The Generalizability of Gender Bias: Testing the Effects of Contextual, Explicit, and Implicit Sexism on Labor Arbitration Decisions

Law and Human Behavior

The Generalizability of Gender Bias: Testing the Effectsof Contextual, Explicit, and Implicit Sexism on LaborArbitration DecisionsErik J. Girvan, Grace Deason, and Eugene BorgidaOnline First Publication, June 1, 2015. http://dx.doi.org/10.1037/lhb0000139

CITATIONGirvan, E. J., Deason, G., & Borgida, E. (2015, June 1). The Generalizability of Gender Bias:Testing the Effects of Contextual, Explicit, and Implicit Sexism on Labor ArbitrationDecisions. Law and Human Behavior. Advance online publication.http://dx.doi.org/10.1037/lhb0000139

The Generalizability of Gender Bias: Testing the Effects of Contextual,Explicit, and Implicit Sexism on Labor Arbitration Decisions

Erik J. GirvanUniversity of Oregon

Grace DeasonUniversity of Wisconsin–La Crosse

Eugene BorgidaUniversity of Minnesota

Decades of social-psychological research show that gender bias can result from features of the socialcontext and from individual-level psychological predispositions. Do these sources of bias impact legaldecisions, which are frequently made by people subject to factors that have been proposed to reduce bias(training and accountability)? To answer the question, we examined the potential for 3 major social-psychological theories of gender bias (role-congruity theory, ambivalent sexism, and implicit bias) topredict outcomes of labor arbitration decisions. In the first study, undergraduate students and professionalarbitrators made decisions about 2 mock arbitration cases in which the gender of the employee-grievantswas experimentally manipulated. Student participants’ decisions showed the predicted gender bias,whereas the decisions of experienced professionals did not. Individual-level attitudes did not predict theextent of the observed bias and accountability did not attenuate it. In the second study, arbitrators’explicit and implicit gender attitudes were significant predictors of their decisions in published cases. Thelaboratory and field results suggest that context, expertise, and implicit and explicit attitudes are relevantto legal decision-making, but that laboratory experiments alone may not fully capture the nature of theireffect on legal professionals’ decisions in real cases.

Keywords: accountability, ambivalent sexism, arbitration, expertise, implicit bias

Supplemental materials: http://dx.doi.org/10.1037/lhb0000139.supp

A little learning is a dangerous thing. Drink deep, or taste not thePierian Spring; There shallow drafts intoxicate the brain, and drinkinglargely sobers us again.

—Alexander Pope (1711), An Essay on Criticism

A central concern at the intersection of psychology and law isthe extent to which research in the former applies in contexts that

matter to the latter (Bornstein, 1999; Mitchell, 2012). This isespecially true for sensitive, politicized topics like sexism andracism, in which the temptation to act on the untested implicationsof research may be great. For example, recently, insights fromdecades of social–cognitive research on implicit attitudes andstereotypes—known generally as “implicit bias”—have capturedthe legal imagination with transformative consequences (Borgida& Girvan, 2015; Greenwald & Krieger, 2006; Levinson & Smith,2012). Notwithstanding the popularity of implicit bias as a catalystof calls for legal doctrinal or procedural change, there is almost noresearch examining whether such theories generalize to the deci-sions of legal professionals. Evidence from laboratory studies inother professional contexts suggests that implicit biases do affectdecisions (see, e.g., Green et al., 2007; Sabin & Greenwald, 2012),and measures of implicit bias have predicted discriminatory out-comes in a variety of workplace behaviors, including hiring (Ag-erstrom & Rooth, 2011; Rooth, 2010), education (van den Bergh,Denessen, Hornstra, Voeten, & Holland, 2010), voting (Payne etal., 2010), and use of force by police during arrests (Goff, Jackson,Di Leone, Culotta, & DiTomasso, 2014). However, in experimen-tal studies involving legal professionals, these individuals fre-quently exhibit implicit bias on standard measures but neverthelessmake unbiased decisions (Levinson & Young, 2010; Rachlinski,Johnson, Wistrich, & Guthrie, 2008). Moving outside of controlledlaboratory experiments, there is no prior research of which we areaware that tests the impact of implicit attitudes or stereotypes ondecisions made by legal professionals in actual cases.

Erik J. Girvan, School of Law, University of Oregon; Grace Deason,Department of Psychology, University of Wisconsin–La Crosse; EugeneBorgida, Department of Psychology, University of Minnesota.

We thank National Academy of Arbitrators Research and EducationFoundation, the American Psychology – Law Society, and The Center forthe Study of Political Psychology at the University of Minnesota for theirfinancial support of this research. None of these organizations, however,bears any responsibility for the scientific work and/or interpretations re-ported herein. The work could not have been accomplished without thediligent help of Ela Gabrys, Karina Kuhrt, Matthew Postma, Joseph DalleMolle, John Alden, and Autumn Chen. Special thanks to the participants inthe 2013 Implicit Bias, Philosophy, and Psychology Conference at theUniversity of Sheffield for their valuable comments on this research and toThe Leverhulme (Grant number IN-058) for making the conference pos-sible.

Correspondence concerning this article should be addressed to Erik J.Girvan, 1221 University of Oregon, Eugene, OR 97403-1221. E-mail:[email protected]

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Law and Human Behavior © 2015 American Psychological Association2015, Vol. 39, No. 4, 000 0147-7307/15/$12.00 http://dx.doi.org/10.1037/lhb0000139

1

Our primary goal is to begin to fill this gap by examiningwhether and how the psychological processes that underlie genderdiscrimination affect the decisions of legal professionals. We pres-ent the results of an experimental laboratory study and a correla-tional field study, each examining the implications of three prom-inent theories of sexism (i.e., role congruity theory, ambivalentsexism, implicit gender attitudes) in the context of labor arbitrationdecisions. In addition, we test the effects of two features of thosesettings proposed to attenuate bias: Training and accountability.Consistent with prior work, the laboratory and field results suggestthat implicit and explicit attitudes are relevant to legal decision-making, but that laboratory experiments may not fully capture thenature of their effect on legal professionals’ decisions in real cases.

Social-Psychological Perspectives on Gender Bias

Decisions are potentially impacted by gender bias in threedifferent ways: Incongruity between general social stereotypesconnecting certain occupations with assumptions about the genderof individuals who work there, explicit or consciously held atti-tudes about gender, and implicit gender bias (Heilman, 2012).

Role Congruity Theory

The magnitude of promale gender bias appears to be rathermodest overall, but becomes pronounced when the gender of anindividual is inconsistent with gendered stereotypes associatedwith the setting (Eagly & Karau, 2002; Wood & Eagly, 2010).Lyness and Heilman (2006), for example, found that for upper-middle and senior-level managers in stereotypically masculine linepositions (e.g., business management and operations management)performance evaluations of men were significantly higher thanthose of women, whereas for managers in stereotypically femininestaff positions (e.g., human resources and administration), differ-ences in performance evaluations for women and men were notsignificantly different (for meta-analyses see Bowen, Swim, &Jacobs, 2000; Swim, Borgida, Maruyama, & Myers, 1989). Ex-perimental and field research into the effects of gender stereotypeson other employment decisions, such as hiring, promotion, andwork assignments, reach a similar conclusion (Davison & Burke,2000; Olian, Schwab, & Haberfeld, 1988; Tosi & Einbender,1985). Thus, at baseline, in stereotypically masculine jobs butperhaps not elsewhere, women tend to be treated more negativelythan men.

Explicit Gender Attitudes

Individual-level psychological predispositions also impact theextent of gender bias. Ambivalent Sexism theory (Glick & Fiske,1996) identifies two main dimensions of attitudes toward women.Hostile sexism involves “sexist antipathy toward women based onan ideology of male dominance, male superiority, and a hostileform of sexuality” (Fiske & Glick, 1995, p. 98). In contrast,benevolent sexism involves “a set of interrelated attitudes towardwomen that are sexist in terms of viewing women stereotypicallyand in restricted roles but that are subjectively positive in feelingtone” (Glick & Fiske, 1996, p. 491; Glick, Diebold, Bailey-Werner, & Zhu, 1997). Hostile and benevolent sexism are knownto guide individual reactions to women in the workplace but in

different ways (Cikara & Fiske, 2009). Levels of hostile sexism,but not benevolent sexism, have been found to predict morenegative evaluations of a female candidate for a masculine occu-pational role (Masser & Abrams, 2004; also see Sakalli-Ugurlu &Beydogan, 2002). Levels of benevolent sexism, in contrast, predictprejudice toward women who violate their traditional spousal roleby engaging in infidelity (Viki & Abrams, 2002), are associatedwith a greater tendency toward victim-blaming in cases of ac-quaintance rape (Abrams, Viki, Masser, & Bohner, 2003; Viki,Massey, & Masser, 2005), and predict paternalistic discriminationin the workplace, including overhelping, taking over, and limitingthe responsibilities of women (Cikara, Lee, Fiske, & Glick, 2009;Glick & Fiske, 2007).

Implicit Gender Attitudes

Implicit attitudes, distinct from consciously endorsed beliefs andevaluations, reflect automatic associations between attitude objectsand categorical attributes (i.e., stereotypes), automatic evaluationsof the attitude object (i.e., attitudes), or both (Amodio & Devine,2006; Blair, Dasgupta, & Glaser, 2015; Rudman & Ashmore,2007). Reaction time measures of implicit attitudes often predictjudgment and behavior better than measures of explicit attitudes,particularly in socially sensitive domains like those involvingintergroup relations (Fazio & Olson, 2003; Greenwald, Poehlman,Uhlmann, & Banaji, 2009; for a critique see Oswald, Mitchell,Blanton, Jaccard, & Tetlock, 2013). Research on implicit genderattitudes in professional settings is relatively sparse, but resultssuggest that such attitudes operate in the workplace. For example,although implicit attitudes toward women tend to be positive(Rudman & Goodwin, 2004), men who anticipated a meeting witha female supervisor revealed more negative implicit attitudes to-ward women, in general, relative to men who expected a meetingwith a subordinate or equal-status female coworker (Richeson &Ambady, 2001).

Possible Moderators of Gender Bias in Legal Decisions

Critics of the relevance of laboratory research to understandingdecisions in professional settings assert that common features ofreal-world situations may reduce or eliminate the effects of suchbias (e.g., Landy, 2008; Mitchell, 2012). Tetlock and Mitchell(2009), for example, argue that “there are daunting obstacles togeneralizing from stylized, low-stakes lab experiments to complex,high-stakes workplaces that have often erected institutional barri-ers—training and accountability procedures—against unlawfuldiscrimination” (p. 16). However, there is little research testingwhether accountability and training do eliminate bias, particularlyin legal decisions.

Accountability

Accountability is “the implicit or explicit expectation that onemay be called on to justify one’ beliefs, feelings, and actions toothers” (Lerner & Tetlock, 1999, p. 255). This expectation mayappear in various forms (Lerner & Tetlock, 1999; Tetlock &Lerner, 1999), but the type of accountability most likely to checkbias is one that induces integrative complexity, an intensive styleof thought characterized by recognition of the multiple facets of a

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

2 GIRVAN, DEASON, AND BORGIDA

situation and preemptive self-criticism. Integratively complexthinking is most likely to occur when individuals know before theymake a decision that they will be accountable to a knowledgeable,legitimate authority for the process by which they make a decision,but do not know the preferences of that authority other than that itis interested in fairness (Lerner & Tetlock, 1999). This thinkingstyle minimizes the use of heuristics and other cognitive shortcuts,and causes participants to incorporate more facts into the decision-making process (Lerner & Tetlock, 1999; Tetlock, Skitka, &Boettger, 1989). In this fashion, accountability may work to pre-vent or attenuate the use of gender stereotypes.

A small number of experiments examine the impact of account-ability on the stereotyping process. In support of the notion thataccountability eliminates bias, Ford, Gambino, Lee, Mayo, andFerguson (2004) found that food industry managers made raciallybiased evaluations of job applicants when they were told to basetheir evaluations on their gut feeling, but not when they wereinstructed to use careful deliberation and to discuss the decisionwith a supervisor. Unfortunately, the study design confoundedaffective versus cognitive decision-making with the absence orpresence of accountability, precluding a clean causal inference. Ina study of gender stereotypes, Pendry and Macrae (1994) showedthat students in a control condition activated stereotypic associa-tions related to women, whereas students in an accountabilitycondition activated stereotypes associated with the more precisesubtype of businesswoman.

The results of other studies suggest that accountability has noappreciable effect on social bias. For example, Foschi (1996)found that accountability did not moderate the extent to whichstudents in a dyadic decision task exhibited gender bias in anevaluation. Similarly, in a series of studies involving undergradu-ates and law students, Girvan (2012) found that the effects ofstereotypes on participants’ mock legal decisions was not moder-ated by accountability or by a financial incentive. Finally, outsideof the laboratory, Dobbin, Schrage, and Kalev (2009) found mixedevidence for the extent to which accountability, along with variousequal opportunity policies and procedures, increased workplacediversity.

Training and Experience

A substantial body of research on expert decision-making sug-gests that experts make more accurate decisions in their domain ofexpertise because they are better able to distinguish betweenrelevant and irrelevant information (Haider & Frensch, 1996;Weiss & Shanteau, 2003). Moreover, uncertainty or ambiguity ina workplace decision is a risk factor for biased judgment (Fiske &Taylor, 2007; Yzerbyt, Schadron, Leyens, & Rocher, 1994). Ifexpertise reduces ambiguity by enabling professionals to focus onaccepted decision criteria that provide an alternative to genderstereotypes and attitudes, then expertise is likely to functionallydisable the use of nonspecified or illegitimate criteria in thedecision-making process (Girvan, 2012).

The results of studies of the effects of training and experienceare also mixed (see Ford et al., 2004), but there is reason to believethat experts may, under certain circumstances, be equipped toavoid bias. Weiss and Shanteau (2003) found that managementstudents but not hiring professionals used irrelevant information(e.g., age and gender) in hiring. In their words, “it is not easy to

ignore something as obvious as age or gender, even if required bythe guidelines. Professionals, however, followed strategies to doprecisely that” (p. 10). This evidence suggests that experts maylearn strategies that allow them to make decisions in their domainof expertise without resorting to stereotypes.

The Arbitration Context

To examine the implications of psychological theories of genderbias in legal decisions, and the potential for accountability andtraining to moderate such biases, we selected the context of laborarbitration. Heavily favored by Federal law (Drahozal, 2014;Sternlight, 2005), arbitration is the primary process through whichemployees who are members of a union resolve disputes with theiremployers. Labor arbitrators make binding judgments as towhether employers’ decisions were discriminatory; as such, thiscontext has substantial theoretical and practical significance forchecking gender bias. The labor arbitration context also offers astringent test of social-psychological theories. The cases that re-quire arbitration tend to represent the more difficult, complex, andambiguous disputes, and labor arbitrators have an almost unparal-leled degree of discretion with respect to the outcome of a matterbefore them—situational features that tend to enable and evenfacilitate unintended social bias (Dovidio & Gaertner, 2000;Souchon et al., 2010). Finally, little is known about how thearbitrators’ attitudes influence their decisions. The results of asmall number of prior studies show that arbitrators’ decisions areinfluenced by their values (Simpson & Martocchio, 1997) and inthe case of work-family conflict, their political ideology (Biernat& Malin, 2008), but none that directly link arbitrators’ explicit andimplicit gender attitudes to their decisions.

Hypotheses

As indicated in the review above, research on the psychology ofgender bias in professional legal decisions is sparse and mixed.Nevertheless, the theoretical processes outlined in the foregoingliterature review have some clear predictions for the factors thatlead to and moderate gender bias. Figure 1 summarizes thoserelationships.

As depicted in Figure 1, we hypothesize the following:

Figure 1. Multifaceted model of gender bias in the workplace.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

3GENERALIZABILITY OF GENDER BIAS

Hypothesis 1: In a stereotypically masculine work domain, par-ticipants will tend to favor male employee-grievants in theirarbitration decisions and disfavor female employee-grievants.

Hypothesis 2: Participants who are high in hostile (benevo-lent) explicit gender bias will tend to favor male (female)employee-grievants in their arbitration decisions and disfavorfemale (male) employee-grievants.

Hypothesis 3: Participants who have relatively positive im-plicit associations with men (women) will tend to favor male(female) employee-grievants in their arbitration decisions anddisfavor female (male) employee-grievants.

Hypothesis 4: Participants with arbitration training and expe-rience will tend to exhibit significantly less gender bias intheir arbitration decisions than those without training andexperience.

Hypothesis 5a: Participants who believe that they will beaccountable to a legitimate authority for making fair decisionswill evidence a higher level of integrative complexity in theexplanations of their arbitration decisions than those who arenot accountable.

Hypothesis 5b: Such participants will exhibit significantly lessgender bias in their arbitration decisions than those who arenot accountable.

We conducted two empirical studies to test these hypotheses.Study 1 was experimental: Participants were randomly assigned tomake decisions about male or female target individuals in twomock arbitration cases. Study 2 was correlational, involving aninvestigation of the relationship between arbitrators’ explicit andimplicit gender attitudes and their published decisions.

Study 1

Method

Participants.Students. Student participants were recruited in undergraduate

psychology courses at a large midwestern university. A total of213 undergraduate psychology students completed both parts ofthe study in exchange for extra credit. Implicit and explicit attitudemeasures and demographic characteristics for 98 of these partici-pants were not recorded due to incompatibility of the surveysoftware platform and the settings, applications, or operating sys-tem on participants’ computers. �2 tests comparing rates of dataloss across the experimental conditions indicated that data loss didnot vary systematically by condition (Student Grievant Conditions:�2(7) � 4.10, p � .77, Student Accountability Conditions: �2(1) �.35, p � .55). Thus, the loss can be considered to be completely atrandom (see, e.g., Sterne et al., 2009). Characteristics of theremaining student participants were as follows: Student partici-pants’ ages ranged from 18 to 40 years old (Mdn � 19); 75% werewomen; 51% White/Caucasian; 17% Asian, 6% Black/AfricanAmerican, 3% Latino, and 4% Other; 42% held, or had held, afull-time job; 9% had been union members; and 5% had priorexperience with arbitration, although none had arbitrated a case.

Arbitrators. To obtain a sample of individuals with arbitrationtraining and experience, we located publicly available lists of thenames of people who are registered, qualified, or certified toarbitrate cases and for whom contact information was given (see,e.g., Minnesota State Courts, 2014). A total of 959 people from 43U.S. states were identified in this way. Multiple contact attemptswere made for each individual using a combination of email, mail,and telephone calls. Of the attempted contacts, 144 resulted in anaffirmative indication that the contact information was no longervalid (e.g., returned e-mail or mail, or message that the phonenumber was out-of-service). Approximately 15% of the remaining815 people who were solicited responded (to e-mail and mailcommunications) or were successfully reached (through telephonecommunications). Approximately 78% of these agreed to partici-pate.

A total of 96 individuals with arbitration training and experiencecompleted both parts of the study. Each participant received a $10Starbucks gift card as compensation. As with the students, implicitand explicit attitude measures and demographic characteristics for44 of the participants were missing, but data loss did not varysystematically by condition (Arbitrator Grievant Conditions:�2(7) � 11.16, p � .13, Arbitrator Accountability Conditions:�2(1) � .12, p � .74). Approximately 21% of the arbitratorparticipants were women, ages ranged from 41 to 88 (Mdn � 64),and 100% self-identified as White/Caucasian. Sixty percent of thearbitrators were lawyers. The highest level of education completedby the arbitrators ranged from a bachelor’s degree to Ph.D.; 63%had received a J.D. The median number of cases arbitrated was500 total and 16 cases involving employment-related issues.

Materials and procedure.Part 1. Measures were administered over the Internet using a

combination of Inquisit (2009) and a customized survey platform.Student and arbitrator participants who indicated willingness toparticipate in the study were emailed a hyperlink that they coulduse to access the study materials. In Part 1, participants completedMale and Female Single-Target Implicit Association Tests (ST-IATs; Bluemke & Friese, 2008), the Ambivalent Sexism Inventory(Glick & Fiske, 1996), a three-item measure of political ideology,and demographics questions. The three attitude measures werepresented in a random order.

Responses to the Hostile Sexism and Benevolent Sexism sub-scales of the Ambivalent Sexism Inventory and the political ide-ology items were averaged, with appropriate items reverse coded,to produce variables for use in the analysis. Reliability of thesescales was acceptable (Hostile Sexism, � � .89, M � 3.44, SD �1.06; Benevolent Sexism, � � .82, M � 3.68, SD � .98; PoliticalIdeology, � � .87, M � 3.58, SD � .84). In addition, differencescores were computed for the ST-IATs using the algorithm rec-ommended by Bluemke and Friese (2008) (Male ST-IAT, M �.05, SD � .31; Female ST-IAT, M � .07, SD � .26). All individ-ual difference measures were converted to standardized scores forthe purposes of the regression analysis, reported below.

Part 2. A few days after completing Part 1, participants re-ceived a hyperlink that they could use to access the materials forPart 2 of the study. On clicking the link, participants who wererandomly assigned to the accountability condition read that a panelof professional arbitrators would contact them to discuss theirdecision, were asked to consent to the contact, and were asked toprovide a phone number at which the professional arbitrator could

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

4 GIRVAN, DEASON, AND BORGIDA

reach them (see the online supplemental materials for the completeverbatim text of the accountability manipulation).

Next, participants read and responded to questions about twomock arbitration cases, one involving a male and the other afemale employee, each of whom had been terminated from theirjob. To prevent order effects, the sex of the grievant and the orderof the cases were randomly assigned. (The parental status of thegrievant was also manipulated. The results of this manipulation didnot affect any of the analyses reported below and will be reportedseparately.) To create the fictional cases, the authors read a randomselection of approximately 20 published arbitration decisions in-volving the termination of an employee, noting common featuresof the facts of the underlying cases as well as the layout andpresentation. Based on this qualitative review, two truncated butrealistic arbitration cases were created (see the online supplemen-tal materials). Both describe, through stipulated facts and argumentsections, the circumstances surrounding the termination of anemployee in a blue-collar, stereotypically masculine job (i.e., aconstruction worker and a railroad yard pump-station operator).Two versions of each case were prepared, one in which theemployee-grievant’s name and relevant pronouns indicated thatshe was a woman, and another in which the name and pronounsindicated that he was a man. In a pilot test of the cases using astudent sample (N � 100), 54% of the participants decided againstthe grievant in each scenario, which was not significantly differentfrom an even 50–50 split (intercept-only binomial-probit model� � 0.04, z � 0.93, p � .353), suggesting that, overall, thescenarios lacked an obviously correct solution and thus wereambiguous. Further, grievant gender was a significant predictor(� � 0.57, p � .010) of decisions, indicating that participants hada tendency to decide to uphold the termination (i.e., to decideagainst the grievant) more often when the grievant was female (seeFigure 2).

After reading each case, participants indicated their decision onthe case by selecting one of two options: The employee’s dis-charge should be upheld (in favor of employer) or the employee’sdischarge should be overturned (in favor of employee/union).Participants also indicated the level of certainty associated withtheir decision on an 11-point Likert scale labeled with percentages

ranging from 0% to 100%, provided an open-ended justificationfor their decision, and completed 16 items designed to assessimpressions of the warmth and competence of the grievant andcompany. After indicating their decision on the second case, as amanipulation check, participants completed multiple-choice itemsfor each case in which they selected the name of the employee andreason for the discharge from a set of four alternatives. The studyconcluded with a debriefing page that provided additional infor-mation about the purpose of the study and the hypotheses.

Results

We conceptualized participants’ dichotomous decisions as re-flecting a latent assessment of the relative strength of the employ-ers’ justifications for their termination decisions (0 � terminationnot justified, 1 � termination justified). Accordingly, all analysesof participants’ termination decisions were conducted usingbinomial-probit generalized linear models, which use the cumula-tive normal distribution as a link function (Kutner, Nachtsheim,Neter, & Li, 2005). All analyses used clustered robust standarderrors to account for lack of independence in observations, whichadjusts standard errors to account for the correlation betweenobservations within a cluster and avoid alpha inflation (Cameron& Miller, 2010; see also Arai, 2009).

Manipulation check. We retained data from 89 arbitrators(92.7%) and 199 students (93.4%) who correctly answered bothmanipulation check questions. Of the participants who passed themanipulation check, complete data (i.e., Part 1 and Part 2) wereavailable for 50 arbitrators and 111 students. To maximize statis-tical power, we tested our experimental/quasi-experimental hy-potheses (H1, H4, H5) using the full set of responses from partic-ipants who passed the manipulation check and the individualdifference hypotheses (H2, H3) using a reduced set of responsesfrom the participants for whom complete Part 1 and Part 2 datawere available.

Descriptive analysis. The zero-order correlations among thegender-related attitudes, participant gender, participant age, andarbitrator status are provided in the online supplemental materials.Table 1 provides the mean, standard deviation, and results of t testscomparing the explicit and implicit gender attitudes for arbitratorsand students. Compared with the students, arbitrators had signifi-cantly lower levels of ambivalent sexism and significantly morepositive implicit attitudes toward both men and women.

Across both samples, 50% and 62% of the participants decidedagainst the grievant in the pump station and construction sitescenarios, respectively. This percentage was not significantly dif-ferent from a 50–50 split for the pump station scenario (intercept-only binomial-probit � � 0.07, CI[-.07, .22], p � .315) but wassignificantly different for the construction site scenario (intercept-only binomial-probit � � 0.23, CI[.08, .37], p � .003), suggestingthat participants in Study 1 found the latter scenario less ambigu-ous than the former.

Role congruity theory. To test Hypothesis 1, using decisions(N � 161) from student participants not assigned to the account-ability condition, we regressed termination decisions on a dummyvariable representing grievant gender (0 � man, 1 � woman) aswell as several control variables: scenario (0 � construction site,1 � pump station), participant age, and participant gender (0 �man, 1 � woman; model Likelihood-Ratio �2(5) � 13.63, p �

Figure 2. Decisions by grievant gender for each sample. � p � .05; �� p �.01.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

5GENERALIZABILITY OF GENDER BIAS

.018; McKelvey & Zavoina, 1975, Pseudo-R2 � .10). A post hocpower analysis indicated that the power to detect small (.02),medium (.15), and large (.35) effect sizes was .28, .99, and 1.00,respectively, exceeding the conventional threshold of .80 for me-dium and large effects (Cohen, 1992).

As predicted, grievant gender was a significant predictor ofparticipants’ decisions (� � 0.52, CI[.06, .99], p � .026), reflect-ing a 20% increased likelihood of an adverse decision for femalecompared to male grievant-employees [see Figure 2, Study 1(Students)]. Omission of any of the control variables did notmaterially affect the results of the analysis. Participant gender didnot interact with grievant gender to significantly moderate thiseffect (� � �0.73, CI[�1.66, .23], p � .140), nor did grievantgender interact significantly with scenario (� � 0.58, CI[�0.13,1.30], p � .111).

Professional training. To Test Hypothesis 4, we regresseddecisions (N � 570) on grievant gender and professional training(student � 0, arbitrator � 1) and the interaction between them, aswell as control variables: scenario, accountability, and grievantparental status (grievant parental status was manipulated but didnot at all impact the Study 1 results; model Likelihood-Ratio�2(6) � 12.54, p � .051, McKelvey & Zavoina, 1975, Pseudo-R2 � .04). A post hoc power analysis of the model indicated thatthe power to detect small (.02), medium (.15), and large (.35)effect sizes (Cohen, 1992) was .74, 1.00, and 1.00, respectively.

The coefficients for grievant gender and professional trainingwere marginally significant (� � 0.24, CI[�.02, .50], p � .070)and significant (� � 0.59, CI[.19, .98], p � .003), respectively.Both effects were qualified by the predicted grievant gender xprofessional training interaction (� � �0.53, CI[�1.02, �.04],p � .035). Removing the control variables did not substantivelyalter this result. In support of Hypothesis 4, post hoc simple slopesanalysis showed that student participants had a marginally signif-icant tendency to decide cases against female grievants (� � 0.24,CI[�.02, .51], p � .069), whereas grievant gender was not asignificant predictor of decisions among arbitrators (� � �0.29,CI[�.71, .13], p � .185) [see Figure 2, Study 1 (Arbitrators)].

Accountability. Two research assistants who had completed acertification program in coding integrative complexity (Suedfeld,2005) and were blind to condition rated the level of integrativecomplexity of more than half of the participants’ justifications fortheir decisions. Based on double coding of a random sample of 36responses, intercoder reliability was adequate for a multidimen-sional construct (72% complete agreement; Type I interclass cor-relation � .51). Disagreements were resolved by averaging thevalues of the two codes.

To Test Hypothesis 5a, we regressed the integrative complexityscores on indicator codes for accountability condition (consented)and accountability condition (declined; 24.7% of participants in

the accountability condition did not consent to be contacted byprofessional arbitrators regarding the reason for their decision),leaving assignment to the control condition as the comparisoncategory (R2 � .01, F(2, 547) � 3.29, p � .038). A post hoc poweranalysis indicated that the power to detect small (.02), medium(.15), and large (.35) effect sizes (Cohen, 1992) in the model was.91, 1.00, and 1.00, respectively. Consistent with Hypothesis 5a,the responses of accountable participants who consented to becontacted evidenced significantly higher levels of integrative com-plexity than those in the control condition (� � 0.21, CI[.01, .42],p � .041). By contrast, the responses of participants assigned tothe accountability condition who declined to be contacted showedlevels of integrative complexity similar to those of participants inthe control condition (� � �0.06, CI[�.36, .24], p � .697). Giventhis result, we combined the accountability condition (declined)with the control group in all other analyses.

To Test Hypothesis 5b, we reran the model testing H1 in the fullsample of participants, adding the higher-order interaction betweenAccountability and Grievant Gender. In addition, we included theinteractions of these with Arbitrator Status to account for thedifferences in the direction of gender bias in the decision betweenStudents and Arbitrators. Specifically, using the full data set, weadded terms for Accountability and the interactions between theindicator code for Accountability � Grievant Gender, Account-ability � Arbitrator, and the three-way interaction between Ac-countability � Grievant Gender � Arbitrator (model Likelihood-Ratio �2(9) � 13.37, p � .147, McKelvey & Zavoina, 1975,Pseudo-R2 � .04). Neither the coefficient for Accountability, norany of the higher-order interactions with it, were significant (allps .35). Removing the nonsignificant three-way interaction totest the moderating effects of Accountability collapsing acrossstudent and arbitrator samples did not change the result (ps .71).A post hoc power analysis of the model indicated that the power todetect small (.02), medium (.15), and large (.35) effect sizes(Cohen, 1992) was .65, 1.00, and 1.00, respectively.

Explicit and implicit attitudes. To Test Hypotheses 2 and 3,using decisions from student and arbitrator participants (N � 214and N � 95, respectively) for whom we had complete data, weregressed decisions on grievant gender, hostile and benevolentsexism scores, Male and Female ST-IAT scores, and the interac-tions between grievant gender and each of the gender attitudemeasures. Also included were controls for professional training,scenario, accountability, participant gender, and number of casesarbitrated (model Likelihood-Ratio �2(13) � 20.90, p � .075;McKelvey & Zavoina, 1975, Pseudo-R2 � .11). A post hoc poweranalysis of the model indicated that the power to detect small (.02),medium (.15), and large (.35) effect sizes (Cohen, 1992) was .28,1.00, and 1.00, respectively. Multicollinearity of the correlatedgender attitude measures was below the most conservative thresh-

Table 1Explicit and Implicit Gender Attitudes of Arbitrators and Students

Attitude Arbitrators Students t test

Hostile Sexism M � 2.79, SD � 1.04 M � 3.71, SD � .98 t(152) � �5.2, p � .001Benevolent Sexism M � 3.17, SD � .96 M � 3.91, SD � .92 t(152) � �4.5, p � .001Male ST-IAT M � .13, SD � .29 M � .02, SD � .32 t(152) � 2.0, p � .053Female ST-IAT M � .13, SD � .25 M � .05, SD � .27 t(152) � 2.0, p � .045

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

6 GIRVAN, DEASON, AND BORGIDA

old for concern and further investigation (all VIFs �4; O’Brien,2007). None of the hypothesized interactions between GrievantGender and any measure of explicit or implicit gender attitude wassignificant (ps .153).

To test the possibility that null results may have occurredbecause gender attitudes operate differently in Students and Arbi-trators, we reran the model including all two-way interactions andthe three-way interactions between Arbitration Status, GrievantGender, and each of the Explicit and Implicit Gender Attitudes(model Likelihood-Ratio �2(22) � 27.48, p � .193; McKelvey &Zavoina, 1975, Pseudo-R2 � .14). The three-way interactions werenonsignificant (ps .235). Multicolinearity across these coeffi-cients was a potential problem (VIFs up to 5.36). Accordingly, weran the analysis separately for students and arbitrators.

For the student subsample (N � 214), we regressed Decision onthe indicator variable for Grievant Gender as well as controls forScenario, Accountability, and participant characteristics (e.g., gen-der and age; model Likelihood-Ratio �2(12) � 17.228, p � .141;McKelvey & Zavoina, 1975, Pseudo-R2 � .14). Again, none of theinteractions between Grievant Gender and participants’ genderattitudes were significant (all ps .190). Multicollinearity of thecorrelated gender attitude measures was below the most conser-vative threshold for concern and further investigation (allVIFs �4; O’Brien, 2007). A post hoc power analysis indicated thatthe power to detect small (.02), medium (.15), and large (.35)effect sizes (Cohen, 1992) was .16, .92, and 1.00, respectively.

Replicating the analysis in the arbitrator sample (N � 95), weregressed Decision on the indicator variable for Grievant Genderas well as controls for Scenario, Accountability, Parental Status,and participant characteristics (e.g., gender, age, square-root of thenumber of arbitration decisions; McKelvey & Zavoina, 1975,Pseudo-R2 � .15). Again, the interactions between Grievant Gen-der and participants’ gender attitudes were not significant (all ps .604). Multicolinearity in the model was below the most conser-vative threshold for further investigation (all VIFs �4; O’Brien,2007). A post hoc power analysis indicated that the power to detectsmall (.02), medium (.15), and large (.35) effect sizes (Cohen,1992) in the model was .10, .61, and .97, respectively. Omission ofthe control variables, none of which was significant, increased thepower of the model (power for small, medium, and large effectsizes .11, .70, and .99, respectively), but did not change the resultsof the analysis.

Study 1 Discussion

The results of Study 1 supported Hypothesis 1. In the stereo-typically masculine work domains of construction and a pumpstation, student participants who were not assigned to the account-ability condition tended to decide cases against female grievants,but in favor of male grievants. Although the design does notpresent a complete test of role congruity theory—there was nostereotypically feminine comparison condition—it is entirely con-sistent with the theory’s predictions regarding the employmentsettings in which gender bias against women is most likely to bepresent.

The results of Study 1 did not support Hypotheses 2 or 3. Ininteraction with grievant gender, neither measures of explicit norimplicit attitudes were significantly related to participants’ deci-sions. Although these findings are inconsistent with the results of

prior research involving student populations suggesting that gen-der attitudes influence similar judgments (e.g., Wiener, Hurt, Rus-sell, Mannen, & Gasper, 1997), this result is similar to prior studiesof legal professionals, in which judges were observed to haveimplicit biases but that these biases were not related to the judges’decisions in mock cases (e.g., Levinson & Young, 2010; Rachlin-ski et al., 2008).

Turning to system-level moderators, the results of Study 1provide strong support for Hypothesis 4. Whereas the students’decisions evidenced the expected gender bias, the decisions madeby the sample of trained, experienced arbitrators did not. To thecontrary, and consistent with the results of studies involving cod-ing of actual arbitration decisions, if anything, arbitrators had a(nonstatistically significant) tendency to exhibit a chivalrous orpaternalistic decision pattern favoring female grievants. Even so,given the known groups design and substantial differences be-tween the student and arbitrator samples beyond just training andexpertise (e.g., age, life experience), drawing strong causal infer-ences from this result is inappropriate.

Finally, in support of Hypothesis 5a, the accountability manip-ulation increased the level of integrative complexity of partici-pants’ open-ended responses in which they explained the reasonsfor their decisions. However, contrary to Hypothesis 5b, account-ability did not significantly reduce the observed gender bias inparticipants’ decisions. On the whole, this result is entirely con-sistent with the weight of the few existing studies that experimen-tally examine the impact of accountability on social biases (i.e.,attitudes and stereotypes) reviewed above but is contrary to theweight of evidence about the moderating effects of accountabilityon judgment heuristics (e.g., Lerner & Tetlock, 1999).

Study 2

The results of Study 1 provided initial evidence that training andexperience, but not accountability, may attenuate gender bias inarbitration decisions. In Study 2, we extended these results by (a)directly examining the prediction of role congruity theory that thegender stereotypicality of a workplace will moderate gender biasand (b) examining the generalizability of the debiasing effects oftraining and experience to the context of actual arbitration deci-sions. To do this, we identified published arbitration opinionsissued by arbitrators who participated in Study 1 to determinewhether arbitrators’ gender-related attitudes are associated withtheir decisions in actual arbitration cases.

Method

Participants. Participants consisted of 40 arbitrators forwhom we had measures of implicit and explicit gender attitudesfrom Study 1 and for whom we were able to locate at least onepublished arbitration decision in Westlaw’s labor arbitrationawards databases (Thomson Reuters, 2012). Demographic charac-teristics of the participants were as follows: 30% Women, agesranged from 48 to 80 (Mdn � 63); 100% self-identified as White/Caucasian; 72% were lawyers; and the median number of casesarbitrated was 650 general and 16 cases involving employment-related issues.

Materials and procedure. Arbitrator demographic character-istics and measures of individual differences were collected in

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

7GENERALIZABILITY OF GENDER BIAS

Study 1. Arbitration opinions used in the analysis were retrievedusing searches of Westlaw arbitration databases for the names andlocations of the arbitrators for whom we had data from Part 1 ofStudy 1. Because of an unequal gender distribution among thegrievants in the actual cases, of these, all of the cases involving afemale grievant (N � 56), and a random sample of those involvinga male grievant (N � 56), were selected for inclusion in Study 2.

The arbitration opinions that we obtained in our search wererandomly ordered before coding. Trained coders, blind to studyhypotheses, coded each case for the gender of the grievant (0 �man, 1 � woman), the state in which the arbitration hearing washeld, the date of the decision, whether the employer was a publicor private entity (0 � public, 1 � private), and for the decisionitself (0 � uphold the grievance, .45 � reinstated with partial backpay, .5 � reinstated without back pay, .55 � entitled to futureopening, and 1 � uphold the employer’s termination decision orother discipline). In addition, as a measure of the gender stereo-typicality of the work setting represented in each case, a trainedcoder recorded the percentage of employees who are women in theemployee-grievant’s occupation (M � 37.1%, Mdn � 26.9%,SD � 30.8%). These percentages were obtained from the Bureauof Labor Statistics (U.S. Department of Labor, Bureau of LaborStatistics, 2012). Intercoder agreement was 100%.

After dropping one case for lack of codable material, the finalsample of cases consisted of 111 labor-arbitration cases fromarbitrations in 29 U.S. states. Decision dates ranged from 1979 to2009 (Mdn � 1999). Fifty-three percent of the cases were Private(vs. Public). Distribution of decisions was as follows: 49% upheldthe termination or other discipline, 17% were compromise deci-sions (e.g., reinstatement without back pay), and 34% found for theemployee/grievant.

Results

For the analysis in Study 2, we again conceptualized the arbi-tration decisions as an assessment of the relative strength of theemployer’s justification for its termination decision (0 � notjustified, 1 � justified). Given the proportional data, we used aquasi-binomial probit model to account for over dispersion (Craw-ley, 2013) and corrected for lack of independence using clusterrobust standard errors.

Role congruity theory. To test an extension of Hypothesis 1,that is, that in stereotypically masculine (feminine) work domains,arbitrators will tend to favor male (female) employee-grievants, weregressed termination decisions on grievant gender, the genderstereotypicality of the relevant work setting, and their interaction,as well as controls for participant age, gender, experience, the yearthe case was decided, and whether it was public or private (modelLikelihood-Ratio �2(8) � 5.52, p � .700, McKelvey & Zavoina,1975, Pseudo-R2 � .08). A post hoc power analysis indicated thatthe power to detect small (.02), medium (.15), and large (.35)effect sizes (Cohen, 1992) was .14, .82, and 1.00, respectively.Neither grievant gender (� � 0.13, CI[�.44, .71], p � .646; seeFigure 2) nor the key interaction between grievant gender andgender stereotypicality (� � �.08, CI[�.69, .52], p � .809) weresignificant predictors. A sensitivity test using a proportional oddsmodel again yielded substantively similar results (Grievant gen-der � Workplace gender stereotypicality interaction � � .04,

CI[�1.00, 1.08], p � .937; model Likelihood-Ratio �2(6) � 3.43,p � .753; R2 � .04).

Explicit and implicit bias. To test Hypotheses 2 and 3, par-ticipants’ implicit and explicit gender attitudes were added to themodel, along with interactions between each of the attitude mea-sures and grievant gender (model Likelihood-Ratio �2(12) �12.24, p � .427; McKelvey & Zavoina, 1975, Pseudo-R2 � .16).Multicollinearity in the model was below the most conservativethreshold for further investigation (all VIFs �4; O’Brien, 2007). Apost hoc power analysis indicated that the power to detect small(.02), medium (.15), and large (.35) effect sizes (Cohen, 1992) was.12, .77, and .99, respectively. Removing the control variablesincreased the power, .13, .81, and 1.00, respectively, withoutsubstantively changing the results of the analysis.

The results indicate that the main effect of grievant genderremained nonsignificant (� � (� � 0.25, CI[�.18, .67], p � .259).However, the Grievant gender � Benevolent sexism (� � �0.63,CI[�1.05, �.21], p � .003) and grievant gender � ST-IATFemale (� � �0.65, CI[�1.30, �.01], p � .048) interactions weresignificant predictors of the termination decision, and the Grievantgender � Hostile sexism interaction was marginally significant(� � 0.46, CI[�.06, .98], p � .085). An ordinal regressionsensitivity test using a proportional odds model yielded substan-tively similar results: Grievant gender � Benevolent sexism(� � �1.08, CI[�2.02, �.22], p � .014); Grievant gender �ST-IAT Female (� � �1.14, CI[�2.06, �.22], p � .015); Griev-ant gender � Hostile sexism (� � 0.80, CI[�.15, 1.76], p � .100;model Likelihood-Ratio �2(10) � 14.75, p � .141; R2 � .14).

Simple-slopes analysis showed that, when the grievant wasfemale, benevolent sexism and implicit attitudes toward womenwere, respectively, marginally significant (� � �0.52, CI[�1.07,.03], p � .066) and significant (� � �0.68, CI[�1.16, �.19], p �.006) negative predictors of upholding the termination decision,whereas hostile sexism was a significant positive predictor ofupholding termination (� � 0.53, CI[.10, .96], p � .015). Bycomparison, none of the arbitrators’ gender-related attitudes weresignificant predictors of their decisions when the grievant wasmale (all ps � .139). As an indication of the practical significanceof gender attitudes on decisions, Table 2 gives the marginal effectsof a one standard deviation change in arbitrators’ explicit andimplicit gender attitudes on the probability of a decision againstthe grievant.

Study 2 Discussion

Study 2 was designed to test the external validity of the findingthat arbitrator participants do not exhibit the same bias against

Table 2Marginal Effects of One Unit Change in Arbitrators’ GenderAttitudes (z Scores) on the Probability of an AdverseArbitration Decision

Attitude Female grievant (%) Male grievant (%)

Hostile Sexism 20.6� �7.3Benevolent Sexism �20.1† 11.2Male ST-IAT 2.4 �1.5Female ST-IAT �26.4�� 1.6

† p � .10. � p � .05. �� p � .01.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

8 GIRVAN, DEASON, AND BORGIDA

women as student participants in decisions involving stereotypi-cally masculine work settings. Consistent with Hypothesis 4 andthe experimental results in Study 1, grievant gender itself did notsignificantly affect the decisions of the professionals in the sampleof actual labor arbitration cases. Study 2 also extended the resultsof Study 1 by exploring the influence that arbitrators’ explicit andimplicit gender attitudes had on their decisions in actual cases.Here, unlike the experimental results, both benevolent sexism andimplicit attitudes toward women were significant predictors ofarbitrators’ decisions, and hostile sexism was a marginally signif-icant predictor, all in the hypothesized direction. Although furtherresearch is needed, this difference in the laboratory and fieldresults raises the possibility that laboratory studies exploring theinfluence of explicit and implicit gender attitudes in decision-making may underestimate the influence of those attitudes inreal-world contexts involving professional legal decision-makers.

General Discussion

Social-psychological research has established several theories ofhow gender bias affects decisions, but critics argue that theoriestested in laboratory research do not generalize to professionalsettings because decision-makers in such contexts have trainingand are subject to accountability. Claims about the effects of theseproposed moderators have not been rigorously tested, and re-searchers’ consideration of the effects of the social-psychologicalprocesses associated with gender bias in professional legal deci-sions in particular has been very limited. In the studies reportedhere, we begin to fill these gaps with an examination of laborarbitration decision-making. Arbitrators have wide discretion tomake binding decisions in some of the most ambiguous employeegrievance cases, making this an ideal context in which to examinethe sources and moderators of social bias in professional decision-making.

In two studies we examined (a) the impact of gender bias fromrole incongruity, ambivalent sexism, and implicit bias on laborarbitration decisions and (b) the potential for accountability andtraining to attenuate gender bias. In support of Hypothesis 1, andconsistent with prior research, we found that the overall decisionsof untrained student participants regarding individuals in stereo-typically masculine settings were biased against female grievants.In contrast, and consistent with Hypothesis 4, professional arbi-trators’ decisions in mock and real arbitration cases did not showevidence of gender bias from role incongruity, suggesting thattraining and experience can attenuate at least the broader, sociallyshared forms of gender bias. We also found, however, that arbi-trators’ gender attitudes predicted their decisions in real arbitrationcases. Consistent with Hypothesis 5a, participants who believedthey would be held accountable to a legitimate authority engagedin a more integratively complex thinking style. Contrary to Hy-pothesis 5b, however, accountability did not attenuate the effectsof the observed gender bias.

Gender Bias in Labor Arbitration Decisions

In our experimental study, the results among students are con-sistent with other research on gender bias in masculine workcontexts and the prediction of role congruity theory that womenare most likely to experience gender prejudice in domains and on

tasks that are stereotypically masculine (Eagly & Karau, 2002;Heilman, 2012). Our results also extend research in this area byincluding a direct comparison between student participants andprofessional arbitrators. This allows us to begin to draw conclu-sions about the generalizability of research conducted on nonpro-fessional convenience samples, and to directly test claims aboutmoderators of gender bias that are likely to arise in real-world legaldecision contexts.

The Potential for Training and Accountability toAttenuate Gender Bias

The different decision patterns among students and arbitrators inthe experimental study suggests that, as critics of the social-psychological research argue, there are important differences be-tween the decisions made by trained and untrained decision-makers that may enable trained professionals to suppress orotherwise control gender bias. Greater accountability in profes-sional contexts is also offered as a reason for the difference of thesort that we found in this study. However, results indicated thatalthough accountability pressure enhanced the integrative com-plexity of the decision-making process, it did not attenuate genderbias in arbitration decisions.

Training and experience remain a compelling explanation forthe relative lack of gender bias in the professional arbitrators’decision pattern. According to the Social Contingency Model,when individuals know before they make a decision that they willbe accountable to a knowledgeable, legitimate authority, and knowthe preferences of that authority, they use the decision criteriapreferred by the authority (Lerner & Tetlock, 1999). Expertise ina given domain may similarly enable the use of specific decisioncriteria that are incompatible with a gender bias or otherwise allowexperts to functionally avoid or account for the use of illegitimatecriteria (e.g., the sex of the target person) in the decision-makingprocess (for a review see Girvan, 2012). Thus, through theirexperience, experts may be equipped with specialized knowledgethat not only activates an integratively complex thinking style, butalso provides information about the correct methods for arriving ata decision.

Implicit and Explicit Attitudes in ActualLegal Decisions

The results also provide evidence that implicit and explicitgender attitudes influence legal professionals when they maketheir actual decisions. With respect to explicit attitudes, we foundthat arbitrators high in benevolent sexism were more likely tooverturn the termination when the grievant was a woman, but notwhen the grievant was a man: Arbitrators who endorse a protec-tive, paternalistic attitude toward women are more inclined toreject termination of women from their jobs. Benevolent sexismmay elicit patronizing discrimination toward women at work(Cikara & Fiske, 2009; Glick & Fiske, 2007; Rudman, Glick, &Phelan, 2007). The results of this study are thus consistent withambivalent sexism theory and with the chivalry-paternalism thesisdescribed in prior research on arbitration decision-making, whichposits that men view themselves as the protectors of all women,who are innocent and defenseless (Bankston, 1976; Bemmels,1988a, 1988b; Ponak, 1987). This extension of the benevolent

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

9GENERALIZABILITY OF GENDER BIAS

sexism research into the legal decision context provides newinsight into how paternalistic gender attitudes may be a dominantinfluence on decisions in naturalistic contexts. In addition, consis-tent with the findings of research conducted on other studentsamples (e.g., Masser & Abrams, 2004), arbitrators higher inhostile sexism were marginally more likely to uphold the termi-nation of a female grievant, but not a male grievant, in actual cases.Thus, among both professional arbitrators and students, those whoendorse an ideology of male dominance and male superiority maytend to act in ways that police women out of masculine occupa-tions.

We found a similarly, potentially paternalistic pattern of influ-ence of implicit attitudes on real case decisions. Arbitrators withpositive implicit attitudes toward women were more likely to findin favor of a female grievant, an implicit form of paternalism that,to our knowledge, has not previously been documented in a legalor other naturalistic decision context. Prior research showed thatimplicit attitudes toward women in positions of authority tend tobe more negative than attitudes toward women, in general(Richeson & Ambady, 2001; Rudman & Kilianski, 2000). How-ever, few studies have examined the association between implicitgender attitudes and decisions. Although it is the first to focus onactual legal decisions, our study joins the growing body of researchdocumenting the impact of implicit attitudes regarding other pro-fessionals on decisions or behaviors they made while on the job,and extends this research by establishing evidence of an implicitgender bias in arbitration decision-making.

The Laboratory and the Field

With respect to role-congruity theory and the potential effects oftraining and experience, the results of our laboratory experimentand analysis of real decisions were consistent. Students, but nottrained arbitrators, decided cases in ways that were biased againstwomen in stereotypically masculine contexts. Not so withindividual-level attitudes. Results showed an association betweenambivalent sexism and implicit attitudes toward women and arbi-trators’ decisions in real cases, although these sources of genderbias did not predict the outcome of cases in the experimental study.This finding suggests that individual differences may be moreinfluential in naturalistic decision contexts than in the laboratory.It also underscores the importance of assessing the impact ofpsychological factors in real-world contexts.

Limitations and Future Directions

Despite many new insights afforded by these studies, they arecharacterized by several limitations. These limitations, in turn,give rise to questions that are worthy of future study. First, thearbitration fact patterns created for the experiment were carefullyconstructed to produce low agreement in the decisions amongparticipants. Arbitration cases that reach a final arbiter are likely tobe similarly ambiguous and difficult, but the findings of thesestudies may not generalize to other contexts in which the correctdecision is more straightforward.

Second, in Study 1 we found evidence of gender bias in stu-dents’ decisions that did not appear in the decisions of arbitrators.Given the known groups design, the experiment does not allow usto pinpoint the precise reason for this difference. Professional

training and/or expertise remains only a possible explanation forarbitrators’ ability to achieve overall gender parity in their deci-sions. Arbitrator age, gender, and variation in arbitration experi-ence were controlled for, but the result is nevertheless confoundedwith other systematic differences in professional and life experi-ence, as well as possible cohort and history effects, controls forwhich were not in the model. Additional research in which trainingand expertise are experimentally manipulated (see, e.g., Girvan,2012) is necessary to rule out, or assess the effects of, theseinfluences.

Third, participants completed measures of implicit and explicitgender attitudes before deciding the arbitration cases. This isconsistent with the causal direction of the underlying theory andensures that any observed attitude-decision relationship is notattributable to the experience of the decision task itself (i.e., thatsignificant effects are not due to the experimental manipulationchanging participant attitudes). But this order raises the possibilitythat participants responses will be sensitized to potential demandcharacteristics created by completion of the measures, a possiblealternate explanation for why the arbitrators in Study 1 demon-strated no promale bias. (In Study 2 measures were taken in mostcases years after the arbitration decisions were made, eliminatingthe possibility of demand characteristics but presenting the alter-nate explanation that arbitrators’ gender attitudes might covarywith their actual decisions because their current attitudes reflecttheir prior experiences deciding cases.)

Several factors suggest that demand characteristics did not ma-terially affect the results here. First, in studies involving assess-ment of gender attitudes where the order of the measures andmanipulation is counterbalanced, order has been found not toinfluence responses (see, e.g., Wiener & Hurt, 2000). Second, weattempted to reduce the potential that participants might be influ-enced by the measures by administering Part 1 (individual differ-ence measures) and Part 2 (the judgment task) of Study 1 severaldays apart. Third, both student and arbitrator participants com-pleted the same attitude measures before deciding the cases, butonly arbitrators did not evidence gender bias in their decisions.Finally, the absence of a main effect of grievant gender in thearbitrator’s decisions in Study 1 is consistent with the weight of theresults of studies of actual arbitration cases, including the results ofStudy 2. That said, we cannot rule out the possibility that height-ened sensitivity to pressure to remain neutral among arbitrators,rather than, in addition to, or as a result of their training andexperience, as a possible explanation for the results of the labstudy. Similarly, as the studies did not involve true random sam-ples, it may be that these null effects and others (e.g., the failure ofaccountability to moderate gender bias even though it was asso-ciated with increases in integrative complexity) are attributable toself-selection effects and other variables not assessed directly orcontrolled for here. Future research should explore whether poten-tial differences like these, especially those between experts andnonexperts, impact responses in laboratory research.

Finally, post hoc analysis indicated that all of the statisticalmodels were adequately powered to detect moderate effects exceptthat in Study 1 testing the effects of gender attitudes on thearbitrator subsample in particular. Even so, because of a loss ofdata, statistical power was not as high as it might otherwise havebeen. It is thus possible that some of the effects which werenonsignificant or trending in these analyses (e.g., the effects of

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

10 GIRVAN, DEASON, AND BORGIDA

grievant gender in the arbitrator sample in Study 1 and the inter-action between Hostile Sexism and grievant gender in Study 2)might also have been significant without this loss. Accordingly,future research replicating these results in professional sampleswill be important to verifying the effects of training and expertiseon the operation of gender bias.

At a practical level, a better understanding of how arbitratorsand other legal experts make decisions is crucial to improvingtraining and reducing gender bias in this and other organizationalcontexts. Identifying the aspects of arbitrator experience that pro-duce this reduction in bias is a worthy goal for future investiga-tions.

References

Abrams, D., Viki, G. T., Masser, B., & Bohner, G. (2003). Perceptions ofstranger and acquaintance rape: The role of benevolent and hostilesexism in victim blame and rape proclivity. Journal of Personality andSocial Psychology, 84, 111–125. http://dx.doi.org/10.1037/0022-3514.84.1.111

Agerström, J., & Rooth, D. O. (2011). The role of automatic obesitystereotypes in real hiring discrimination. Journal of Applied Psychology,96, 790–805. http://dx.doi.org/10.1037/a0021594

Amodio, D. M., & Devine, P. G. (2006). Stereotyping and evaluation inimplicit race bias: Evidence for independent constructs and uniqueeffects on behavior. Journal of Personality and Social Psychology, 91,652–661. http://dx.doi.org/10.1037/0022-3514.91.4.652

Arai, M. (2009). Cluster-robust standard errors using R. Retrieved fromhttp://people.su.se/ma/clustering.pdf

Bankston, E. W. (1976). Value differences between attorney and economistlabor arbitrators. In J. Stern & B. Dennis (Eds.), Proceedings of theTwenty-Ninth Annual Winter Meetings, Industrial Relations ResearchAssociation (pp. 151–160). Madison, WI: IRRA.

Bemmels, B. (1988a). The effect of grievants’ gender on arbitration deci-sions. Industrial & Labor Relations Review, 41, 251–262. http://dx.doi.org/10.2307/2523635

Bemmels, B. (1988b). Gender effects in discharge arbitration. Industrial &Labor Relations Review, 42, 63–76. http://dx.doi.org/10.2307/2523172

Biernat, M., & Malin, M. H. (2008). Political ideology and labor arbitra-tors’ decision making in work-family conflict cases. Personality andSocial Psychology Bulletin, 34, 888–899. http://dx.doi.org/10.1177/0146167208316689

Blair, I. V., Dasgupta, N., & Glaser, J. (2015). Implicit attitudes. In E.Borgida & J. Bargh (Eds.), APA handbook of personality and socialpsychology, attitudes and social cognition. Washington, DC: APABooks. http://dx.doi.org/10.1037/14341-021

Bluemke, M., & Friese, M. (2008). Reliability and validity of the Single-Target IAT (ST-IAT): Assessing automatic affect toward multiple atti-tude objects. European Journal of Social Psychology, 38, 977–997.http://dx.doi.org/10.1002/ejsp.487

Borgida, E., & Girvan, E. J. (2015). Social cognition in law. In E. Borgida& J. Bargh (Eds.), APA handbook of personality and social psychology,attitudes and social cognition. Washington, DC: APA Books. http://dx.doi.org/10.1037/14341-024

Bornstein, B. H. (1999). The ecological validity of jury simulations: Is thejury still out? Law and Human Behavior, 23, 75–91. http://dx.doi.org/10.1023/A:1022326807441

Bowen, C. C., Swim, J. K., & Jacobs, R. R. (2000). Evaluating genderbiases on actual job performance of real people: A meta-analysis. Jour-nal of Applied Social Psychology, 30, 2194–2215. http://dx.doi.org/10.1111/j.1559-1816.2000.tb02432.x

Cameron, A. C., & Miller, D. L. (2010). Robust inference with clustered

data. In A. Ullah & D. E. Giles (Eds.), Handbook of empirical economicsand finance (pp. 1–28). Boca Raton, FL: Chapman & Hall/CRC. http://dx.doi.org/10.1201/b10440-2

Cikara, M., & Fiske, S. T. (2009). Warmth, competence, and ambivalentsexism: Vertical assault and collateral damage. In M. Barreto, M. K.Ryan, & M. T. Schmitt (Eds.), The glass ceiling in the 21st century:Understanding barriers to gender equality. Washington, DC: AmericanPsychological Association. http://dx.doi.org/10.1037/11863-004

Cikara, M., Lee, T. L., Fiske, S. T., & Glick, P. (2009). Ambivalent sexismat home and at work: How attitudes toward women in relationshipsfoster exclusion in the public sphere. In J. T. Jost, A. C. Kay, & H.Thorisdottir (Eds.), Social and psychological bases of ideology andsystem justification. New York, NY: Oxford University Press. http://dx.doi.org/10.1093/acprof:oso/9780195320916.003.018

Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159.http://dx.doi.org/10.1037/0033-2909.112.1.155

Crawley, M. J. (2013). The R book (2nd ed.). Chichester, UK: Wiley.Davison, H. K., & Burke, M. J. (2000). Sex discrimination in simulated

employment contexts: A meta-analytic investigation. Journal of Voca-tional Behavior, 56, 225–248. http://dx.doi.org/10.1006/jvbe.1999.1711

Dobbin, F., Schrage, D., & Kalev, A. (2009). Someone to watch over me:Coupling, decoupling, and unintended consequences in corporate equalopportunity. Working Paper, Department of Sociology, Harvard Univer-sity, Cambridge, MA.

Dovidio, J. F., & Gaertner, S. L. (2000). Aversive racism and selectiondecisions: 1989 and 1999. Psychological Science, 11, 315–319. http://dx.doi.org/10.1111/1467-9280.00262

Drahozal, C. R. (2014). FAA preemption after Concepcion. BerkeleyJournal of Employment and Labor Law, 35, 153–174.

Eagly, A. H., & Karau, S. J. (2002). Role congruity theory of prejudicetoward female leaders. Psychological Review, 109, 573–598.

Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition.research: Their meaning and use. Annual Review of Psychology, 54,297–327. http://dx.doi.org/10.1146/annurev.psych.54.101601.145225

Fiske, S. T., & Glick, P. (1995). Ambivalence and stereotypes cause sexualharassment: A theory with implications for organizational change. Jour-nal of Social Issues, 51, 97–115. http://dx.doi.org/10.1111/j.1540-4560.1995.tb01311.x

Fiske, S. T., & Taylor, S. E. (2007). Social cognition: From brains toculture. New York, NY: McGraw-Hill.

Ford, T. E., Gambino, F., Lee, H., Mayo, E., & Ferguson, M. A. (2004).The role of accountability in suppressing managers’ preinterview biasagainst African-American sales job applicants. Journal of PersonalSelling & Sales Management, 24, 113–124.

Foschi, M. (1996). Double standards in the evaluation of men and women.Social Psychology Quarterly, 59, 237–254. http://dx.doi.org/10.2307/2787021

Girvan, E. J. (2012). Habits of meaning: When legal education and otherprofessional training attenuate bias in social judgments (Doctoral dis-sertation). University of Minnesota Digital Conservancy: http://purl.umn.edu/128753

Glick, P., Diebold, J., Bailey-Werner, B., & Zhu, L. (1997). The two facesof Adam: Ambivalent sexism and polarized attitudes toward women.Personality and Social Psychology Bulletin, 23, 1323–1334. http://dx.doi.org/10.1177/01461672972312009

Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory: Dif-ferentiating hostile and benevolent sexism. Journal of Personality andSocial Psychology, 70, 491–512. http://dx.doi.org/10.1037/0022-3514.70.3.491

Glick, P., & Fiske, S. T. (2007). Sex discrimination: The psychologicalapproach. In F. J. Crosby, M. S. Stockdale, & S. A. Ropp (Eds.), Sexdiscrimination in the workplace. Malden, MA: Blackwell.

Goff, P. A., Jackson, M. C., Di Leone, B. A. L., Culotta, C. M., &DiTomasso, N. A. (2014). The essence of innocence: Consequences of

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

11GENERALIZABILITY OF GENDER BIAS

dehumanizing Black children. Journal of Personality and Social Psy-chology, 106, 526–545. http://dx.doi.org/10.1037/a0035663

Green, A. R., Carney, D. R., Pallin, D. J., Ngo, L. H., Raymond, K. L.,Iezzoni, L. I., & Banaji, M. R. (2007). Implicit bias among physiciansand its prediction of thrombolysis decisions for black and white patients.Journal of General Internal Medicine, 22, 1231–1238. http://dx.doi.org/10.1007/s11606-007-0258-5

Greenwald, A. G., & Krieger, L. H. (2006). Implicit bias: Scientificfoundations. California Law Review, 94, 945–967. http://dx.doi.org/10.2307/20439056

Greenwald, A. G., Poehlman, T. A., Uhlmann, E. L., & Banaji, M. R.(2009). Understanding and using the Implicit Association Test: III.Meta-analysis of predictive validity. Journal of Personality and SocialPsychology, 97, 17–41. http://dx.doi.org/10.1037/a0015575

Haider, H., & Frensch, P. A. (1996). The role of information reduction inskill acquisition. Cognitive Psychology, 30, 304–337. http://dx.doi.org/10.1006/cogp.1996.0009

Heilman, M. E. (2012). Gender stereotypes and workplace bias. Researchin Organizational Behavior, 32, 113–135. http://dx.doi.org/10.1016/j.riob.2012.11.003

Kutner, M. H., Nachtsheim, C. J., Neter, J., & Li, W. (2005). Applied linearstatistical models. New York, NY: McGraw-Hill International.

Landy, F. J. (2008). Stereotypes, bias, and personnel decisions: Strange andstranger. Industrial and Organizational Psychology: Perspectives onScience and Practice, 1, 379–392. http://dx.doi.org/10.1111/j.1754-9434.2008.00071.x

Lerner, J. S., & Tetlock, P. E. (1999). Accounting for the effects ofaccountability. Psychological Bulletin, 125, 255–275. http://dx.doi.org/10.1037/0033-2909.125.2.255

Levinson, J. D., & Smith, R. J. (Eds.). (2012). Implicit racial bias acrossthe law. New York, NY: Cambridge University Press. http://dx.doi.org/10.1017/CBO9780511820595

Levinson, J. D., & Young, D. (2010). Implicit gender bias in the legalprofession: An empirical study. Duke Journal of Gender Law & Policy,18, 1.

Lyness, K. S., & Heilman, M. E. (2006). When fit is fundamental: Perfor-mance evaluations and promotions of upper-level female and malemanagers. Journal of Applied Psychology, 91, 777–785. http://dx.doi.org/10.1037/0021-9010.91.4.777

Masser, B. M., & Abrams, D. (2004). Reinforcing the glass ceiling: Theconsequences of hostile sexism for female managerial candidates. SexRoles, 51, 609–615. http://dx.doi.org/10.1007/s11199-004-5470-8

McKelvey, R. D., & Zavoina, W. (1975). A statistical model for theanalysis of ordinal level dependent variables. The Journal of Mathemat-ical Sociology, 4, 103–120. http://dx.doi.org/10.1080/0022250X.1975.9989847

Minnesota State Courts. (2014). Minnesota Statewide ADR-Rule 114Neutrals Roster. State Court Administrator’s Office. Retrieved fromhttp://www.mncourts.gov/apps/adr/adr_query.asp

Mitchell, G. (2012). Revisiting truth or triviality: The external validity ofresearch in the psychological laboratory. Perspectives on PsychologicalScience, 7, 109–117. http://dx.doi.org/10.1177/1745691611432343

O’Brien, R. M. (2007). A caution regarding rules of thumb for varianceinflation factors. Quality & Quantity: International Journal of Method-ology, 41, 673–690. http://dx.doi.org/10.1007/s11135-006-9018-6

Olian, J. D., Schwab, D. P., & Haberfeld, Y. (1988). The impact ofapplicant gender compared to qualifications on hiring recommendations:A metaanalysis of experimental studies. Organizational Behavior andHuman Decision Processes, 41, 180–195. http://dx.doi.org/10.1016/0749-5978(88)90025-8

Oswald, F. L., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. E.(2013). Predicting ethnic and racial discrimination: A meta-analysis ofIAT criterion studies. Journal of Personality and Social Psychology,105, 171–192. http://dx.doi.org/10.1037/a0032734

Payne, B. K., Krosnick, J. A., Pasek, J., Lelkes, Y., Akhtar, O., &Tompson, T. (2010). Implicit and explicit prejudice in the 2008 Amer-ican presidential election. Journal of Experimental Social Psychology,46, 367–374. http://dx.doi.org/10.1016/j.jesp.2009.11.001

Pendry, L. F., & Macrae, C. N. (1994). Stereotypes and mental life: Thecase of the motivated but thwarted tactician. Journal of ExperimentalSocial Psychology, 30, 303–325. http://dx.doi.org/10.1006/jesp.1994.1015

Ponak, A. (1987). Discharge arbitration and reinstatement in the provinceof Alberta. The Arbitration Journal, 42, 40–46.

Rachlinski, J. J., Johnson, S. L., Wistrich, A. J., & Guthrie, C. (2008). Doesunconscious racial bias affect trial judges. The Notre Dame Law Review,84, 1195.

Richeson, J. A., & Ambady, N. (2001). Who’s in charge? Effects ofsituational roles on automatic gender bias. Sex Roles, 44, 493–512.http://dx.doi.org/10.1023/A:1012242123824

Rooth, D. (2010). Automatic associations and discrimination in hiring:Real world evidence. Labour Economics, 17, 523–534. http://dx.doi.org/10.1016/j.labeco.2009.04.005

Rudman, L. A., & Ashmore, R. D. (2007). Discrimination and the implicitassociation test. Group Processes & Intergroup Relations, 10, 359–372.http://dx.doi.org/10.1177/1368430207078696

Rudman, L. A., Glick, P., & Phelan, J. E. (2007). From the laboratory tothe bench: Gender stereotyping research in the courtroom. In E. Borgida& S. T. Fiske (Eds.), Beyond common sense: Psychological science inthe courtroom. Malden, MA: Blackwell. http://dx.doi.org/10.1002/9780470696422.ch5

Rudman, L. A., & Goodwin, S. A. (2004). Gender differences in automaticin-group bias: Why do women like women more than men like men?Journal of Personality and Social Psychology, 87, 494–509. http://dx.doi.org/10.1037/0022-3514.87.4.494

Rudman, L. A., & Kilianski, S. E. (2000). Implicit and explicit attitudestoward female authority. Personality and Social Psychology Bulletin, 26,1315–1328. http://dx.doi.org/10.1177/0146167200263001

Sabin, J. A., & Greenwald, A. G. (2012). The influence of implicit bias ontreatment recommendations for 4 common pediatric conditions: Pain,urinary tract infection, attention deficit hyperactivity disorder, and asth-ma. American Journal of Public Health, 102, 988–995. http://dx.doi.org/10.2105/AJPH.2011.300621

Sakalli-Ugurlu, N., & Beydogan, B. (2002). Turkish college students’ attitudestoward women managers: The effects of patriarchy, sexism, and gender differ-ences. The Journal of Psychology: Interdisciplinary and Applied, 136, 647–656. http://dx.doi.org/10.1080/00223980209604825

Simpson, P. A., & Martocchio, J. J. (1997). The influence of work historyfactors on arbitration outcomes. Industrial & Labor Relations Review,50, 252. http://dx.doi.org/10.2307/2525085

Souchon, N., Cabagno, G., Traclet, A., Dosseville, F., Livingstone, A.,Jones, M., & Maio, G. (2010). Referees’ decision-making and playergender: The moderating role of the type of situation. Journal ofApplied Sport Psychology, 22, 1–16. http://dx.doi.org/10.1080/10413200903250476

Sterne, J. A., White, I. R., Carlin, J. B., Spratt, M., Royston, P., Kenward,M. G., . . . Carpenter, J. R. (2009). Multiple imputation for missing datain epidemiological and clinical research: Potential and pitfalls. BritishMedical Journal (Clinical Research Ed.), 338, b2393. http://dx.doi.org/10.1136/bmj.b2393

Sternlight, J. R. (2005). Creeping mandatory arbitration: Is it just? StanfordLaw Review, 57, 1631–1675.

Suedfeld, P. (2005). Welcome to the Electronic Complexity Workshop.Retrieved from http://www2.psych.ubc.ca/~psuedfeld/Workshop.html

Swim, J., Borgida, E., Maruyama, G., & Myers, D. G. (1989). Joan McKayversus John McKay: Do gender stereotypes bias evaluations? Psycho-logical Bulletin, 105, 409–429. http://dx.doi.org/10.1037/0033-2909.105.3.409

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

12 GIRVAN, DEASON, AND BORGIDA

Tetlock, P. E., & Lerner, J. S. (1999). The social contingency model:Identifying empirical and normative boundary conditions on the error-and-bias portrait of human nature. In S. Chaiken & Y. Trope (Eds.),Dual-process theories in social psychology (pp. 571–585). New York,NY: Gilford Press.

Tetlock, P. E., & Mitchell, G. (2009). Implicit bias and accountabilitysystems: What must organizations do to prevent discrimination? Re-search in Organizational Behavior, 29, 3–38. http://dx.doi.org/10.1016/j.riob.2009.10.002

Tetlock, P. E., Skitka, L., & Boettger, R. (1989). Social and cognitivestrategies for coping with accountability: Conformity, complexity, andbolstering. Journal of Personality and Social Psychology, 57, 632–640.http://dx.doi.org/10.1037/0022-3514.57.4.632

Thomson Reuters. (2012). Westlaw classic. New York, NY. Retrievedfrom http://www.westlaw.com

Tosi, H. L., & Einbender, S. W. (1985). The effects of the type and amountof information in sex discrimination research: A meta-analysis. Academyof Management Journal, 28, 712–723. http://dx.doi.org/10.2307/256127

U.S. Department of Labor, Bureau of Labor Statistics. (2012). Householddata annual averages: 18. Employed persons by detailed industry, sex,race, and Hispanic or Latino ethnicity. Retrieved from http://www.bls.gov/cps/cpsaat18.pdf

Van den Bergh, L., Denessen, E., Hornstra, L., Voeten, M., & Holland,R. W. (2010). The implicit prejudiced attitudes of teachers relations toteacher expectations and the ethnic achievement gap. American Educa-tional Research Journal, 47, 497–527. http://dx.doi.org/10.3102/0002831209353594

Viki, G. T., & Abrams, D. (2002). But she was unfaithful: Benevolentsexism and reactions to rape victims who violate traditional gender role

expectations. Sex Roles, 47, 289 –293. http://dx.doi.org/10.1023/A:1021342912248

Viki, G. T., Massey, K., & Masser, B. (2005). When chivalry backfires:Benevolent sexism and attitudes toward Myra Hindley. Legal and Crim-inological Psychology, 10, 109 –120. http://dx.doi.org/10.1348/135532504X15277

Weiss, D. J., & Shanteau, J. (2003). Empirical assessment of expertise.Human Factors, 45, 104–116. http://dx.doi.org/10.1518/hfes.45.1.104.27233

Wiener, R. L., & Hurt, L. E. (2000). How do people evaluate social sexualconduct at work? A psycholegal model. Journal of Applied Psychology,85, 75–85. http://dx.doi.org/10.1037/0021-9010.85.1.75

Wiener, R. L., Hurt, L., Russell, B., Mannen, K., & Gasper, C. (1997).Perceptions of sexual harassment: The effects of gender, legal standard,and ambivalent sexism. Law and Human Behavior, 21, 71–93. http://dx.doi.org/10.1023/A:1024818110678

Wood, W., & Eagly, A. H. (2010). Gender. In S. T. Fiske, D. T. Gilbert,& G. Lindzey (Eds.), Handbook of social psychology (5th ed., Vol. 1, pp.629 – 667). New York, NY: Wiley. http://dx.doi.org/10.1002/9780470561119.socpsy001017

Yzerbyt, V. Y., Schadron, G., Leyens, J. P., & Rocher, S. (1994). Socialjudgeability: The impact of meta-informational cues on the use ofstereotypes. Journal of Personality and Social Psychology, 66, 48–55.http://dx.doi.org/10.1037/0022-3514.66.1.48

Received August 4, 2014Revision received April 21, 2015

Accepted April 25, 2015 �

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

13GENERALIZABILITY OF GENDER BIAS