers of high- versus low-status demographicgroups.

The two-way race by objective performance in-teractions as a set explained a significant amountof incremental variance in the dependent vari-able (R2 � .08, p � .05), providing preliminarysupport for Hypothesis 1. Inspection of the indi-vidual regression weights showed that the physi-cian productivity by race and physician qualityby race interactions were significant (p � .05).Figure 2 again shows the forms of these interac-tions. Simple slope analysis revealed that thecoefficient of the simple slope of quality on cus-tomer satisfaction was significant and positivefor white physicians (b � .29, p � .01) but nega-tive and nonsignificant for nonwhite physicians(b � �.14, n.s.). The simple slope of productivitybehaviors on customer satisfaction was nonsig-nificant and positive for white physicians (b �.15, n.s.) but significant and negative for non-white physicians (b � �.32, p � .01), whichagain indicates a crossover interaction effect.Overall, we found support for four of the relation-ships predicted in Hypothesis 1.

TABLE 2Moderating Effects of Physician Race, Gender, andObjective Performance on Patient Satisfaction with

Physician, Medcorp Studya

Variables Model 1 Model 2 Model 3 Model 4

ControlsPractice busyness �.23** �.18* .26** �.22*Full-time equivalent �.17 �.23 �.20 �.24*Number of patients

in panel�.04 �.03 .01 .01

Panel age �.23* �.27* �.20 �.24*Chronic sickness of

panel.23* .25* .20 .22*

Tenure withMedcorp

.34** .35* .38** .39**

Age �.13 �.12 �.17 �.16Productivity .09 .09 .12 .14Quality .17* .20* .16 .17*Accessibility .15 .13 .07 .05Female �.22* �.26* �.17* �.20*Nonwhite �.12 �.09 �.18* �.17*

InteractionsProductivity �

tenure.06 .09 .01 .03

Quality � tenure .13 .13 .14 .14Accessibility �

tenure.12 .01 �.16 .06

Productivity �female

.04 .00

Quality � female �.18** �.18**Accessibility �

female�.16* �.16*

Productivity �nonwhite

�.18* �.21**

Quality � nonwhite �.18* �.16*Accessibility �

nonwhite�.13 �.10

Adjusted R2 .17 .21 .23 .25R2 .25 .32 .33 .40�R2 from model 1 .07** .08** .15**

a n � 113. The sample consisted of physicians, 100 whites, 10Asians or Pacific Islanders, 2 blacks, and 1 Native American.

* p � .05** p � .01


Discussion

Our first study explored whether customers, whoin this case were patients of an HMO, expressedtheir race- and gender-based biases in customersatisfaction judgments. We found that objectivelymeasured behaviors were only positively related tocustomer satisfaction for physicians who werewhite or male. We also found that one type ofcustomer-centered behavior was significantly neg-atively related to customer satisfaction for womenand nonwhite physicians. This second finding wasan even stronger result than we anticipated becauselogically one might expect the relationship be-tween customer-benefiting behaviors and customersatisfaction to be weaker, but still positive, forwomen and nonwhites than for men and whites.

The observed pattern of relationships indicatesthat biases against nonwhite and female employeesmay creep into satisfaction judgments. However,we must also consider this study’s shortcomings.First, the relationships between the dependentvariable and both panel age and physician age haveopposite signs in the correlation table versus theregression table. This pattern suggests that these

variables may have somehow influenced our re-sults by suppressing variance in the dependentvariable that was irrelevant to its prediction(Tabachnick & Fidell, 2003). However, the correla-tions are not statistically significant, and our re-sults are substantively unchanged regardless ofwhether these variables are included in the model,so we believe our results are not due to suppressionby those variables. A more serious limitation is thatour Medcorp study only included a small percent-age of nonwhites. Moreover, many of the non-whites were Asians rather than African Americans.Biases against African Americans are more negativethan those associated with Asians (Song, 2004),and so a study that included African Americansmight be better able to detect the influence of suchbiases on customer satisfaction judgments. Wewere also not able to control for employees’ accentsor differences in their language and communica-tion styles, or whether customers felt certain em-ployees had names that sounded nonwhite. It ispossible that the biases we observed were the resultof some contextual variable such as employee lan-guage skill and not of customer prejudices. Finally,

FIGURE 2Interactive Effects of Physician Objective Performance and Demographic Characteristics on Patient

Satisfaction with Physician, Medcorp Studya

PatientSatisfaction

Physician Quality

(2c) Physician Quality by Race

(2a) Physician Quality by Gender

(2d) Physician Productivity by Race

(2b) Physician Accessibility by Gender

Nonwhite physician (n.s.)

Low

Low

High

High White physician (**)

PatientSatisfaction

Physician Productivity

White physician (**)

Low

Low

High

High

Nonwhite physician (**)

PatientSatisfaction

Physician Quality

Male physician (**)

Female physician (n.s.)

Low

Low

High

High

PatientSatisfaction

Physician Accessibility Low

Low

High

High

Male physician (n.s.)

Female physician (n.s.)

a Notations in parentheses denote the significance of the simple slopes. “**” � p � .01; “n.s.” � p � .10.


we did not measure whether customer raters hadpreexisting bias against women and minorities.That is, we had no assessment of the raters’ stereo-types or racial/gender biases as potential causes oftheir ratings. We designed our second study to ad-dress the limitations of our Medcorp study. Theoccupation we chose for our laboratory study wasservice employees working in a university book-store, and our raters were college students. We alsoused an experimental design to control for extrane-ous variables that might have influenced the resultsof our Medcorp study.

BOOKCORP STUDY

In our second study, Bookcorp, student raterswere asked to observe a video of an employee-customer interaction in a university bookstore, toevaluate the employee’s behavior, and to providesatisfaction judgments of the store environment.This study differed from our Medcorp study in avariety of ways. First, we controlled for the job-related behavior of the employees with a scriptedinteraction and varied only whether the behaviorwas performed by a male versus a female or a whiteversus an African American employee. This aspectof the Bookcorp study’s design allowed us to re-duce variability in employee behavior, thereby pro-viding a better test of whether the same behaviorwould produce different customer satisfactionjudgments depending on the employee’s gender orrace. Second, we assessed how participants, whowere asked to assume the role of customers, notonly evaluated the employee (as in our Medcorpstudy) but also evaluated the organizational context(the bookstore) in which the employee-customerinteraction took place. Third, we assessed each par-ticipant’s implicit bias toward women and non-whites to see if these unconscious attitudes mightpartly explain gender or racial bias in the ratings.

Sample

Eighty-six students from a major northwesternU.S. public university watched two videos of abookstore employee interacting with a customerand were asked to evaluate the employee and thebookstore. The bookstore in the video clips was at alarge eastern U.S. university, and it is highly un-likely that any of the participants had visited it.The “employees” and “customers” were hired pro-fessional actors, and the scripted interaction hadbeen filmed before the bookstore opened in themorning, although our raters taking the customerperspective were not aware of this. Participantswere randomly assigned to view either the white

male employee (n � 33), white female employee (n� 21), or black male employee (n � 34). Overall,substantial percentages of our participants werenonwhite (43 percent) or female (38 percent). Thisheterogeneous sample of raters was representativeof the population of people using the book store.

Design

Our design was a mixed factorial, with one be-tween-subjects factor (employee demographic char-acteristics) and one within-subjects factor (employ-ee-customer interaction). We treated employeedemographic characteristics as a between-subjectsfactor to reduce participants’ awareness that theywere participating in a race- or gender-relatedstudy. We presented all participants with two vid-eos depicting different employee-customer interac-tions. One video involved the employee ringing upa book and telling the customer that the book’sprice in the computer was higher than its price onthe shelf. The other video showed the same em-ployee being unable to help a customer find a bookthe customer wanted. Each video was about oneminute in length. Each participant saw both videosof the same employee. We randomly assigned theorder of the videos’ presentation within each ex-perimental condition and found no evidence thatorder influenced customer ratings. The customerand employee interactions were scripted to ensurethat behavior was equivalent across conditions. Allvideos were filmed in the same bookstore so thestore background was equivalent across conditions.The interactions in the video were designed to de-pict a moderate level of service quality to avoid“floor effects” (nearly everyone would rate perfor-mance as very poor) and “ceiling effects” (nearlyeveryone would rate performance as very good).We reasoned that making employee performancemore ambiguous would allow us to more effec-tively detect the influence of unconscious biases onratings of service quality and context because it hasbeen shown that the effects of stereotypic biasescan be weakened when people are asked to evalu-ate employees who are performing extremely wellor poorly (Dovidio & Gaertner, 2000). We assessedwhether the videos did in fact produce ambiguousjudgments of employee service quality by askingparticipants to report their satisfaction with theemployee using the customer satisfaction with em-ployee items described in the next section. Resultsshowed that the employee behaviors were per-ceived as being somewhat negative, as evidencedby satisfaction ratings of the videos averaging 3.27on a seven-point scale. That this rating was notextremely negative suggests that our videos were


not susceptible to floor effects. After watching eachvideo, participants completed a distraction task (anunrelated questionnaire) to clear short-term mem-ory before they rated satisfaction and context. Thisprocedure simulated the real-world situation inwhich customers typically do not complete servicequality ratings immediately after observing an em-ployee, but only after having engaged in other ac-tivities (e.g., dinner, shopping) that can take theirattention away from their encounter with theemployee.

Dependent Variables

Customer satisfaction with the employee. Rat-ers were asked how satisfied they were with (1)speed of service, (2) quality of service, (3) availabil-ity of staff for assistance, and (4) employee respon-siveness to customers’ issues and concerns (1 �“very poor,” 7 � “excellent”). Coefficient alpha forthe items was .74. This measure was adapted froman existing customer satisfaction survey we ob-tained from a large organization. The Appendixgives the customer satisfaction items used in thethree studies.

Customer satisfaction with the context. Raterswere asked how satisfied they were with (1) thebookstore’s appearance, (2) the degree to which thebookstore was conducive to learning, (3) whetherthe bookstore had up-to-date equipment, (4) thedegree to which the bookstore’s facilities were vi-sually appealing, (5) whether the bookstore’s ap-pearance was in keeping with the type of servicesprovided, (6) the bookstore relative to their expec-tations, and (7) their likelihood of recommendingthe bookstore to others (1 � “very poor,” “less thanexpected,” “definitely would not,” or “strongly dis-agree”; 7 � “excellent,” “better than expected,”“definitely would,” or “strongly agree”). This mea-sure (� � .76) was also adapted from an existingcustomer satisfaction survey we obtained from alarge organization.

Predictor Variables

Experimental condition. We had two condi-tions—one for race and one for sex. The sex con-dition included participants who viewed the whitemale or white female employee (1 � “participantviewed two videos of a white female employee,”0 � “participant viewed two videos of a white maleemployee”). The race condition included partici-pants who viewed either the white or the nonwhitemale (1 � “participant viewed two videos of anonwhite male employee,” 0 � “participantviewed two videos of a white male employee”).

Participants completed survey questions only afterwatching both videos.

Implicit bias. To measure raters’ racial and gen-der prejudices, we administered two Implicit Asso-ciation Tests (IATs) constructed to capture eachparticipant’s level of unconscious bias against non-whites and women (Greenwald, Nosek, & Banaji,2003). We should note that shortcomings of IATmeasurement (Blanton & Jaccard, 2006), such as noabsolute zero point and equal intervals, makes theinterpretation of scores somewhat unclear. How-ever, we chose to use the IAT as opposed to othertypes of bias measures (e.g., the modern racismscale [McConahay, 1986]) because it is more diffi-cult for participants to hide prejudices on the IATthan on explicit measures (Nosek, 2005). The gen-der IAT was administered after the participantssaw the videos and made their customer satisfac-tion judgments, but the race IAT was administeredbetween the videos and the ratings. Prior researchindicates no evidence of order effects for the IATand dependent variables—probably because sub-jects still respond in socially desirable ways on theexplicit measures (Greenwald, Poehlman, Uhl-mann, & Banaji, 2009).

Importantly, implicit attitudes appear to be betterpredictors of behavior than their explicit counter-parts, especially when social sensitivity concernsare high (Greenwald et al., 2009). For instance,implicit (but not explicit) attitudes about AfricanAmericans have been shown to predict desire towork with an African American partner on an in-tellectual task (Ashburn-Nardo, Knowles, & Mon-teith, 2003), and nonverbal actions (eye contact andother “friendly” behaviors) toward African Ameri-can interaction partners (McConnell & Leibold,2001). Though the correlation between implicit andexplicit attitudes varies across domains (Nosek,2005), the predictive validity of each suggests thatthey represent independent processes that explainunique variance in behavioral outcomes (seeGreenwald et al. [2009] for a meta-analysis of thepredictive validity of the IAT).

Control Variables

We controlled for rater race, gender, and age toaccount for rater demographic characteristics thatmight plausibly influence reactions to employeedemographic characteristics.

Results

We regressed the employee and organizationalcontext judgments on our controls, predictors, andinteraction to determine the degree to which the


customer judgments reflected race and gender bias.Tables 3 and 4 present results of the regressionmodels we used to test Hypotheses 1, 2, and 3.

Hypothesis 1 states that the relationship betweenemployee objective performance and customer sat-isfaction judgment is less positive for employeesbelonging to low-status demographic groups than

for employees belonging to high-status demo-graphic groups. Recall that the objective perfor-mance of each employee was made comparable byusing an equivalent behavioral script. Model 2 inTable 3 shows that the raters were significantly lesssatisfied with women employees than with theirequally performing white male counterparts (�R2 �

TABLE 3Effects of Employee Race and Gender on Customer Satisfaction with Employee, Bookcorp Studya

Variablesb

Customer Satisfaction with Employee

Female Condition Nonwhite Condition

Model 1 Model 2 Model 3 Model 1 Model 2 Model 3

Nonwhite �.25* �.26 �.27 �.23* �.21 �.20Female �.15 �.23 �.21 �.15 �.15 �.16Age �.09 �.05 �.05 �.22* �.23* �.21IAT score .05 .15 �.13 �.15 �.14 �.19Female condition �.28* �.26*IAT score � female condition �.06Nonwhite condition �.02 �.01IAT score � nonwhite condition �.28**

Adjusted R2 .03 .08 .08 .08 .08 .14R2 .10 .16 .16 .14 .14 .22�R2 from previous model .06* .00 .00 .08**

a For the nonwhite experimental condition, 1 � “nonwhite male employee,” 0 � “white male employee”; n � 67. For the femalecondition, 0 � “white male employee,” 1 � white female employee”; n � 54.

b “IAT” is the Implicit Association Test.* p � .05

** p � .01

TABLE 4Effects of Employee Race and Gender on Customer Satisfaction with Organizational Context, Bookcorp Studya

Variablesb

Customer Satisfaction with the Context

Female Condition Nonwhite Condition

Model 1 Model 2 Model 3 Model 1 Model 2 Model 3

Nonwhite �.03 �.03 �.07 �.16 �.08 �.08Female .09 �.03 .09 .16 .01 .01Age �.26 �.19 �.14 �.17 �.04 �.04IAT score �.12 .03 �.04 .13 .09 .02Female condition �.45** �.38**IAT score � female condition �.23*Nonwhite condition �.44*** �.46***IAT score � nonwhite condition �.18*

Adjusted R2 .00 .15 .17 .00 .18 .20R2 .07 .24 .29 .07 .22 .26�R2 from previous model .17** .04** .15*** .04*

a For the nonwhite experimental condition, 1 � “nonwhite male employee”; 0 � “white male employee”; n � 67. For the femalecondition, 0 � “white male employee,” 1 � white female employee”; n � 54.

b “IAT” is the Implicit Association Test.* p � .05

** p � .01*** p � .001


.06; b � �.28, p � .05). However, we did not findevidence of bias in customer satisfaction judgmentsof the nonwhite employee (b � �.02; n.s.). Thus,Hypothesis 1 received partial support.

Hypothesis 2 states that people report lower cus-tomer satisfaction judgments of an environmentwhen an employee in that environment belongs toa low-status demographic group (i.e., female, Afri-can American) rather than a high-status one (i.e.,male, white). Model 2 of Table 4 shows there is asignificant main effect of race and gender on judg-ments of the store environment. Indeed, model 2 ofTable 4 shows a main effect of the conditions inwhich a female (�R2 � .17; b � �.45, p � .01) anda nonwhite were rated (�R2 � .15; b � �.44, p �.001), suggesting that raters’ biases influence judg-ments of organizational context. We found strongsupport for Hypothesis 2.

Hypothesis 3a suggests that people report evenlower customer satisfaction judgments of an em-ployee belonging to a low-status demographicgroup when they have negative implicit attitudestoward that group. Model 3 of Table 3 shows thatthe coefficient for the interaction of IAT score by

nonwhite condition is significant and in the ex-pected direction for customer satisfaction with theemployee (�R2 � .08; b � �.28, p � .01). To gainmore insight into this effect, we plotted the inter-action and analyzed the simple slopes, as shown inFigure 3. Individuals with high levels of implicitbias (�1 s.d.) were significantly more likely to re-port lower satisfaction with the nonwhite male’sperformance than with the white male’s (p � .01).However, the coefficient for the interaction of IATscore by gender condition was not significant.

As for Hypothesis 3b, model 3 of Table 4 showsthat the coefficient for the interaction of IAT scoreand race is significant and in the expected directionfor customer satisfaction with the rated context(�R2 � .04; b � �.18, p � .05). We plotted theinteractions and conducted a simple slope analysis(see Figure 3). Customer IAT score (�1 s.d.) waspositively related to customer satisfaction with thecontext when customers were observing a whitemale employee (b � .33, p � .01) but was nega-tively related to customer satisfaction with the con-text when customers were observing a nonwhitemale employee (b � �.21, p � .05). The coefficient

FIGURE 3Interactive Effects of Employee Demographic Characteristics and Customer Implicit Association Test Score

on Customer Satisfaction with Employee and Store Context, Bookcorp Studya

Satisfaction with Context

Implicit Association Test Score

(3a) Employee Satisfaction by Race

(3b) Context Satisfaction by Race (3c) Context Satisfaction by Gender

White male employee (*)

Low

Low

High

High

White female employee (*)

Satisfaction with Employee

Implicit Association Test Score Low

Low

High

High

White male employee (n.s.)

Nonwhite male employee (**)


Implicit Association Test Score Low

Low

High

HighWhite male employee (**)

Nonwhite maleemployee (*)

a Notations in parentheses refer to the equation represented by the slope. “**” � p � .01; “*” � p � .05; “n.s.” � p � .10.


for the interaction term IAT score by gender condi-tion was significant and in the expected directionfor customer satisfaction with the context (�R2 �.04; b � �.23, p � .05). Customer IAT score (�1s.d.) was positively related to customer satisfactionwith the context when customers were observing awhite male employee (b � .23, p � .05) but wasnegatively related to customer satisfaction with thecontext when customers were observing a whitefemale employee (b � �.21, p � .05). Overall, wefound three significant coefficients supporting Hy-pothesis 3. These results and plots suggest thatjudgments of both employees and organizationalcontexts are vulnerable to unconscious biases.

Discussion

We found that student participants taking a cus-tomer perspective rated employees and an organi-zational context as worse when observing the per-formance of a low-status employee, and this wasespecially true if the raters held implicit biasesabout that low-status group. The results for Hy-pothesis 1 were not as strong as we anticipated,suggesting that it may be possible to minimize bi-ases by changing the setting in which rating takesplace. The laboratory context was less anonymousthan a typical customer satisfaction questionnairesetting, which may have weakened the influence ofbias on customer satisfaction judgments of employ-ees. Although we told the participants that theirresponses were anonymous, they may have feltscrutinized because they provided their judgmentswhen an experimenter was present and wrote theirnames on a separate sign-in sheet to receive partic-ipation credit for a class. Additionally, administra-tion of the IAT for race prior to the customer satis-faction ratings may have alerted participants thatthey were in a race and gender study, which mayhave weakened the Bookcorp study results for race.We observed, though, that the effects of bias onjudgments of the organizational context were stillquite strong, which makes sense to us because par-ticipants may have been unaware of and thereforeunable to suppress biases that spilled over ontotheir judgments of the organizational context.

As summarized by Hypothesis 4, we expectedthe relationship between an organizational unit’sobjective performance and customer satisfaction tobe less positive for organizational units with higherpercentages of employees belonging to low-statusdemographic groups (i.e., women and racial minor-ities) than for units with higher percentages of em-ployees belonging to high-status demographicgroups (i.e., men and whites). By returning to thefield to test this hypothesis, we completed a full

cycle of research and assessed the generalizabilityof our theory to a different organization.

GOLFCORP STUDY

Our sample was drawn from a large country cluborganization, hereafter referred to as Golfcorp. Golf-corp has 66 country clubs across the United Statesand roughly 70,000 customer-members, and it em-ploys approximately 8,000 people. Our sampleconsisted of all 66 Golfcorp country clubs. In oursample, 31.4 percent of employees were women,18.1 percent were Latino, 6.7 percent were AfricanAmerican, and 1.7 percent were Asian American orNative American.

Measures

Golfcorp routinely collects customer satisfactionratings as well as objective indicators of facilityperformance that are assumed to have a direct,positive impact on customers’ service experiences.The dependent variable in our study was customersatisfaction with a facility. The independent vari-ables were each club’s employee demographiccharacteristics (percents nonwhite and femaleemployees) and two types of objective clubperformance.

Customer satisfaction with facility. Like manyorganizations, Golfcorp measures customer satis-faction with a quarterly survey, which is mailed toa percentage of each facility’s customers. An aver-age of 63.8 customers rated each facility. The aver-age response rate per facility was 27.3 percent (anaverage of 234 surveyed customers per facility).The marketing company hired to do the customersurvey randomly sampled each facility’s customerseach quarter until they got either 20 respondents or3 percent of the total customer base (whichever waslarger). The items used for this measure reflected afocus on the facility context (quality of its club-house and golf course) and overall ratings of thefacility; it was thus similar to what was used in theBookcorp study. Customers rated each of theseitems (� � .81): “How would you rate the followingaspects of your club . . . ” (1) “maintenance ofgrounds/appearance of clubhouse,” (2) “lockerrooms and restrooms,” (3) “quality of greens,” (4)“condition of course,” (5) “pace of play,” (6) “con-dition of practice facilities,” (7) “ability to obtaindesired tee times” (items 1–7 were rated as 1 �“very poor,” 5 � “very good”); (8) “club meetsexpectations” (1 � “less than expected,” 5 � “bet-ter than expected”); and (9) “likelihood of recom-mending club to others” (1 � “definitely will not,”5 � “definitely will”). In our analyses, we lagged


this measure six months after the independent vari-ables to more conclusively show that employeedemographic characteristics and objectively mea-sured performance influence customer ratings,rather than the other way around.

Percent nonwhite. Golfcorp identified the per-centages of white and nonwhite employees in eachfacility. Across the 66 clubs in the sample, 26.4percent of employees were nonwhite. According tothe U.S. Census Bureau’s 2000 census, the percent-age of ethnic minority employees in our samplewas consistent with the national average of 28 per-cent of the U.S. population.

Percent female. We also obtained this variablefrom Golfcorp records. Thirty-one percent of em-ployees in our sample were women, which Golf-corp leaders believed to be consistent with thecountry club industry average. However, the per-centage of women in our sample was lower than thepercentage of women in all industries, which is 46percent according to the Bureau of Labor Statisticsat the U.S. Department of Labor.

Objective facility characteristics. As in our firsttwo studies, we wanted to clearly identify the por-tion of variance attributable to customer bias versusthe portion attributable to better facility perfor-mance. We therefore used two attributes as ourindicators of objective facility performance: facilityproductivity and facility attribute quality. Both di-mensions reflect facility characteristics that benefitcustomers. Facilities with more productive em-ployees create more value, both for Golfcorp andfor customers. Indeed, Golfcorp executives told usthat facilities with higher productivity valuescharged lower dues to members, were more profit-able, and were simply better-run. Higher attributequality benefits customers by allowing them to en-joy newer and better facilities. Managers wereshown how they compared to other facilities interms of quality and productivity. Employee com-pensation was tied to the productivity measure, butnot to the quality measure. Employees in facilitiesthat were above average in productivity were givena bonus.

Facility productivity. Facility productivity wascalculated by Golfcorp’s central accounting officefor the calendar year ending six months before thedependent variable was collected. This variablewas each club’s annual profits divided by its aver-age number of employees in that year. The numberof employees at each club was centrally controlled,with clubs with more members allotted proportion-ally more employees by the central office. There-fore, facility productivity was determined by theemployees’ effectiveness at creating value.

Facility quality. Over time, the condition of golfcourses and clubhouses deteriorates, and they needto be rebuilt or refurbished. Golfcorp assessed thequality of the course and clubhouse of each club toensure that customers were receiving a high stan-dard of service. For the courses, the rating scale was1, “more than 40% of course is crab grass or deadspots,” to 5, “less than 5% of course is crab grass ordead spots.” For the clubhouses, 1 was “built orrefurbished more than 15 years ago” and 5 was“built or refurbished 2 years ago or less.” The over-all facility attribute quality variable was the com-posite of these scales. The two component variablesapproached normality and were added together.

Control variables. Customers may be more sat-isfied with larger facilities because they offer moreamenities, and they may be more satisfied withfacilities that employ large percentages of young ortemporary employees, who may be more energetic.Therefore we controlled for facility size, averageemployee age, and percent temporary employees.Customers may be more satisfied if they have beenmembers for a long time and have not quit, if theyare men, or if they are older. Therefore, we alsocontrolled for average customer tenure (months),percent male customers, and average customer age.

Results

Table 5 reports the means, standard deviations,and correlation coefficients for the dependent, in-dependent, and control variables. We used hierar-chical moderated regression models to examine thehypothesized interaction effects, centering all vari-ables involved in the interaction terms to minimizemulticollinearity between the interaction termsand their individual components (Aiken & West,1991). Table 6 presents results. We entered all ofthe control variables in model 1. In model 2 weentered the interactions involving sex and the twodimensions of objective performance. In model 3we entered the interactions involving race and thetwo dimensions of objective performance, and inmodel 4 we entered all four interactions.

Hypothesis 4 states that the association betweenan organizational unit’s objective performance andcustomer satisfaction will be attenuated for organ-izational units that employ higher percentages ofemployees belonging to low-status demographicgroups compared to units that employ higher per-centages of employees belonging to high-status de-mographic groups. The two-way gender by objec-tive performance interactions as a set explained asignificant amount of incremental variance in thedependent variable (�R2 � .04, p � .05) providingsome further support for this hypothesis. Inspec-


tion of the individual regression weights in the fullmodel showed that the facility quality by genderinteraction and the facility productivity by genderinteraction were significant (p � .05). We probedthe pattern of the interactions by examining thesimple slopes of the objective performance mea-sures for facilities with high and low percentages offemale employees (Aiken & West, 1991).

As in the Medcorp study plots, a stronger posi-tive relationship between objective performanceand customer satisfaction emerged for facilities thathad low percentages of female employees than forfacilities that had high percentages of female em-ployees. We do not report the plots owing to spacelimitations, but they are available from the firstauthor. Facility quality was significantly more pos-

TABLE 5Descriptive Statistics and Correlations, Golfcorp Studya

Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10

1. Customer satisfaction withfacility

3.90 0.22

2. Size 129.05 157.51 .163. Average employee age 39.20 4.04 �.18 .094. Percent temporary employees 0.04 0.08 .15 �.07 .165. Average customer tenure 60.66 22.10 �.15 .14 .15 .016. Percent male customers 0.56 0.07 .00 �.03 �.13 .06 �.037. Average customer age 54.23 8.40 �.14 .17 .34 .01 .42 �.168. Percent nonwhite employees 0.26 0.18 .08 �.12 .11 .20 �.28 .15 .089. Percent female employees 0.31 0.12 �.07 �.21 .07 .05 �.07 .20 �.06 �.08

10. Facility quality 3.32 1.04 .08 .16 �.14 .16 �.04 .16 �.03 .10 �.0811. Facility productivity 14,614.26 7,703.06 .13 .14 .12 .16 .12 .13 .16 .14 .09 .15

a n � 66; all correlations greater than .21 are significant at p � .05.

TABLE 6Interactive Influence of Percent Nonwhite Employees, Percent Female Employees, and Objective Facility Performance

on Customer Satisfaction with Facility, Golfcorp Studya

Variables

Customer Satisfaction with Facility

Model 1 Model 2 Model 3 Model 4

ControlsSize .21 .27* .27* .30**Average employee age �.21 �.24* �.24* �.19Percent temporary employees .18 .16 .16 .18Average customer tenure �.12 �.06 �.06 �.06Percent male customers �.06 �.08 �.08 �.11Average customer age �.09 �.13 �.13 �.10Facility quality �.04 �.01 �.01 �.03Facility productivity .17 .01 .01 �.05Percent nonwhite employees �.05 �.15 �.15 �.19Percent female employees .04 �.07 �.07 �.14

InteractionsPercent female � quality �.25* �.23*Percent female � productivity �.22 �.31*Percent nonwhite � quality �.25* �.25*Percent nonwhite �

productivity�.34** �.49**

Adjusted R2 .00 .03 .12 .20R2 .13 .17 .28 .37�R2 from model 1 .04* .15** .24***

a n � 66 country clubs.* p � .05

** p � .01*** p � .001


itively associated with customer satisfaction for fa-cilities with low percentages of female employeesthan for facilities with high percentages of femaleemployees. Likewise, facility productivity wasmore positively associated with customer satisfac-tion for facilities with low percentages of nonwhiteemployees than for facilities with high percentagesof nonwhite employees. Again, even though thesimple slopes were not significantly differentfrom zero, the significant regression coefficient inthe full model demonstrates that they were sig-nificantly different from each other. By looking atthe plots, one can see that the interaction is acrossover, which shows that the direction of therelationship is the opposite, and statistically sig-nificant, for members of high- versus low-statusdemographic groups. These results support Hy-pothesis 4 for both of the gender by objectiveperformance relationships.

The two-way race by objective performance in-teractions as a set explained a significant amount ofincremental variance in the dependent variable(�R2 � .15, p � .05) providing preliminary supportfor our hypothesis. Inspection of the individualregression weights from the full model showed thatthe facility attribute quality by race and facilityproductivity by race interactions were significant(p � .05). Simple slope analysis revealed that theassociation between facility quality and customersatisfaction was significantly more positive for fa-cilities employing low percentages of nonwhitesthan for facilities employing high percentages ofnonwhites. Likewise, the association between facil-ity productivity and customer satisfaction wasmore positive for facilities employing low percent-ages of nonwhites (b � .32, p � .01) than for facil-ities employing high percentages of nonwhites (b ��.23, p � .05). These results support Hypothesis 4for both of the race by objective performancerelationships.

Discussion

In this Golfcorp study, we found that objectivelymeasured behaviors that benefited customers werepositively related to customer satisfaction, but onlyfor facilities with low percentages of nonwhite andfemale employees. These results parallel the resultsof our Medcorp study.

GENERAL DISCUSSION

We set out to determine if and how customersatisfaction judgments are influenced by racial andgender biases. In three samples, we found converg-ing evidence that customer satisfaction judgments

are susceptible to systematic and predictable racialand gender biases. Customers tended to be lesssatisfied with the services provided by women andnonwhite employees rather than by men and whiteemployees, even when objective indicators of per-formance were controlled for (Medcorp study). Wealso found that these biases operated on judgmentsabout the service context when a third-party eval-uator observed an employee interacting with a cus-tomer (Bookcorp study), particularly if the observerhad implicit negative attitudes about women orminorities. Finally, we found that evaluations of anorganizational unit as a whole were negatively as-sociated with the presence of nonwhite and femaleemployees (Golfcorp study). It is worth noting thatwe found evidence for the operation of racial biasesregardless of whether the nonwhite employeeswere predominantly Asian (Medcorp study), Afri-can American (Bookcorp study), or Latino (Golf-corp study).

The consistency of our results across three differ-ent samples and methodologies testifies to the ro-bustness and generality of the systematic biases weobserved and to the internal validity of our theoret-ical model. The pattern of these biases may helpexplain the persistence of demographic inequali-ties in organizations. To cite just a few examples,per 2006 U.S. Census Bureau data women and non-whites make 25 percent less than their male andwhite counterparts in equivalent jobs (http://www.census.gov/hhes/www/income/histinc/p38N.html),women and nonwhites are twice as likely as whitemen to be unemployed and underemployed (Na-tional Institute for Occupational Safety and Health,2002), and women and ethnic minorities are not wellrepresented among the ranks of highly paid managersand professionals in U.S. corporations and in presti-gious occupations like law and medicine (e.g., Baldi& McBrier, 1997; Eagly & Karau, 2002; Wilson,Sakura-Lemessy, & West, 1999). Economists have of-ten been perturbed by these demographic inequalitiesbecause orthodox economic theory would predictthat some of these inequalities should be erased whenemployers compete for women and nonwhite appli-cants, whose wages are 25 percent less costly thantheir white and male counterparts (Economist, 2008).Our results suggest that the evidence available to or-ganizational decision makers is that customers tendto be less satisfied with nonwhite and female employ-ees; however, without the benefit of the research con-ducted here, decision makers would not be able todetermine that these lower satisfaction ratings areattributable at least in part to customer biases. There-fore, decision makers might rationally choose, giventheir limited information, to preferentially selectwhite and male employees, as their data are likely to


indicate that such personnel are better performers.Evidence from highly publicized lawsuits (e.g.,Shoney’s Restaurant; Abercrombie & Fitch) suggeststhat managers are keenly aware of the fact that somecustomers prefer white and male employees. Execu-tives in these cases admitted to deliberately favoringwhite employees in hiring and promotion decisionsto enhance customer satisfaction and organizationalprofitability (Brief et al., 2000).

One could argue that when customers view em-ployees or units as performing less well because ofemployee demographic characteristics such as raceor gender, these employees and units should re-ceive fewer rewards, bonuses, and promotional op-portunities. But it is important to note that thecustomer judgments in our studies were inconsis-tent with other, objective indicators of perfor-mance. In other words, nonwhites and women mayhave behaved the same way as their white and malecounterparts in trying to provide satisfactory cus-tomer service, but if compensation and other organ-izational benefits are linked to customer satisfac-tion ratings, then they may not be rewardedsimilarly for identical behavior, which would vio-late the principle of equity that most business en-terprises claim to follow.

Our results suggest that if customer evaluationsbecome widely and uncritically used to determinepay and promotion opportunities, the job outcomesof women and ethnic minorities could be adverselyimpacted. For example, consider what would hap-pen if managers noticed which employees rou-tinely received the highest customer satisfactionscores and used this information to make promo-tion decisions or if university administrators reliedheavily on student ratings of teaching effectivenessto influence promotion and tenure decisions. Athigher levels of an organization, executives mightexamine which of their organizational unitsachieved the highest levels of customer satisfactionand promote those managers further up the organ-izational hierarchy. Our data suggest that one pos-sible consequence of these decisions is that whitesand men will be much more likely than their non-white or women counterparts to receive favorablecustomer satisfaction judgments, which should ac-celerate their journeys up the organizational lad-der. Likewise, managers who purposely stock theirorganizational units with whites and men are likelyto have more career success than managers whodo not.

Our finding that customer biases can spill overonto the surrounding organizational context con-tributes to the literature on contamination and sig-naling and also illustrates the subtle operation ofracial and gender bias. Marketing researchers have

shown how observable customer characteristicssuch as physical attractiveness can influence othercustomers’ desire to purchase a product (Argo etal., 2009). To our knowledge, ours is the first studyto show how an observable characteristic of anemployee, such as race or gender, can influencecustomer perceptions of an organization’s contex-tual quality. This finding may provide insight intothe phenomenon known as “white flight,” whereinwhites move out of a neighborhood once a criticalmass of nonwhites has moved in (Gladwell, 2000;Kruse, 2005). In an organizational setting, a similarphenomenon may operate in which customers linktheir conscious or unconscious negative attitudestoward members of low-status groups to employeeswho belong to these groups. In turn, these associa-tions “contaminate” customer perceptions of an or-ganizational context. This process of contextualspillover may partly explain why managers haveoften been reluctant to pursue diversity, despite theknown performance advantages of having a diverseworkforce (Joshi, Liao, & Jackson, 2006; O’Reilly,Williams, & Barsade, 1997). It may be that managersare aware that diversity has hidden costs because itincreases the possibility of “customer flight” to anorganization that has fewer employees who belongto low-status demographic groups. Future studiesusing customer satisfaction as an outcome variableshould take into account the demographic make-upof organizations’ employees as well as objectivelymeasured organizational characteristics.

Our findings cast doubt on the ability of custom-ers to accurately perceive the quality of customerservice organizations. The theoretical attributes wesuggested as possible causes of customer bias weremeant to explain why these biases occur, but likeany useful theory they also suggest potential rem-edies. Our theorizing suggests that racial and gen-der biases in customer satisfaction judgments maybe reduced by (1) making customer ratings lessanonymous, (2) changing the standards customersuse to make their ratings so they emphasize behav-ior rather than subjective judgments, and (3) intro-ducing customer debiasing education or traininginto the evaluation process. Identifiability (i.e.,nonanonymity) may provide an especially strongdebiasing effect in service contexts characterizedby repeated interactions (e.g., the doctor-patientrelationship) because customers would want tomake sure their ratings do not jeopardize the qual-ity of future received service. Behaviorally an-chored rating scales may be especially helpful forremoving bias in customer expectations, becauseindividuals often hold nonwhites and women tohigher standards (Biernat & Kobrynowicz, 1997).Organizations could consider only accepting cus-


tomer satisfaction surveys from customers whohave completed debiasing training, although such amove would be logistically and perhaps scientifi-cally problematic (owing to potential selectionbias).

In addition to addressing factors that cause biasin customer ratings, organizations can take practi-cal steps to minimize the potential adverse impactof customer biases on nonwhite and female em-ployees’ careers. For example, organizations mightconsider only using satisfaction surveys from fre-quent customers to ensure that raters have hadsufficient exposure to targets. Organizations couldalso ask for customer feedback during service en-counters so that customers will be most likely to bepaying attention and less likely to rely on informa-tion that is subject to memory bias when judgingtheir experience. Organizations might also want tolet customers know that the data will be used tomake career progression decisions so that they aremore motivated to judge responsibly. Organiza-tions could also insert bias-sensitive questions intocustomer satisfaction judgments so that responsesfrom potentially biased customers could be givenless weight or discarded. However, we urge cautionif organizations choose to remove outlier ratings,because this tactic may actually decrease ratingaccuracy if most judges are aware of their biasesand therefore tend to overcorrect their judgments(Zitzewitz, 2006). Alternatively, organizations maybe able to statistically correct for bias when calcu-lating customer satisfaction judgments. Finally, us-ing different survey formats for customer ratingscales might also be helpful for circumventing raterbiases; possibilities are forced choice, behaviorallyanchored rating scales (citing specific valued be-haviors), and unweighted and weighted checklists.Organizations should consider the trade-offs theseformats entail and choose the one that is mostlikely to reduce the effects of customer judgmentbiases on the career prospects of those who are themost vulnerable targets.

Limitations

We believe our findings provide strong, consis-tent support for our theoretical predictions. How-ever, like all research, ours has its share of limita-tions. First, role congruence may have been anissue in our Medcorp study, as patients may expecttheir physicians to be white and male and thereforejudge nonwhite or women doctors more harshly.However, there is not much evidence to suggestthat they do expect their doctor to be a white male.Indeed, patients prefer their doctor to look likethemselves, in that women prefer women doctors,

and nonwhite patients prefer nonwhite doctors(Chen, Fryer, Phillips, Wilson, & Pathman, 2005;Cooper-Patrick, Gallo, & Gonzales, 1999). Likewise,role congruence should not have been an issue inour Bookcorp study, so the consistent results foundfor these two samples provide us with some confi-dence that role congruence alone is not responsiblefor our findings. Still, future research would bewell served to test our hypotheses in a variety ofsamples as well as to see whether other observabledemographic characteristics, such as age, might in-fluence people’s career progress (see Wang, Adams,Beehr, & Shultz, 2009) as a result of the mecha-nisms we described.

Another potential issue is rater-target congru-ence. Although including patient gender and racein the analysis slightly strengthened our Medcorpstudy results, we did not control for these variablesin that study because of multicollinearity (that is,the respective correlations between customer gen-der and race and physician gender and race weregreater than .90). In our other two studies, we foundno evidence that rater-target congruence influencedcustomer satisfaction judgments. Indeed, in posthoc analyses, the interactions of customer race byemployee race and customer gender by employeegender were not significant. To maintain compli-ance with the rule of thumb that there should befive cases per variable (Tabachnick & Fidell, 2003),we do not report these post hoc analyses.

Another potential limitation is that unobservedvariables may be responsible for our results. In ourGolfcorp study, employee demographic character-istics may have masked the facility’s strategy. Fa-cility executives pursuing a low-cost strategy mayhave hired a large number of women and nonwhiteemployees, whereas those pursuing a premium-pricing strategy may have hired a large number ofwhites and men. We ran some post hoc analyses totest this idea and found no supporting evidence.Specifically, we ran the interactions of several vari-ables measuring club strategy (i.e., services offered,turnover rate) by objective performance, and al-though our four race and gender interactions re-mained significant, none of the additional interac-tions were significant (we do not report theseanalyses to maintain an adequate case-to-variableratio). Relatedly, we also did not test or reportwhether nonwhite women faced a double jeopardy(Berdahl & Moore, 2006) in terms of customer sat-isfaction to maintain compliance with the five-to-one rule of thumb (that is, we were not able to testthe three-way interactions involving employee ob-jective performance, employee race, and employeegender), but future research should do so. Finally,we did not test whether the Implicit Association


Test score influenced ratings of organizationalunits because we found that it influenced subcom-ponents of the organizational unit ratings—that is,the employee and context ratings.

We should also mention that our method of test-ing for possible bias in performance evaluationswas a significant improvement over past studies.First, we used an objective performance standardso that we could compare subjective judgmentswith this standard and therefore determinewhether race and gender might influence the cus-tomer judgments of performance. Second, our sub-jective judgments were based on aggregated judg-ments from a large number of customers rather thanthe judgments of a single supervisor. This is impor-tant because the large number of raters provided ahighly reliable subjective performance rating foreach individual, context, or organization. The IATis new, and its predictive validity is relatively un-tested (Blanton & Jaccard, 2006), so another contri-bution our study offers is the first demonstration ofthe IAT’s predictive validity in a management jour-nal. Finally, we controlled for several variables thatcould provide alternative explanations for our re-sults, such as average employee age and averagecustomer tenure with an organization.

Given these methodological strengths of our re-search, it is unsettling to find that customers may notrespond favorably to organizational characteristicsdesigned to benefit them when these organizationshave a high percentage of low-status employees. AtGolfcorp, employees at clubs with a high percentageof female and nonwhite employees can in fact beeconomically harmed by customer satisfaction eval-uations because clubs that fail to achieve the targetlevel of customer satisfaction (i.e., below the organi-zational average) do not receive a salary bonus. Thepractical implications of our results become moreapparent when we examine the effect sizes in oursamples. Across our three studies, the racial and gen-der bias effects on customer satisfaction judgmentsexplained between 15 and 24 percent of the variancein customer satisfaction. According to Cohen’s (1988)“ballpark” descriptors of effect sizes, a large effectsize is associated with an R2 of .25; a medium one,with an R2 of .09; and a small one, with an R2 of .01.Therefore, the average effect size of the racial andgender biases observed in our three samples is be-tween medium and large.

Conclusions

In these different samples, we demonstrated thatcustomer ratings are biased against women and ra-cial minorities. We conducted two field studies andone laboratory study, thus utilizing a full cycle

research strategy. The effects were demonstratedfor three different minority groups and three differ-ent contexts involving employee-customer contact.In all three settings, we controlled for actual objec-tive behavior or performance along with a series ofother controls appropriate for that context. The ef-fects are demonstrated for individual targets as wellas for the contexts or organizations in which thetargets worked. In the laboratory sample, weshowed that implicit racial or gender bias exacer-bated the biased ratings. In short, these are fairlyrobust findings across jobs, contexts, raters,and ratees.

If these results are replicated and generalizable,they have significant implications for organization-al practice. If managers are serious about the fairtreatment of their employees and the promotion ofdiversity, they need to treat customer ratings differ-ently. More specifically, the rating process can bechanged by increasing information, responsibility,or training for raters and by changing how customerratings are used. In the latter context, organizationscan perhaps measure and discount such biases orstatistically adjust the ratings to remove the bias.Without such actions, given the increasing depen-dence on customer ratings, society is not onlylikely to maintain existing levels of inequitablecompensation and advancement for women andminorities, but also likely to increase these inequi-ties. This outcome is unacceptable in a society thatis committed legally, morally, and socially to fairtreatment for all in the workplace.

REFERENCES

Aiken, L. S., & West, S. G. 1991. Multiple regression:Testing and interpreting interactions. NewburyPark, CA: Sage.

Anderson, E. W., Fornell, C., & Mazvancheryl, S. K. 2004.Customer satisfaction and shareholder value. Jour-nal of Marketing, 68(4): 172–185.

Anderson, E. W., & Mittal, V. 2000. Strengthening thesatisfaction-profit chain. Journal of Service Re-search, 3: 107–120.

Anderson, N. H. 1981. Foundations of information inte-gration theory. New York. Academic Press.

Argo, J. J., Dahl, D. W., & Morales, A. C. 2009. Positiveconsumer contamination: Responses to attractiveothers in a retail context. Journal of Marketing Re-search, 45: 690–701.

Ashburn-Nardo, L., Knowles, M. L., & Monteith, M. J.2003. Black Americans’ implicit racial associationsand their implications for intergroup judgment. So-cial Cognition, 21: 61–87.

Aquino, K., & Bommer, W. H. 2003. Preferential mistreat-


ment: How victim status moderates the relationshipbetween organizational citizenship behavior andworkplace victimization. Organization Science, 4:274–285.

Baldi, S., & McBrier, D. B. 1997. Do the determinants ofpromotion differ for blacks and whites? Evidencefrom the U.S. labor market. Work and Occupations,24: 478–497.

Baltes, B. B., & Parker, C. P. 2000. Reducing the effects ofperformance expectations on behavioral ratings. Or-ganizational Behavior and Human Decision Pro-cesses, 82: 237–267.

Berdahl, J. L., & Moore, C. 2006. Workplace harassment:Double-jeopardy for minority women. Journal ofApplied Psychology, 91: 426–436.

Berger, J., Conner, T. L., & Fisek, M. H. 1974. Expectationstates theory: A theoretical research program.Cambridge, MA: Winthrop.

Berger, J., Fisek, M. H., Norman, R. Z., & Zelditch, M., Jr.1977. Status characteristics and social interaction:An expectation states approach. New York:Elsevier.

Bertrand, M., & Mullainathan, S. 2004. Are Emily andGreg more employable than Lakisha and Jamal? Afield experiment on labor market discrimination.American Economic Review, 94: 991–1013.

Biernat, M., & Kobrynowicz, D. 1997. Gender and race-based standards of competence: Lower minimumstandards but higher ability standards for devaluedgroups. Journal of Personality and Social Psychol-ogy, 72: 544–557.

Blanton, H., & Jaccard, J. 2006. Arbitrary metrics inpsychology. American Psychologist, 61: 27–41.

Bobo, L. 1998. Race, interests, and beliefs about affirma-tive action—Unanswered questions and new direc-tions. American Behavioral Scientist, 41: 985–1003.

Bracken, D. W., Church, A. H., & Timmreck, C. W. 2001.Handbook of multisource feedback: The compre-hensive resource for designing and implementingMSF processes. New York: Jossey-Bass.

Brief, A. P., Dietz, J., Cohen, R. R., Pugh, S. D., & Vaslow,J. B. 2000. Just doing business: Modern racism andobedience to authority as explanations for employ-ment discrimination. Organizational Behavior andHuman Decision Processes, 81: 72–97.

Chatman, J. A., & Flynn, F. J. 2005. Full-cycle micro-organizational behavior research. Organization Sci-ence, 16: 434–447.

Chen, F. M., Fryer, G. E., Jr., Phillips, R. L., Jr., Wilson, E.,& Pathman, D. E. 2005. patients’ beliefs about racism,preferences for physician race, and satisfaction withcare. Annals of Family Medicine, 3: 138–143.

Cialdini, R. B. 1995. A full-cycle approach to social psy-chology. In G. G. Brannigan & M. R. Merrens (Eds.),

The social psychologists: Research adventures:53–72. New York: McGraw-Hill.

Cohen, J. 1988. Statistical power analysis for the be-havioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.

Cooper-Patrick, L., Gallo, J. J., & Gonzales, J. J. 1999. Race,gender, and partnership in the patient-physician re-lationship. Journal of the American Medical Asso-ciation, 282: 583–589.

Crandall, C. S., & Eshelman, A. 2003. A justification-suppression model of the expression and experienceof prejudice. Psychological Bulletin, 129: 414–446.

Davis, W. E., & Davis, D. R. 1999. The university presi-dency: Do evaluations make a difference? Journal ofPersonnel Evaluation in Education, 13: 119–140.

DeNisi, A. S., Robbins, T., & Cafferty, T.P. 1989. Organi-zation of information used for performance apprais-als: Role of diary keeping. Journal of Applied Psy-chology, 74: 124–129.

Dovidio, J. F., & Gaertner, S. L. 1981. The effects of race,status, and ability on helping-behavior. Social Psy-chology Quarterly, 44: 192–203.

Dovidio, J. F., & Gaertner, S. L. 2000. Aversive racism andselection decisions: 1989 and 1999. PsychologicalScience, 11: 315–319.

Dubois, R. W., Alexander, C. M., Wade, S., Mosso, A.,Markson, L., Lu, J. D., Nag, S., & Berger, M. L. 2002.Growth in use of lipid lowering therapies: Are wetargeting the right patients? American Journal ofManaged Care, 8: 81–86.

Eagly, A. H., & Karau, S. J. 2002. Role congruity theory ofprejudice toward female leaders. Psychological Re-view, 109: 573–598.

Eagly, A. H., Makhijani, M. G., & Klonsky, B. G. 1992.Gender and the evaluation of leaders: A meta-analy-sis. Psychological Bulletin, 111: 3–22.

Economist. 2008. Black America: Nearer to overcoming.May 8: 34–36.

Feinglass, J., & Salmon, J. W. 1990. Corporatization ofmedicine: The use of medical management informa-tion systems to increase the clinical productivity ofphysicians. Journal of Health Services, 20: 233–252.

Fernandez, J. P. 1981. Racism and sexism in corporatelife. Lexington, MA: Heath.

Figueroa, E. B., & Woods, R. A. 2007. Industry output andemployment projections to 2016. Monthly Labor Re-view, 130(November): 53–85.

Folkes, V. S., & Patrick, V. M. 2003. The positivity effectin perceptions of services: Seen one, seen them all?Journal of Consumer Research, 30: 125–137.

Gaertner, S. L., & Dovidio, J. F. 1986. The aversive form ofracism. In J. F. Dovidio & S. L. Gaertner (Eds.), Prej-udice, discrimination, and racism: 61–89. Orlando,FL: Academic Press.


Gerstein, H. C., Yusuf, S., Mann, J. F. E., Hoogwerf, B.,Zinman, B., Held, C., Fisher, M., Wolffenbuttel, B.,Bosch, J., Richardson, L., Pogue, J., & Halle, J. P.2000. Effects of ramipril on cardiovascular and mi-crovascular outcomes in people with diabetes melli-tus: Results of the HOPE study and MICRO-HOPEsubstudy. Lancet, 355: 205–212.

Giannopoulos, G., Conway, M., & Mendelson, M. 2005.The gender of status: The laypersons’ perception ofstatus is gender-typed. Sex Roles, 53: 795–806.

Gilovich, T., Griffin, D. W., & Kahneman, D. 2002. Heu-ristics of judgment: The psychology of intuitivejudgment. Cambridge, U.K.: Cambridge UniversityPress.

Gladwell, M. 2000. The tipping point: How little thingscan make a big difference. Little, Brown.

Goldin, C., & Rouse, C. 2000. Orchestrating impartiality:The impact of “blind” auditions on female musi-cians. American Economic Review, 90: 715–741.

Greenhaus, J. H., Parasuraman, S., & Wormley, W. M.1990. Effects of race on organizational experiences,job performance evaluations, and career outcomes.Academy of Management Journal, 33: 64–86.

Greenwald, A. G., & Banaji, M. R. 1995. Implicit socialcognition: Attitudes, self-esteem, and stereotypes.Psychological Review, 102: 4–27.

Greenwald, A. G., Nosek, B. A., & Banaji, M. R. 2003.Understanding and using the implicit associationtest: I. An improved scoring algorithm. Journal ofPersonality and Social Psychology, 85: 197–216.

Greenwald, A. G., Poehlman, T. A., Uhlmann, E., & Ba-naji, M. R. 2009. Understanding and using the Im-plicit Association Test: III. Meta-analysis of predic-tive validity. Journal of Personality and SocialPsychology: 97: 17–41.

Gruca, T. S., & Rego, L. L. 2005. Customer satisfaction,cash flow, and shareholder value. Journal of Mar-keting, 69(3): 115–130.

Gupta, S., & Zeithaml, V., 2006. Customer metrics andtheir impact on financial performance. MarketingScience, 25: 718–739.

Haas, J. S., Cook, E. F., Puopolo, A. L., Burstin, H. R.,Cleary, P. D., Brennan, T. A. 2000. Is the professionalsatisfaction of general internists associated with pa-tient satisfaction? Journal of General Internal Med-icine, 15: 122–128.

Hagan, C. M., Konopaske, R., Bernardin, H. J., & Tyler,C. L. 2006. Predicting assessment center perfor-mance with 360-degree, top-down, and customer-based competency assessments Human ResourceManagement, 45: 357–390.

Hsaio, W. C., Braun, P., Becker, E. R., & Thomas, S. R.1988a. The resource-based relative value system.Journal of the American Medical Association, 258:799–802.

Hsaio, W. C., Braun, R., Dunn, D., & Becker, E. R. 1988b.

Resource-based relative values: An overview. Jour-nal of the American Medical Association, 260:2347–2353.

Isles, C. G. 2002. Patients with acute coronary syndromeshould start a statin while still in hospital. Heart, 88:5–6.

Ittner, C. D., & Larcker, D. F. 1998. Are non-financialmeasures leading indicators of financial perfor-mance? An analysis of customer satisfaction. Jour-nal of Accounting Research, 36: 1–35.

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

Judge, T. A., & Ferris, G. R. 1993. Social-context of per-formance evaluation decisions. Academy of Man-agement Journal, 36: 80–105.

Kane, J. S., Bernardin, H. J., Villanova, P., & Peyrefitte, J.1995. Stability of rater leniency: Three studies.Academy of Management Journal, 38: 1036–1051.

Keller, K. L. 2003. Strategic brand management. UpperSaddle River, NJ: Prentice Hall.

Kleiner, K. D., Akers, R., Burke, B. L., & Werner, E. J.2002. Parent and physician attitudes regarding elec-tronic communication in pediatric practices. Pedi-atrics, 109: 740–744.

Kruse, K. M. 2005. White flight: Atlanta and the makingof modern conservatism. Princeton, NJ: PrincetonUniversity Press.

Laine, C., & Davidoff, F. 1996. Patient-centered medicine:A professional evolution. Journal of the AmericanMedical Association, 275: 152–156.

Landy, F. J., Shankster, L. J., & Kohler, S. S. 1994. Per-sonnel-selection and placement. In M. R. Rosenz-weig & L. W. Porter (Eds.), Annual review of psy-chology, vol. 45: 261–296. Palo Alto, CA: AnnualReviews.

LaRosa, J. C., He, J., & Vupputuri, S. 1999. Effect of statinson risk of coronary disease: A meta-analysis of ran-domized controlled trials. Journal of the AmericanMedical Association, 282: 2340–2346.

Latham, G. P., & Wexley, K. N. 1977. Behavioral obser-vation scales for performance-appraisal purposes.Personnel Psychology, 30: 255–268.

Lerner, J. S., & Tetlock, P. E. 1999. Accounting for theeffects of accountability. Psychological Bulletin,125: 255–275.

McConahay, J. B. 1983. Modern racism and modern dis-crimination—The effects of race, racial-attitudes,and context on simulated hiring decisions. Person-ality and Social Psychology Bulletin, 9: 551–558.

McConahay, J. B. 1986. Modern racism, ambivalence,and the modern racism scale. In J. F. Dovidio & S. L.Gaertner (Eds.), Prejudice, discrimination, and rac-ism: 91–126. New York: Academic.


McConnell, A. R., & Leibold, J. M. 2001. Relations amongthe Implicit Association Test, discriminatory behav-ior, and explicit measures of racial attitudes. Journalof Experimental Social Psychology, 37: 435–442.

Mercer Consulting Group. 2007.

Morales, A. C., & Fitzsimons, G. J. 2007. Product conta-gion: Changing consumer evaluations through phys-ical contact with “disgusting” products. Journal ofMarketing Research, 44: 272–283.

Moshavi, D. 2004. He said, she said: Gender bias andcustomer satisfaction with phone-based service en-counters. Journal of Applied Social Psychology, 34,162–176.

Murphy, K. R. 1991. Criterion issues in performanceappraisal research: Behavioral accuracy versus clas-sification accuracy. Organizational Behavior andHuman Decision Processes, 50: 45–50.

National Institute for Occupational Safety and Health.2002. The changing organization of work and thesafety and health of working people (NIOSH pub-lication no. 2002–116). Washington, DC: NIOSH.

Nosek, B. 2005. Moderators of the relationship betweenimplicit and explicit evaluation. Journal of Experi-mental Psychology: General, 134: 565–584.

O’Leary, V. E., & Ickovics, J. R. 1992. Cracking the glassceiling: Overcoming isolation and alienation. In U.Sekaran & F. Leong (Eds.), Woman power: Manag-ing in times of demographic turbulence: 7–30.Thousand Oaks, CA: Sage.

O’Reilly, C. A., Williams, K. Y., & Barsade, S. G. 1998.Group demography and innovation: Does diversityhelp? In D. Gruenfeld, B. Mannix, & M. Neale (Eds.),Research on managing groups and teams: 183–207. Stamford, CT: JAI Press.

Pellegrin, K. L., Stuart, G. W., Maree, B., Frueh, B. C., &Ballenger, J. C. 2001. A brief scale for assessing pa-tients’ satisfaction with care in outpatient psychiat-ric services. Psychiatric Services, 52: 816–819.

Pulakos, E. D., White, L. A., Oppler, S. H., & Borman,W. C. 1989. Examination of race and sex effects onperformance ratings. Journal of Applied Psychol-ogy, 74: 770–780.

Richman, W. L., Kiesler, S., Weisband, S., & Drasgow, F.1999. A meta-analytic study of social desirabilitydistortion in computer-administered questionnaires,traditional questionnaires, and interviews. Journalof Applied Psychology, 84: 754–775.

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

Roch, S. G., & O’Sullivan, B. J. 2003. Frame of referencerater training issues: Recall, time and behavior ob-servation training. International Journal of Train-ing and Development, 7: 93–107.

Rotundo, M., & Sackett, P. S. 1999. Effect of rater race onconclusions regarding differential prediction in cog-

nitive ability tests. Journal of Applied Psychology,84: 815–822.

Rozin, P., Millman, T., & Nemeroff, C. 1986. Operation ofthe laws of sympathetic magic in disgust and otherdomains. Journal of Personality and Social Psy-chology, 50: 703–712.

Rynes, S. L., Bretz, R. D., & Gerhart, B. 1991. The impor-tance of recruitment in job choice—A different wayof looking. Personnel Psychology, 44: 487–521.

Rynes, S. L., Heneman, H. G., III, & Schwab, D. P. 1980.Individual reactions to organizational recruiting: Areview. Personnel Psychology, 33: 529–542.

Rynes, S. L., & Miller, H. E. 1983. Recruiter and jobinfluences on candidates for employment. Journalof Applied Psychology, 68: 147–154.

Salam, S., Cox, J. F., & Sims, H. P., Jr. 1997. In the eye ofthe beholder: How leadership relates to 360-degreeperformance ratings. Group and Organization Man-agement, 2: 185–209.

Schneider, B., Ehrhart, M. G., Mayer, D. M., Saltz, J. L., &Niles-Jolly, K. 2005. Understanding organiza-tion-customer links in service settings. Academy ofManagement Journal, 48: 1017–1032.

Sidanius, J., & Pratto, F. 1999. Social dominance: Anintergroup theory of social hierarchy and oppres-sion. Cambridge, U. K.: Cambridge University Press.

Simonet, D. 2005. Patient satisfaction under managedcare. International Journal of Health Care QualityAssurance, 18: 424–440.

Simonin, B. L., & Ruth, J. A. 1998. Is a company knownby the company it keeps? Assessing the spillovereffects of brand alliances on consumer brand atti-tudes. Journal of Marketing Research, 35: 30–42.

Sixma, H. J., Spreeuwenberg, P. M., & van der Pasch,M. A. 1998. Patient satisfaction with the generalpractitioner: A two-level analysis. Medical Care, 36:212–229.

Song, M. 2004. Who’s at the bottom? Examining claimsabout racial hierarchy. Ethnic and Racial Studies,27: 859–877.

Spence, M. 1973. Job market signalling. Quarterly Jour-nal of Economics, 85: 355–374.

Stewart, M., Brown, J. B., Donner, A., McWhinney, I. R.,Oates, J., Weston, W. W., & Jordan, J. 2000. Theimpact of patient-centered care on outcomes. Jour-nal of Family Practice, 49: 796–804.

Tabachnick, B. G., & Fidell, L. S. 2003. Using multivar-iate statistics (5th ed.). Boston: Allyn & Bacon.

Taylor, H. 2002. Patient/physician online communication:Many patients want it, would pay for it, and it wouldinfluence their choice of doctors and health plans.Harris Interactive Health Care News, 2(8): 1–3.

Wang, M., Adams, G. A., Beehr, T. A., & Shultz, K. S.2009. Career issues at the end of one’s career: Bridgeemployment and retirement. In S. G. Baugh & S. E.


Sullivan (Eds.), Maintaining focus, energy, and op-tions through the life span: 135–162. Charlotte, NC:Information Age Publishing.

Westphal, J. D., & Stern, I. 2007. Flattery will get youeverywhere (especially if you are a male caucasian):Ingratiation, boardroom behavior, and demographicminority status affect the likelihood of gaining addi-tional board appointments at U.S. companies. Acad-emy Management Journal, 50: 1–22.

Wherry, R. J., & Bartlett, C. J. 1982. The control of bias inratings: A theory of rating. Personnel Psychology,35: 521–551.

Wilkinson, T. J., & Fontaine, S. 2002. Patients’ globalratings of student competence. Unreliable contami-nation or gold standard? Medical Education, 36:1117–1121.

Willerson, J. T., & Cohn, J. N. (Eds.). 2000. Cardiovascu-lar medicine. Philadelphia: Churchill Livingstone.

Wilson, G., Sakura-Lemessy, I., & West, J. P. 1999. Reach-ing the top: Racial differences in mobility paths toupper-tier occupations. Work and Occupations, 26:165–186.

Woehr, D., & Roch, S. G. 1996. Context effects in perfor-mance evaluation: The impact of ratee gender andperformance level. Organizational Behavior andHuman Decision Processes, 66: 31–41.

Wyer, R. S., Jr., & Schrull, T. K. 1989. Person memory andjudgment. Psychological Review, 96: 58–83.

Yarkin, K. L., Town, J. P., & Wallston, B. S. 1982. Blacksand women must try harder: Stimulus person’s raceand sex attributions of causality. Personality andSocial Psychology Bulletin, 8: 21–24.

Yusuf, S., Sleight, P., Pogue, J., Bosch, J., Davies, R., &Dagenais, G. 2000. Effects of an angiotensin-convert-ing-enzyme inhibitor, ramipril, on cardiovascularevents in high-risk patients. New England Journalof Medicine, 342: 145–153.

Zitzewitz, E. 2006. Nationalism in winter sports judgingand its lessons for organizational decision making.Journal of Economics & Management Strategy, 15:67–99

APPENDIX

Customer Satisfaction Items

Satisfaction with Employee

Medcorp“How would you rate the following attributes of your

provider?” (1 � “very poor”; 5 � “excellent”):

1. Attention provider paid2. Thoroughness and competence of provider3. Ability to ask questions of this provider

Bookcorp“How would you rate the following?” (1 � “very

poor”; 7 � “excellent”):

1. Speed of service2. Quality of service3. Availability of staff for assistance4. Employee responsiveness to customers’ issues and

concerns


Bookcorp“How would you rate the following aspects of the

bookstore?”

1. Appearance of bookstore (1 � “very poor”; 7 �“excellent”)

2. Environment of the bookstore was conducive to learn-ing/reading (1 � “strongly disagree”; 7 � “stronglyagree”)

3. The bookstore has up-to-date equipment4. This bookstore’s physical facilities are visually

appealing5. The appearance of this bookstore is in keeping with

the type of services provided6. Bookstore meets expectations (1 � “less than ex-

pected,” 7 � “better than expected”)7. Likelihood of recommending bookstore to others (1 �

“definitely would not,” 7 � “definitely would”)

Golfcorp (Satisfaction with Facility)“How would you rate the following aspects of your

club?” (1 � “very poor”; 5 � “very good”):

1. Maintenance of grounds/appearance of clubhouse2. Locker rooms and rest rooms3. Quality of greens4. Condition of course5. Pace of play6. Condition of practice facilities7. Ability to obtain desired tee times8. Club meets expectations (1 � “less than expected”;

5 � “better than expected”)9. Likelihood of recommending club to others (1 � “def-

initely will not”; 5 � “definitely will”)


David R. Hekman ([email protected]) is an assistantprofessor of management at the Lubar School of Busi-ness, University of Wisconsin–Milwaukee. He earned hisPh.D. in management from the University of Washington.He is interested in improving organizational health byminimizing organizational problems such as weak em-ployee attachment, contagious harmful behaviors, andpersistent workplace inequality.

Karl Aquino ([email protected]) is the RichardPoon Professor of Organizations and Society at theSauder School of Business, University of British Colum-bia. His research interests are moral behavior, workplacevictimization, revenge and forgiveness, and status andpower dynamics in organizations. He received his Ph.D.in organizational behavior from Northwestern Univer-sity.

Bradley P. Owens ([email protected]) is a post-doctoral fellow at the Center for Positive OrganizationalScholarship at the University of Michigan. He earned hisPh.D. in management from the University of Washington.His primary research interests include humility, leader-ship, race and gender issues, and team dynamics.

Terence R. Mitchell ([email protected]) receivedhis Ph.D. from the University of Illinois in organizationalpsychology. He is the Carlson Professor of Managementat the University of Washington. Research interests in-clude motivation, leadership, turnover, and decisionmaking.

Pauline Schilpzand ([email protected]) isan assistant professor of character development at theArmy Center of Excellence for the Professional MilitaryEthic at West Point. She earned her Ph.D. in managementfrom the Warrington College of Business Administrationat the University of Florida. Her primary research inter-ests include behavioral ethics, leadership, and the selec-tion process.

Keith Leavitt ([email protected]) is an assistantprofessor of character development at the Army Center ofExcellence for the Professional Military Ethic at WestPoint. He earned his Ph.D. in management from the Fos-ter School of Business at the University of Washington.His primary research interests include behavioral ethics,implicit social cognition, research methods, and organi-zational citizenship.


Copyright of Academy of Management Journal is the property of Academy of Management and its content may

not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

permission. However, users may print, download, or email articles for individual use.

Documents

AN EXAMINATION OF WHETHER AND HOW RACIAL …leeds-faculty.colorado.edu/dahe7472/Hekman AMJ Best Paper Finalist...GENDER BIASES INFLUENCE CUSTOMER SATISFACTION ... We examined whether