Representative Bureaucracy and Organizational Justice in Mediation

Abstract

Studies of representative bureaucracy (RB) argue public organizations reflective of the public they serve exhibit better outcomes, especially when serving underrepresented groups. RB theory attributes improved outcomes either to the actions representative bureaucrats take (active representation), or a greater perception of trust and legitimacy toward them by service recipients (symbolic representation), largely treating active and symbolic representation as separate phenomena. We explore the intricate relationship between bureaucracies and the populations they serve by observing the cross-influence between active and symbolic representation, as revealed by self-reported outcomes in discrimination complaints (N = 1,372) referred for voluntary mediation in the United States Postal Service, the REDRESS© program, a context in which mediators are highly limited in representing a claimant’s interests given the requirement of impartiality. In exit surveys measuring employee perceptions of organizational justice, we observed the impact of race and gender representation by gauging changes in reported satisfaction when a mediator’s race or gender matched the nature of the complaint in cases of race or sex discrimination and sexual harassment, via multivariate regression estimation. These analyses support RB theory regarding sexual harassment complaints, where complainants rated outcomes significantly more favorably for female mediators. We found a negative correlation between female mediators and sex discrimination complaints, as well as African American mediators and race discrimination complainants. To explain this discrepancy, we argue that interactions between symbolic and active representation determine the expectations and perceptions placed on bureaucrats. When a bureaucrat does not meet those expectations, service recipients tend to have a more negative view of organizational justice outcomes.

Introduction

Representative bureaucracy (RB) theory holds that higher levels of representation of the public’s characteristics—including race, sex, and other characteristics—among bureaucrats in a public organization will lead to higher levels of desired outcomes (Kennedy 2014; Kingsley 1944; Krislov 2013; Kenneth John Meier 1975; 2019; Mosher 1982; see also Bishu and Kennedy 2019 for a meta-analytic review). Theory predicts that the identification bureaucrats have with those of similar demographic characteristics (Mosher 1982; Meier 2019), and that service recipients have with bureaucrats (Theobald and Haider-Markel 2008), will lead to favorable outcomes for that group and its members, as well as better agency performance (Fernandez 2020; Fernandez, Koma, and Lee 2018). This performance is a manifestation, RB theory suggests, of bureaucrats’ shared characteristics translating to outcomes that benefit those being represented through active representation (AR), symbolic representation (SR), other mechanisms (e.g., nonrepresenting bureaucrats becoming socialized to plight of minority groups by associating with representing bureaucrats within the organization), or an interaction of these elements (Meier 2019; Riccucci and Van Ryzin 2017; see also Abner, Kim, and Perry 2017). The dynamics between these mechanisms have yet to be fully defined.

Researchers have employed a variety of methods for measuring represented groups’ outcomes. In employment, perceptions of procedural (Lind and Tyler 1988) and organizational justice (Colquitt et al. 2001) correlate with performance outcomes, particularly in the arena of workplace diversity (Choi 2009; Choi and Rainey 2010; Page 2008; Wise and Tschirhart 2000). To incorporate this knowledge into practice, organizations designed dispute resolution systems to address workplace conflict (Amsler, Avtgis, and Jackman 2017; Amsler and Sherrod 2017). Under Equal Employment Opportunity Commission (EEOC) rules (https://www.eeoc.gov/federal/adr/federal-adr.cfm; 29 C.F.R. § 1614.603), federal agencies and organizations address workplace discrimination through voluntary alternative dispute resolution (ADR) programs.

These dispute resolution systems present an opportunity for measuring outcomes associated with representation, especially mediation. Not only does mediation offer a forum in which employees can hold employers accountable, but also provides a platform to influence perceptions of justice (Lind and Tyler 1988). The US Postal Service’s (USPS) implementation of the workplace mediation program REDRESS© embodies one such effort. Following initial pilot projects (1994-1997), the USPS conducted a complete national rollout over 18 months from 1998 to 1999; the program continues to provide outside, neutral (i.e., impartial) mediation for employee EEO complaints (https://about.usps.com/what-we-are-doing/redress/about.htm). One of the effects of REDRESS© (and mediation processes in general) is a stronger sense of procedural and organizational justice; both employee and supervisor satisfaction with the process, mediator, and outcome are logical and common validated measures of mediation program success (Amsler, Avtgis, and Jackman 2017; Bingham 1997; Bingham et al. 2009).

The USPS designed REDRESS© to meet organizational goals (Nabatchi and Bingham 2010), including employee concerns (e.g., organizational justice to improve workplace relationships). REDRESS© amounts to a one-party control design in that the USPS had “the power to make choices regarding what cases are subject to the process, which process, or sequence of processes are available…what due process rules apply, and other structural aspects of a private justice system” (Bingham et al. 2009, 4). Thus, its mediators meet classical definitions of bureaucrats (e.g., Weber 1947/1997). While some ambiguity exists in the definition of “bureaucrat,” mediators are representatives of a process within the USPS institutional structures.¹ REDRESS© provides a field experiment for research on RB.

RB in the context of ADR also remains an open question. Does the demographic representativeness of a mediator influence perceptions of organizational justice in terms of satisfaction with outcome? That is, how does satisfaction change if an employee who alleges workplace discrimination participates in mediation with a demographically matched mediator? Do employees’ sexual harassment and sex discrimination claims end with a higher or lower level of satisfaction when the mediator is female? Do employees with claims of race discrimination have differing outcome satisfaction with African American mediators?

Mediation provides a unique opportunity to examine SR, which is generally difficult to detect in situations where AR is possible. In mediation, AR is much less likely because ethical codes mandate impartiality, which may eclipse the necessary condition of discretion and dampen the salience of group identity (Keiser et al. 2002; Kennedy 2014; Selden 1997; Sowa and Selden 2003). For the claimant, however, salience of group identity is evident in the very nature of their claim (e.g., race discrimination), and they may perceive that a mediator has discretion to act in their interest. Since mediation, especially transformative mediation, functions best when participants trust the process and perceive it as legitimate (Bingham et al. 2009), this question is important.

To the best of our knowledge, no research to date has addressed how RB theory applies to dispute resolution, particularly as to cases entailing discrimination claims; nor has any literature explored SR in a context that limits AR to this extent. We explore the effect demographic representation of mediators has on employee complainants’ perceptions of procedural and distributive justice in the context of the REDRESS© workplace mediation program. We analyze complainants’ exit surveys on perceptions of procedural and distributive justice for relative fluctuations observed when a mediator’s gender or race matched the nature of the complaint or purview as sex discrimination, sexual harassment, or race discrimination. Within this context we explore the effects SR has on outcomes in the absence of AR, investigating the dynamic relationship between these two elements by controlling for the latter.

Representative Bureaucracy

RB theory generally falls into three categories: passive, active, and symbolic. Passive representation (PR), sometimes referred to as descriptive representation (Kennedy 2014; Pitkin 1967), refers to relative proportionality of demographic groups between an organization and constituents, or a demographic match between a bureaucrat and service recipient (Meier 1975; Riccucci and Van Ryzin 2017). Research has shown that representation not only increases social equity, embodies democratic values, and leads to affirmative policy outcomes for underrepresented groups, but also positively affects performance (Choi and Rainey 2010; Page 2008).

The other two other primary perspectives on RB, active representation (AR) and symbolic representation (SR), constitute causal mechanisms that produce outcomes (Marinin and Singer 1988; Nicholson-Crotty et al. 2016). Riccucci and Van Ryzin (2017, 25) proposed a model with two separate causal channels through which PR leads to outcomes. Abner, Kim, and Perry (2017, 148) built upon this model adding organizational and individual complexities. Irrespective of proposed models, nearly all studies regarding PR leading to outcomes have occurred in contexts difficult, if not impossible, to discern which causal mechanism led to measured outcomes (see Theobald and Haider-Markel 2008).

Studies spanning four decades found a link between PR and desired outcomes, from Meier’s (1975) groundbreaking work to the more recent work by Riccucci, Van Ryzin, and Jackson (2018) and Johnston and Holt (2019). As organizations begin to resemble those they serve, previously underrepresented groups within the organization effect change benefitting their demographic group (Wilkins and Keiser 2006). Research has found representation influences an entire organization by minimizing biases and increasing awareness of a specific social group; even actors within the organization that do not belong to that group begin to manifest benefits toward them (Hong 2016; Lim 2006). Additionally, Riccucci and Van Ryzin (2017, 24) indicated that better representation, “translates into better treatment of not only minorities but of all citizens” (citing Hong 2016).

Active representation refers to decisions and actions taken by representing bureaucrats (Mosher 1982; Riccucci and Van Ryzin 2017). Theory predicts that the identification bureaucrats have with those of similar demographic characteristics will inform their beliefs and attitudes, which lead to actions to benefit the identifying group (Meier and Nigro 1976; Meier 2019). Research has shown this is truer for street-level bureaucrats than the hierarchy (Andrews, Ashworth, and Meier 2014).

Meier and Nigro (1976) put forth four sequential factors that link PR to AR: social origins, socialization experiences, attitudes, and behaviors. Though their initial research showed a tendency for representatives to become socialized to the organization, rather than actively represent their social-origin group’s interests, subsequent studies found significant evidence that social origins impact individual bureaucrats’ actions (e.g., Dolan 2002; Rosenbloom and Featherstonhaugh 1977; Meier and Stewart 1992).

SR describes service recipient perceptions of representation leading to higher levels of trust in institutions, bureaucrats within them, and coproduction (Banducci, Donovan, and Karp 2004; Bobo and Gilliam 1990; Gay 2002; Tate 2004; Theobald and Haider-Markel 2008; see also Pitkin 1967). Rather than an organization’s actions that benefit specific groups, “symbolic representation works cognitively on the audience of those who belong to a group that is to be represented” (Theobald and Haider-Markel 2008, 410). Several studies have linked PR to SR and downstream outcomes, though few studies have provided a convincing path from PR solely through SR to outcomes.

Thielemann and Stewart (1996) investigated the preferences of AIDS clinic patients and found significant inclination for demographic and sexual orientation alignment. These preferences can lead to coproduction, improving proximate outcomes. For instance, Ken Meier and Jill Nicholson-Crotty (2006) found female victims of sex crimes more amenable to reporting these crimes, and working with investigators, when dealing with female police officers. They also found police agencies with better female representation manifest higher proportions of arrests for sex crimes, pointing toward interactions between AR and SR. In this endogenous relationship, Meier (2019, 41) predicts representing bureaucrats “are more likely to act for clients who engage in coproduction.”

Extending beyond immutable characteristics like race and gender, Gade and Wilkins (2013) explored veterans’ preferences in a Veterans Administration employment program. They found clients perceived counselors who were also veterans to act more in their interest. The authors also indicated, through qualitative interviews with counselors of the program, that “veteran status of the counselor has the potential to change the nature of the relationship between the client and the counselor, and that Veterans nearly always inquire as to whether their counselor is a Veteran” (277, emphasis added), further evidence of interaction between SR and AR.

In “micro-theory” of RB (individual interactions, rather than aggregate measures; Meier 2019, 48), which our analysis focuses on, there are contextual factors and necessary conditions that affect representation and how it translates to outcomes (Keiser et al. 2002; Meier and Stewart 1992). First, representatives must operate within a policy domain that affects the social group in question (Gade and Wilkins 2013; Thompson 1976). The more a group is impacted within the policy domain, such as racial profiling by police, the greater the impacts of representation (Hong 2017). The size, purpose, and position (both hierarchically and geographically) of an organization can also affect representation (Abner, Kim, and Perry 2017; Thompson 1976). Empirical evidence has indicated that such contextual issues, like the amount of segregation in public schools, influence other factors of representation, such as the salience (discussed more below) of race (Roch and Edwards 2017).

Evidence thus far indicates two necessary conditions for PR to translate to outcomes through AR: salience of the demographic identity and discretion of the representative to act on behalf of service recipients (Keiser et al. 2002; Kennedy 2014; Selden 1997; Sowa and Selden 2003). Meier (2019, 40) posits, “The bare bone’s theory of representation holds that the translation of PR into AR is contingent on the salience of the identity in question (race, gender, age, etc.) and the discretion of the bureaucrat that is linked to that identity.”

Salience applies to SR in that a service recipient must consider the demographic identity as important, like the AR context, but SR may include the client’s perception that the identity is also important to the bureaucrat (Theobald and Haider-Markel 2008). The discretion requirement, on the other hand, seems to be one more of perception from the SR perspective. To make a somewhat converse argument presented in Meier (2019), if a service recipient assumes that a demographically matched bureaucrat has the authority to act on their behalf, they may have more trust in the process and, thus, be more likely to actively engage. A problem may arise, however, if the bureaucrat does not (or cannot) act on the client’s behalf—breaking the perceived trust.

This notion of perception between parties affecting coproduction in public organizations manifest in recent work by Lucila M. Zamboni (2019). The author explored decision-making processes of emergency responders, examining Meier’s (2019) hypotheses regarding bureaucrats’ willingness to use discretion on behalf of clients. Bureaucrats tend to use discretion based upon perceptions of whether clients deserve to be actively represented: “Decisions to actively represent clients are triggered by the responder’s assessment of the client’s deservedness of service, based on whether the responder considers the client to be properly using or misusing the 911 emergency service” (Zamboni 2019, 6). It appears that not only are clients influenced by their perceptions of a bureaucrat, but the actions of service recipients also influence bureaucrats.

Considering RB research thus far, and the gap regarding dynamics between AR and SR, we propose a simple model (figure 1). Service recipients perceive representation (1) and, having an improved sense of trust and legitimacy, participate in coproduction (2), which increases the likelihood that bureaucrats will perform AR, which clients notice (3), and ultimately leads to improved policy outcomes (PO) (4). The cycle between AR and SR (2 and 3) can continue in various stages, especially in cases of multiple contacts to achieve specific outcomes or may happen briefly. RB literature has thus far failed to test this interaction or convincingly isolate the independent effects of either channel.

Figure 1.

Theoretical Relationships in Representative Bureaucracy.Source. Adapted from Abner, et al. (2017) and Riccucci and Van Ryzin (2017).

We examine complainant perceptions of organizational justice in the context of equal employment opportunity (EEO) cases entailing claims of sex discrimination, sexual harassment, and race discrimination (also called purviews under Title VII of the Civil Rights Act of 1964); logically, in these cases the identity of complainants is highly salient because it is legally relevant to the purview of the claim. The bureaucrats in question, however, are independent contractors serving as third party impartial mediators in an employment dispute resolution program. Mediation is a process in which a third party, defined as neutral or impartial, aids the disputants in negotiating a resolution to their dispute (Senger 2004; Wall and Dunne 2012). This context limits the discretion of the mediator as bureaucrat.

Gender Representation

Representation research has shown that gender matters. Male and female bureaucrats experience socialization differently, and have different attitudes toward issues salient to women (Hale and Branch 1992). Female representation leads to better results in relevant policy areas such as child support enforcement (Wilkins and Keiser 2006). Guul (2018) found that gender matched counselors in an employment program led to better outcomes, postulating that it may be due to more effort by clients in the matched condition. Work by Fernandez, Malatesta, and Smith (2013), though, alluded to the complex nature of representation. While they found racial/ethnic minority representation increased public contracts with minority-owned businesses, their work discovered no such impact for women-owned firms, indicating a possible “queen bee syndrome” (113, citing Staines, Tavris, and Jayaratne 1974).

Riccucci, Van Ryzin, and Lavena (2014) offer a telling look into the effects SR can have on legitimacy and personal interactions of women with police. Their study examined the effects of varying levels of female representation in a hypothetical domestic violence unit have on perceptions of trustworthiness, fairness, and job performance. The online experiment indicated that not only women, but all of their subjects scored well-represented (6 female, 4 male) units higher than underrepresented (1 female, 9 male). The authors replicated results concerning a recycling program (Riccucci, Van Ryzin, and Li 2016), but were unable to do so regarding emergency preparedness (Van Ryzin, Riccucci, and Li 2017), highlighting the importance of policy area relevance mentioned above.

Meier and Nicholson-Crotty (2006) not only found women more likely to report sexual assault to representative police agencies, and that these more representative agencies produced more arrests, but that male and female police officers differed significantly in attitudes toward sexual assault (largely an issue affecting women). Following a comprehensive review of the history of male-dominated policing’s dismal record with sexual assault in the US (“Manipulating statistics, discouraging individuals from reporting sexual assaults, and failing to pursue cases that are brought to them,” 855), the authors present qualitative data showing how different perceptions can be. Male detectives “are always mindful of the likelihood that the victim is lying” (854), while women officers focused on the importance of gathering evidence as quickly and thoroughly as possible. Notably, this work points implicitly to the importance of perceptions between bureaucrats and clients. Police officers make judgements on whether to actively represent the interests of clients based upon whether they consider those interests to be legitimate, however misguided such perceptions may be.

These studies provide convincing evidence of female representation increasing women’s willingness to coproduce public goods. The current analysis happens within the context of mediation, so this notion of coproduction is consistent with the desired outcomes of mediation; participants in mediation are considered an integral part of the process; hence, they are coproducers. This literature provides some indication, as well, that the more immediately relevant an issue is, the more impactful representation.

Definitions within RB theory, especially those regarding relevance of policy domain and salience of demographic identity, fit within the context of sex discrimination and sexual harassment claims. Modern theory on sexual harassment characterizes the phenomenon as more than simply a sexual, male-domination act, but as a manifestation of “women’s systematic disadvantage…as creative, committed workers” (Schultz 1998, 1690). The very organization a woman confronts when attempting to pursue a sexual harassment claim represents, in some part, the context that made the indignity possible. Furthermore, much of the structure serves as “a means to protect hegemonic masculine work status and identity” (Schultz 2018, 46). Sexual harassment claimants, then, have good reason to prefer mediation facilitated by women.

Race/Ethnicity Representation

Like gender representation, RB research has found race and ethnicity important factors for public organizations. Theobald and Haider-Markel (2008) presented a study into citizen perceptions of legitimacy for law enforcement agencies improving with higher levels of PR. This did not require any specific actions by bureaucrats but indicated better policy outcomes via more citizen participation. Interestingly, they point out that people’s “perceptions of situations have real importance even when perceptions might be wrong. In a very real sense, an individual’s perception is his or her reality” (411, emphasis added).

Riccucci, Van Ryzin, and Jackson (2018) performed a survey experiment exploring perceived legitimacy of police actions given various configurations of representation. They found African American citizens perceive the actions taken by African American officers to be more legitimate. The perception of legitimacy held even when the actions presented amounted to misconduct.

This complexity may, in part, come from the endogenous relationship between SR and AR. Theobald and Haider-Markel (2008) offer insight into this idea. Citing Claudine Gay (2002), they argue that a service recipient “simply needs to believe that race is an indicator of values or experiences and that these values and experiences influence representatives’ actions” (412, emphasis added). That is, an expectation of AR comes from the very presence of a representative (SR). Scholarship investigating African American school superintendents shows that these bureaucrats not only hold a belief that they must “identify with Black-directed endeavors to resolve the needs of Blacks in a racist society” (Scott 1990, 168), but also that constituents expect them to actively represent their interests (Mann 1974; 1976).

Though some research has explored the endogeneity between SR and AR (e.g., Theobald and Haider-Markel 2008; Meier and Nicholson-Crotty 2006), none presents context in which one of these can be held constant. The current paper presents such a context. As mediators, by definition, are required to be impartial as to all claimants, they are much less likely to perform AR. This provides an opportunity to evaluate if SR acts independently or in a relationship with AR.

Administrative Dispute Resolution Context and RB

RB theory and research emphasize the importance of context in the translation of representation to positive policy outcomes. In 1990, Congress passed the Administrative Dispute Resolution Act, an amendment to the Administrative Procedure Act made permanent in 1996 (Bingham and Wise 1996; Nabatchi 2007). It authorized federal agencies to use all forms of alternative or administrative dispute resolution (or ADR, including negotiation, mediation, factfinding, ombuds programs, and arbitration); mandated that agencies appoint a dispute resolution specialist; and directed agencies to adopt policies on how they use dispute resolution. Agencies first responded by adopting programs for employment and procurement disputes (Bingham and Wise 1996; Nabatchi 2007), providing mediation to supplement and/or substitute for administrative adjudication.

Under the Alternative Dispute Resolution Act of 1998, Congress directed federal civil trial courts to develop ADR programs; many programs entailed mediation and the use of various designs to encourage settlement (Crowne 2001). Since 1998, there has been tremendous growth in the use of dispute resolution, including both mediation and arbitration, in the state and federal courts (Stipanowich 2004; Stipanowich and Lamare 2014; Wissler 2004). States also adopted mediation in administrative adjudications. Despite the size and national scope of these programs, relatively little public administration scholarship addresses them.

Mediation usually entails identifying issues, using problem-solving communication techniques, and caucusing with parties in confidential settings. ADR and mediation within the government workplace amount to accountability forums for public officials (Amsler and Sherrod 2017). Thus, notions of representativeness are germane, especially as they relate to a workplace mediation program like REDRESS©, which aims to improve organizational justice.

An important distinction for the mediation context is the necessity of impartiality. Mediators, like judges and arbitrators, “when effectively acting as public officials in accountability forums… should be impartial and maintain fair process” (Amsler, Avtgis, and Jackman 2017, 927). The context and realities of ADR practice illustrate the role of discretion in the translation of PR to AR. An ethical code approved by multiple professional associations (including the American Bar Association and Association of Conflict Resolution) mandates that mediators be impartial. The Model Standards of Ethics for Conduct for Mediators (2005) provides the following:

STANDARD II. IMPARTIALITY

1. A mediator shall decline a mediation if the mediator cannot conduct it in an impartial manner. Impartiality means freedom from favoritism, bias or prejudice.

2. A mediator shall conduct a mediation in an impartial manner and avoid conduct that gives the appearance of partiality.

   1. A mediator should not act with partiality or prejudice based on any participant’s personal characteristics, background, values and beliefs, or performance at a mediation, or any other reason.

Scholars generally construe this definition of impartiality to mean that mediators should not actively advocate for any individual or group (Amsler et al. 2017); in theory, mediators are ethically bound to eschew active representation.

Across the entirety of the federal government, mediators are serving as third parties to assist federal employers and EEO complainants in resolving allegations of employment discrimination. How do mediators’ demographic characteristics and the way they do or do not match those of EEO complainants shape the translation of PR to outcomes? Ashforth and Mael (1989) posit a self-classification based upon demographic characteristics within organizations, that individuals prefer interactions with similar individuals. Additionally, people may hold a higher view of those similar to them (Kossek and Zonia 1993). Thus, discrimination complainants from underrepresented groups (e.g., minorities or women) may have a higher expectation of AR than a mediator has discretion to confer. Moreover, it is conceivable mediators may believe an employee from a certain minority or gender group actually is the subject of active discrimination, and that this requires an intervention to balance power in mediation, that is, AR.

Though little research investigates the relationship between mediator and arbitrator demographic characteristics and outcomes, LaFree and Rack (1996) explored variations in monetary outcomes from court and mediation cases relative to the gender and ethnicities of claimants, respondents, and judges or mediators. They found that much of the variation in adjudicated court outcomes were dependent upon the types of cases; when researchers added case types as controls, variations in outcome related to participant demographics disappeared. However, mediation outcomes differed from court outcomes: white claimants achieved better outcomes with white mediators and minority women received worse outcomes when both mediators were women (using two mediators is called a co-mediation model).

Charkoudian and Wayne (2010) investigated the interaction between mediator and participant gender or race and effects on mediation outcomes. The study focused on community co-mediation programs and considered three conditions for analysis: a demographic match, in which case at least one of the mediators’ characteristic of interest aligned with the participant; a non-match; and other match only, in which case the demographic only matched the opposing party. They analyzed conflict behavior, mediator behavior, and the interaction between mediator and participant gender or mediator and participant racial or ethnic group.

Mediation participants with no gender match were less satisfied with the mediation process. They reported significant findings where the participant and mediator did not match by gender and the mediator’s gender matched that of the other party; participants who attended a mediation with no same-gender mediator present saw the mediator(s) as listening judgmentally and as taking sides. When the mediator’s gender matched only that of the opponent, these perceived bias effects worsened. The authors found that matching participant and mediator by racial or ethnic identity had little or no impact on perceptions of mediation process fairness or mediator ability and fairness.

Finally, the notion of “impartial enough” applies in this context (Geyh 2014). Though some contemporary work dissuades from the idea of a purely impartial arbitrator (including in the mediation arena), the goal of mediation stays the same: to reach a solution that both parties support. The entire REDRESS© process is voluntary, including resolution. This means that both parties have to agree that the matter is resolved, otherwise the matter escalates to the formal EEO process. Hence, any active representation by the mediator would likely push the other party away from resolution. We can therefore conclude that REDRESS mediators had very little discretion and little motivation to use what discretion they might have. Mediators’ mandate of impartiality serves as something of a foil to the fundamental partiality necessary for active representation.

Organizational Justice for Measuring Outcomes of Representation

Philosophers, politicians, and scholars have debated “justice” throughout human history in the context of resolving conflict (Bingham 2008, 9). Societal norms—including prevailing philosophical systems and social constructions—inform what is “just” or “fair” (Colquitt et al. 2001). Academics and researchers have defined distributive justice through consensus (mainly among empirical research to-date) regarding distribution and allocation fairness (Adams 1965; Deutsch 1975; Homans 1974; Leventhal 1976). Procedural justice concerns the fairness of procedures used to determine distribution and allocation (Leventhal 1980; Leventhal, Karuza, and Fry 1980; Lind and Tyler 1988; Thibaut and Walker 1975). Proposed organizational justice models agree on central tenets; employees’ perceptions of fairness within their organization tend to influence specific work-related variables, such as prosocial behaviors and productivity (Colquitt et al. 2001; Moorman 1991; Nabatchi, Bingham, and Good 2007). Most organizational justice literature focuses on two-way interactions between supervisors and employees about organizational decisions and employee perceptions of them. Jason Colquitt and colleagues (2001) proposed a four-factor model for organizational justice: distributive justice, procedural justice, and two classes of interactional justice—interpersonal and informational justice.

Nabatchi, Bingham, and Good (2007), however, recast the theory to reflect the ADR and workplace mediation context. This expanded model acknowledges “the divergent directions of interaction among the parties, the potential multiplicity of agent-referenced outcomes, and the reduced power imbalances among the organizational members” (152). They distinguish between perceptions of a mediator (as a procedural representative) and perceptions of the mediation process itself. Mediation in the workplace is a three-way relationship: complainant, respondent, and the mediator. Nabatchi, et al. (2007) propose a model addressing the multiple interactions unique to the mediation context—between disputants and the mediator, in addition to between each other—by including interpersonal justice dimensions for each. The authors tested this new model using confirmatory factor analysis and found it works well within workplace mediation. More recent work has assessed this six-factor model in state employment mediation contexts (Coggburn et al. 2020). The current study uses the same survey instrument to assess organizational justice as a desired outcome (Amsler, Avtgis, and Jackman 2017; citing Nabatchi, Bingham, and Good 2007).

The USPS Workplace Mediation Program: REDRESS©

The Civil Rights Act of 1991 authorized new compensatory and punitive damages for discrimination claims, jury trials for complainants, and encouraged employers to use a wide variety of ADR processes. By 1994, the USPS faced roughly 28,000 formal EEO complaints annually. The USPS took these to represent both a problem intrinsically and symptomatic of deeper issues. It needed more than settlement; it needed a process to address the root causes of workplace conflict. It determined that ADR for EEOC complaints might not only save resources, but also create a forum for organizational justice through mediation (Bingham et al. 2009). Mediation could improve workplace culture through increased communication and understanding among employees and supervisors (Bingham et al. 2009). The USPS created REDRESS© to provide mediation for complaints of discrimination under Title VII of the Civil Rights Act of 1964, the Age Discrimination in Employment Act of 1967, and the Rehabilitation Act of 1973, which forbid race, national origin, gender, religion, age, disability, as well as sexual harassment and retaliation for raising discrimination claims (Bingham et al. 2009). The USPS implemented the program nationwide between 1998 and 1999, at which time the USPS had over 800,000 employees, making it then the largest employment mediation program in the world (Amsler et al. 2017; Bingham et al. 2009).

The USPS selected the transformative style of mediation (see Folger and Bush 1996) for instrumental reasons. The USPS trained, selected, and paid the mediators. In USPS EEO cases handled through the traditional adversary process of administrative adjudication and litigation, the USPS prevailed in over 95% of all complaints. To avoid the appearance of structural bias, it was important for the mediators not to give an opinion on the merits of the case. This would be permissible in an evaluative mediation model. In the transformative model, however, the mediator focuses on empowering the parties to express their views and assisting the parties in recognizing each other’s perspective. In contrast to evaluative mediators, transformative mediators are ethically prohibited from opining on the merits of a case or attempting to pressure the parties into settlement. The USPS worked with Robert A. Baruch Bush and Joseph Folger to train 100 trainers who fanned out across the country in 1998 to train 3,000 mediators for the initial roster (Hallberlin 2000). Once trained, USPS EEO ADR specialists observed one case for each mediator on the roster to ensure they implemented the transformative model (Bingham and Nabatchi 2001). The USPS had 85 geographic regions. Once the roster was complete and culled to 1,500 mediators, EEO ADR specialists in each region assigned mediators to cases. Given the scale and demand for services, mediator assignment was virtually random.

Data

Data come from two sources that we merged for analysis. We derived claimant satisfaction, type of claim, and other control variables from exit surveys. Mediator surveys provided demographics for specific mediators. The master mediator roster was used to match self-reported mediator demographic information with claimant exit surveys.

Claimant Exit Surveys

Mediators distributed anonymous exit surveys for EEO complainants immediately after they completed the mediation session. Employees mailed completed surveys postage prepaid to Indiana University (see the online Supplementary Appendix for full survey instrument). This eliminated recall error and provided an opportunity to evaluate performance of the program and mediators. These confidential instruments included questions regarding the roles each party played (complainant, supervisor, or a representative of either the complainant or supervisor), their employment position (supervisor, manager, or craft employee), and the type of representation (attorney, union official, coworker, or other such as friend or family member, if any). Surveys also asked participants whether they believed the issue to be fully, partially, or not resolved. Participant (complainant or respondent, employee or supervisor) satisfaction was measured regarding procedure, mediator, and outcome with questions using a five-point Likert-type scale—ranging from very satisfied to very dissatisfied or strongly agree to strongly disagree on a five-point scale. These surveys did not, however, include demographic information for claimants.

Importantly, for a limited time one early version of the survey included a question regarding legal purviews of claims under federal discrimination laws:

What was the nature of the EEO complaint in this case? Circle all that apply, and please specify on the space provided.

1. Race _____

2. Color _____

3. Reverse Discrimination (white) _____

4. National Origin _____

5. Sex Discrimination _____

6. Sexual Harassment _____

7. Disability (physical or mental) _____

8. Religion _____

9. Age _____

10. Retaliation _____

11. Other _____

Exit surveys only included this question from 1997 to 1999, the time period for analysis. These were early versions of the instrument administered before full national roll out of the REDRESS© program. We use purview (the nature of the claim) as a proxy variable to infer a claimant is female for claims of sex discrimination and sexual harassment, and African American for claims of race discrimination. Given the time period, we did not include indicators for non-conforming gender identities which are, therefore, not a part of this study.

A sexual harassment or sex discrimination claimant’s gender may, in fact, be irrelevant. Contemporary sexual harassment and discrimination theory dimensionalizes the concepts as reflections of “maintaining a sense of masculine prerogative and status in and through their work—one that depends on displaying mastery and superiority over women, and denigrates men they do not consider ‘real’ men” (Schultz 2018, 47). The claimant’s gender matters much less than that of the mediator.

Similarly, if some of the race discrimination claimants were not African American, the well-known and well-established discrimination experienced by this population should inform claimants’ perceptions in a similar way. A non-African American person of color pursuing such a claim would likely feel an alignment with an African American mediator’s experience. We can say with confidence that claimants in this category were not white, given the presence of the “Reverse Discrimination (white)” alternate category listed above.

Of the 2,609 indicated purviews the most frequent were for race discrimination (19.2%). Sexual discrimination claims totaled 395 (15.1%) and combining them with sexual harassment claims (3.8%) comes close to those regarding race at 494 (18.9%). Table 1 lists purview distribution.

Table 1.

Nature of Claim or EEOC Purview

Race discrimination claim	501	19.2%
Sexual discrimination claim	395	15.1%
Other	340	13.0%
Disability claim	312	12.0%
Age claim	300	11.5%
Retaliation	284	10.9%
Color claim	189	7.2%
Sexual harassment claim	99	3.8%
Reverse-discrimination claim	80	3.1%
National origin claim	75	2.9%
Religion claim	34	1.3%
Totala	2,609

Note: ^aTotal is more than number of surveys as some claimants marked more than one claim.

Mediator Surveys

Researchers mailed mediators a survey to evaluate use of transformative and directive/evaluative behaviors in mediation sessions (Amsler, Avtgis, and Jackman 2017). Surveys included questions regarding age, race, gender, and education level, providing information necessary for this analysis. Of mediators who reported race or gender (N = 1,372), approximately 46% (628) were female. Most mediators—85% (1,161)—were white. Almost 12% were African American (161); other demographic groups comprised the balance (see table 2).

Table 2.

Mediator Race

Mediator gender	White	African American	Latinx	Asian-American	Native American	Other	Total
Male	633 (85.08%)	80 (10.75%)	24 (3.23%)	5 (0.67%)	0 (0%)	2 (0.27%)	744 (54.23%)
Female	528 (84.08%)	81 (12.9%)	2 (0.32%)	11 (1.75%)	4 (0.64%)	2 (0.032%)	628 (45.77%)
Total	1,161 (84.62%)	161 (11.73%)	26 (1.9%)	16 (1.17%)	4 (0.29%)	4 (0.29%)	1,372 (100%)

Combining Claimant and Mediator Surveys

We paired claimant surveys that included the purview question (i.e., those from 1997 to 1999) and mediator USPS identification number with mediator demographics via the mediator roster, which resulted in a dataset of 1,387 observations.

Dependent Variables

Exit surveys included three indices: One on satisfaction with the mediation process, one on the mediator, and another the outcome. All questions were scored on a one to five, Likert-type scale. Both the outcome and mediator indices had a maximum score of 20 and the Process index a maximum of 35. We highlight the Outcome and Mediator Indices in table 3 as these are our primary focus. For the sake of brevity, we pass over the Process Index as it provided no significant results.

Table 3.

Descriptive Statistics on Organizational Justice Indices

Variable2	Obs	Mean	SD	Min	Max
Outcome Index	1,307	13.991	5.439	4	20
Outcome—Overall	1,307	3.692	1.339	1	5
Outcome—Met Expectations	1,307	3.683	1.345	1	5
Outcome—Control	1,307	3.902	1.191	1	5
Outcome—Improved Relationship	1,307	3.759	1.230	1	5
Mediator Index	1,337	18.371	4.360	4	20
Mediator—Respect	1,337	4.868	0.438	1	5
Mediator—Impartiality	1,337	4.770	0.581	1	5
Mediator—Fairness	1,337	4.795	0.536	1	5
Mediator—Performance	1,337	4.765	0.591	1	5

Independent Variables

To explore these questions, the primary variables of interest are the interaction terms combining specific types of claims and mediator demographics. As all of these are binary, the absence of either condition, claim type or demographic of mediator, causes the interaction term to “switch off” (multiplied by zero).

For mediator race, we only consider whether the mediator was African American. First, race itself is an ambiguous and evolving construct, both socially and in the literature. Latinx was considered a race category during our research time period (1997–99), but today authorities like the US Census cite the label as an ethnicity (see Rodriguez 2000). Second, during this period African Americans were the most well-represented group in the USPS craft worker class, while being underrepresented in higher-level positions (McAllister 1998). We concentrate analysis on the most clearly defined category that also includes the most well-developed literature for the time period.

Women mediated 188 of the 395 sex discrimination claims, and 51 of the 99 sexual harassment claims. Of the 501 race discrimination claims, 64 were mediated by African Americans.

Control Variables

We included variables for previous EEO and REDRESS© experience to control for the influence such prior experiences might have on setting expectations. Also, since the point of mediation has more to do with procedural than distributive justice, and the design of the REDRESS© system bent toward fostering understanding and reconciliation rather than some notion of winners and losers (Bingham 1997; Bingham et al. 2009; Nabatchi, Bingham, and Good 2007), we follow Charkoudian and Wayne (2010) by including a control variable for whether the claimant considered the issue to be resolved or not.³ This provided clear analysis of the role a mediator’s demographic match plays in a claimant’s outcome score indices by isolating the independent effects.

Methods

This study explores two layers of RB. First, we examine is the simple association of matched demographic characteristics and its impact has on organizational justice outcomes: whether a match between mediator and claim purview leads to a higher sense of organizational justice for discrimination claimants. The next, more nuanced layer explores the interaction between SR and AR. Given the expansive body of RB literature showing a positive correlation between demographic representation and outcomes, we would expect an increase in perceptions of justice in the matched condition, especially for underrepresented groups. When considering the proposed endogenous relationship between SR and AR, however, a situation in which service providers cannot perform AR may disrupt the usual positive correlation.

To explore the issue of variance between male and female mediators for sexual harassment and sex discrimination claims, we used the following model:⁴

$${\displaystyle \begin{array}{ll}{\displaystyle y_{im}=}& {\mathrm{\beta }}_0+{\beta }_{1}Clai{m}_i+{\beta }_{3}MedDem{o}_m\\ & +{\beta }_{4}Clai{m}_i\ast MedDem{o}_m++{\beta }_5{X}_{im}+{\epsilon }_{im}\end{array}}$$

y _im = satisfaction index score for claimant i and mediator m; Claim_i = binary variable indicating claim purview claimant i; MedDemo_m = binary variable indicating if mediator m was female for sexual harassment and sex discrimination claims, or African American for race discrimination claims; X_im = vector of control variables.

As our analysis pertains to mediator demographic matches with claim purview, we include both claims relevant to female mediators (sex discrimination and sexual harassment) in the same model.

We use this model to test the following hypotheses, which RB and organizational justice theories imply:

H ₁ : For disputes in the USPS REDRESS© system, the presence of a female mediator will be positively correlated with sexual harassment claimants’ satisfaction.

H ₂ : For disputes in the USPS REDRESS© system, the presence of a female mediator will be positively correlated with sex discrimination claimants’ satisfaction.

H ₃ : For disputes in the USPS REDRESS© system, the presence of an African American mediator will be positively correlated with race discrimination claimants’ satisfaction.

Mediators’ inability to act in claimants’ interest under impartiality rules may disrupt the hypothesized positive correlations. Since this study is among the first to explore the endogenous relationship between SR and AR, we cannot make a precise prediction. Thus, we simply predict that the lack of actions on behalf of claimants by mediators will somehow disrupt the usual positive correlation. We expect to see any effect from the hypothesized relationship (2 and 3 from Figure 1) in the interaction term coefficients (${\beta }_4$).

A note is necessary here ensuring that the use of female and African American mediators, as well as female and African American claimants (inferred from claim purviews) is not construed as analysis of anything particular to women or African Americans as individuals or as groups. This analysis operates under the assumption that any individual or group in the position contemporary women and African Americans find themselves would exhibit similar results.

Results

Overall, the most significant impact on satisfaction with outcome (Outcome Index scores) came from whether the claimant considered the issue resolved. Germane to our hypotheses, though, two of three interaction term coefficients of interest were significant and showed substantial effects.

The results only support one of our hypotheses: Hypothesis 1 regarding the positive correlation between female mediators mediating sexual harassment claims and outcome score indices (table 4). We found no correlation between female mediators and sex discrimination claims regarding overall outcome score indices but did find negative correlations with individual elements of this index.

Table 4.

Female Mediators and Sexual Harassment/Discrimination Claims

Dependent variable: Outcome Index (if entirely filled)
	(1)	(2)	(3)
Female mediator	−0.347 (0.298)	−0.338 (0.317)	−0.353 (0.415)
Sexual harassment claim		−0.633 (0.199)	−1.562* (0.032)
Sexual discrimination claim		−0.12 (0.678)	0.0592 (0.842)
Female mediator × sexual harassment claim			1.957* (0.031)
Female mediator × sexual discrimination claim			−0.418 (0.477)
Resolved	2.706*** (0.000)	2.704*** (0.000)	2.714*** (0.000)
Not resolved	−4.651*** (0.000)	−4.633*** (0.000)	−4.612*** (0.000)
Previous EEO experience	0.197 (0.466)	0.19 (0.483)	0.183 (0.494)
Previous REDRESS experience	−0.0582 (0.841)	−0.0673 (0.817)	−0.0696 (0.813)
Intercept	15.13*** (0.000)	15.21*** (0.000)	15.21*** (0.000)
N	1277	1276	1276
R 2	0.313	0.314	0.316

Note: p-Values in parentheses. Statistically significant values in bold.

*p < .05, **p < .01, ***p < .001.

Sexual harassment claims mediated by women saw an outcome index score increased by 1.957 points, while sex discrimination claims had no significant effect. African Americans mediating race discrimination claims experienced a drop of 1.516 points. All effects came in at a 95% confidence level.⁵

African American mediators received lower scores than their counterparts when mediating race discrimination claims for both the Outcome (table 5) and Mediator (table 6) indices. These mediators received generally higher scores otherwise, with coefficients coming in highly significant with substantial effect sizes—in some cases, higher than whether a claimant felt the issue was resolved. Results concerning the Mediator Index (table 6), however, show much lower R², so explain less of the variation.

Table 5.

African American mediators and race discrimination claims

Dependent variable: Outcome Index (if entirely filled)
	(1)	(2)	(3)
African American mediator	−0.0792 (0.869)	−0.0721 (0.881)	0.539 (0.261)
Race discrimination claim		−0.13 (0.649)	0.0582 (0.848)
African American mediator × race discrimination claim			−1.516* (0.027)
Resolved	2.721*** (0.000)	2.723*** (0.000)	2.733*** (0.000)
Not resolved	−4.643*** (0.000)	−4.636*** (0.000)	−4.616*** (0.000)
Previous EEO experience	0.192 (0.478)	0.191 (0.481)	0.195 (0.473)
Previous REDRESS experience	−0.0887 (0.766)	−0.087 (0.771)	−0.0888 (0.766)
Intercept	15.00*** (0.000)	15.04*** (0.000)	14.96*** (0.000)
N	1,277	1,276	1,276
R 2	0.312	0.312	0.314

Note: p-Values in parentheses. Statistically significant values in bold.

*p < .05, **p < .01, ***p < .001.

Table 6.

African American Mediators and Race Discrimination Claims

DV: Mediator Index (if entirely filled)
	(1)	(2)	(3)
African American mediator	0.804** (0.002)	0.782** (0.002)	1.222*** (0.000)
Race discrimination claim		0.347 (0.176)	0.482 (0.095)
African American mediator × race discrimination claim			−1.089* (0.014)
Resolved	0.886** (0.003)	0.882** (0.003)	0.889** (0.003)
Not resolved	−0.262 (0.374)	−0.276 (0.340)	−0.261 (0.370)
Previous EEO experience	0.153 (0.449)	0.149 (0.463)	0.152 (0.456)
Previous REDRESS experience	0.298 (0.261)	0.3 (0.256)	0.298 (0.257)
Intercept	18.01*** (0.000)	17.89*** (0.000)	17.84*** (0.000)
N	1,277	1,276	1,276
R 2	0.022	0.024	0.026

Note: p-Values in parentheses. Statistically significant values in bold.

*p < .05, **p < .01, ***p < .001.

Disaggregating elements of the outcome index illuminate the measures’ inner workings. Splitting up the scores for female mediators for sexual harassment and sex discrimination claims indicates that the interaction term for female mediators and sexual harassment claims is only statistically significant in aggregate, though the individual score indicating whether the claimant felt the outcome of their mediation improved the relationship with the other party came the closest to significance and saw the largest effect size. The effect of sexual harassment claims on the outcome index was also driven by the improved relationship question, showing a 9% drop, though it dropped just out of statistical significance. Sexual discrimination claim effects became statistically significant for the overall outcome and control-over-the-outcome questions, with each showing negative correlation.

Performing the same disaggregation on the African American mediator and race discrimination claims question relative to the Outcome Index shows that the negative effect seen in the index is mainly driven by the overall outcome question, though the question regarding whether the outcome met the claimant’s expectations came close to statistical significance (88.9%) and presents a not insignificant factor in the index score.

A similar disaggregation for African Americans mediating race discrimination claims relative to the Mediator Index results in most coefficients of interest dropping out of statistical significance. The models that come closest to this, though, provide more insight into the overall score, with the question about mediator impartiality coming in at 89.8% significance and that regarding mediator fairness at 84.6%.

Discussion

The results for female mediators facilitating sexual harassment claims pose an interesting observation, particularly when compared to sex discrimination claims. Discrimination “involves treating someone (an applicant or employee) unfavorably because of that person’s sex” (EEOC n.d., “Sex-Based Discrimination”), but sexual harassment is more personal, involving “sexual advances, requests for sexual favors, and other verbal or physical conduct of a sexual nature” (EEOC, “Facts About Sexual Harassment”). Thus, a claim filed for sexual harassment may be more emotionally fraught than one for sex discrimination.

Though our analysis time period is 1997–99, this evidence highlights a trend occurring in the contemporary #metoo movement. As victims of sexual harassment and sexual violence perceive a more receptive audience to communicate such personal issues, more and more are coming forward. Similarly, sexual harassment claimants in the REDRESS© system may have felt more empowered to communicate their experiences to other women, as compared to men, as indicated by the higher indices. We can see in the disaggregated index scores in table 7 that the most important component of the index score is whether the claimant felt the outcome improved their relationship with the other party. In other words, they felt that they had been heard, understood, and validated while experiencing an ordeal, which led to a better relationship with the other party.

Table 7.

Female Mediators and Sexual Harassment/Discrimination Claims—Disaggregated Scores

Dependent variable→	Outcome Index	Outcome-Overall	Outcome-Met Expectation	Outcome-Control	Outcome-Improved Relationship
Female mediator	−0.353 (0.415)	−0.043 (0.586)	0.040 (0.669)	−0.062 (0.478)	−0.102 (0.223)
Sexual harassment claim	−1.562* (0.032)	−0.248 (0.163)	−0.211 (0.303)	−0.145 (0.334)	−0.351 (0.057)
Sexual discrimination claim	0.059 (0.0.842)	−0.173* (0.022)	−0.0691 (0.402)	−0.164* (0.045)	−0.112 (0.165)
Female mediator × sexual harassment claim	1.957* (0.031)	0.203 (0.346)	0.125 (0.609)	0.269 (0.190)	0.367 (0.094)
Female mediator × sexual discrimination claim	−0.418 (0.477)	0.010 (0.931)	0.0004 (0.997)	0.134 (0.310)	0.069 (0.562)
N	1,276	1,276	1,276	1,276	1,276
R 2	0.316	0.450	0.368	0.294	0.294

Note: p-Values in parentheses. Statistically significant values in bold.

*p < .05, **p < .01, ***p < .001.

We find this argument particularly compelling since participation in the REDRESS© process is voluntary. It is likely that this volitional component attenuates the effects toward zero, since the most egregious claims would probably not volunteer for an alternative to the formal EEOC process. In this case the interaction between SR and AR may be the least pronounced. Claimants’ perceptions of trust and legitimacy required no affirmative actions be taken by mediators; the mere presence of a woman eased the process and influenced claimants toward coproduction, as evidenced by the disaggregated scores for improving the relationship with the other party. We find no disruption of the correlation between female mediators and claimants’ outcome scores due to mediators’ inability to actively represent a claimant’s interest.

On the other hand, these results may indicate some type of active representation despite mediators’ mandate of neutrality. While this possibility cannot be ignored, we do not find it convincing. As mentioned above, the purpose of transformative mediation is to cultivate better relationships between parties; and the USPS explicitly employed this method to improve overall culture dynamics. Favoring one party over the other run counter to these goals. Further work is needed to determine whether our results indicate mediators are more likely to act in a claimant’s interest when dealing with sexual harassment claims, or if active representation is not necessary.

The results for women mediating sex discrimination claims and African Americans mediating race discrimination claims (table 7), on the other hand, seem to indicate a disconnect between the expectations claimants had of a demographically matched mediator, or the amount of control they would have over the outcome, and the actual results (table 8 and 9). We get a hint of this in the disaggregated scores. For female mediators and sex discrimination claims, claimants indicate less satisfaction with the outcome generally, and their control over that outcome. For African American mediators and race discrimination claims, the overall outcome score had the biggest effect but the question regarding whether the claimant’s expectations were met seems to have played a major role (table 8).

Table 8.

African American Mediators and Race Discrimination Claims—Disaggregated Scores

Dependent variable→	Outcome-Index	Outcome-Overall	Outcome-Met Expectation	Outcome-Control	Outcome-Improved Relationship
African American mediator	0.539 (0.261)	0.130 (0.282)	0.016 (0.887)	−0.142 (0.243)	−0.070 (0.645)
Race discrimination claim	0.058 (0.848)	0.020 (0.780)	−0.053 (0.490)	−0.053 (0.392)	0.040 (0.522)
African American mediator × race discrimination claim	−1.516* (0.027)	−0.464* (0.012)	−0.309 (0.111)	0.041 (0.815)	−0.268 (0.233)
N	1,276	1,276	1,276	1,276	1,276
R 2	0.314	0.449	0.369	0.293	0.293

Note: p-Values in parentheses. Statistically significant values in bold.

*p < .05, **p < .01, ***p < .001.

Table 9.

African American Mediators and Race Discrimination Claims—Disaggregated Scores

Dependent variable→	Mediator-Index	Med-Respect	Med-Impartial	Med-Fairness	Med-Performance
African American mediator	1.222*** (0.000)	0.004 (0.936)	0.070 (0.183)	0.072 (0.129)	0.035 (0.478)
Race discrimination claim	0.482 (0.095)	−0.636* (0.037)	−0.068 (0.075)	−0.057 (0.127)	−0.061 (0.108)
African American mediator × race discrimination claim	−1.089* (0.014)	−0.018 (0.839)	−0.168 (0.102)	−0.146 (0.154)	−0.009 (0.928)
N	1,276	1,276	1,276	1,276	1,276
R 2	0.022	0.028	0.036	0.039	0.055

Note: p-Values in parentheses. Statistically significant values in bold.

*p < .05, **p < .01, ***p < .001.

When considering the usual relationship between RB theories—PR, AR, and SR—the presence of a neutral representative (as mediators are by definition) may be disruptive. The traditional theoretical flow from PR to outcome does not fit well within the context of mediation, as mediators are forbidden from actively pursuing the interests of any specific social group. At the same time, symbolic representation’s constructs of trust and legitimacy may rouse an expectation of certain processes or outcomes for claimants confronting a demographically matched mediator, resulting in lower satisfaction once the expectations are not met. Our analysis confirms an idea Charkoudian and Wayne (2010, 30) presented: “Culture is complex” (see also Keiser et al. 2002 for analysis of the complexity between sex and gender). They point out that individuals not only exist within their race group, but additionally have cultural influences of work and social life that affect perceptions and expectations. Here, the interaction between SR and AR seems to be more pronounced. Sex discrimination claimants indicate the presence of a female mediator may have signaled some control over outcomes that did not materialize, leading to lower satisfaction scores—keeping in mind that this analysis is relative to male mediators. Race discrimination claimants seem to have had more of an expectation from African American mediators to act in their interest. When this did not happen, the outcome did not meet claimants’ expectations.

Taking these results together, and the contrasts between them, we see a complex interplay between AR and SR. The usual mechanisms proposed for improved outcomes via SR are trust and legitimacy (Riccucci and Van Ryzin 2017; Theobald and Haider-Markel 2008), and “psychological satisfaction with government and its services” (Riccucci and Van Ryzin 2017, 26). Expectations associated with complaint purview may moderate this effect.

The impact of mediator neutrality may create a tension between active and symbolic representation. If we infer that sex and race discrimination claimants had some expectation of active representation from a demographically matched mediator, and that mediators did not meet that expectation, then satisfaction, trust, and legitimacy may be compromised. Results altogether point to a nuanced relationship between AR and SR. When the expectations of clients amount to being heard and understood, as in sexual harassment claims above, then satisfaction increases. When the expectation has more to do with positive actions for the client’s benefit, which are not met, then satisfaction decreases.

Conclusion

While substantial work has offered RB as a vehicle for improving social, distributive, and procedural justice, as well as increasing trust and legitimacy, the underlying mechanisms seem to be more complicated than purported. Though it is certainly useful to examine these issues using demographic specifications, our results confirm previous work cautioning against assuming generalized race or gender effects (e.g., Keiser et al. 2002). Our analysis indicates a complex, interactive relationship between AR and SR as causal mechanisms linking PR to outcomes. The new avenue of inquiry merits incorporation into the RB canon and further exploration.

This work shows that representation in mediation presents challenges concerning expectations mediation claimants may have going into the process. Being aware of these challenges, organizations employing mediation may work toward better results by educating participants and attempting to manage expectations on the front end. Furthermore, acknowledging the nuanced differences between claim types allows mediation programs to provide services customized to specific claims, rather than proceeding uniformly. To put it the organizational justice context, ensuring informational justice—at least, along the lines of managing expectations and information about mediation elements (i.e., mediator neutrality)—would likely provide returns along the procedural and distributive justice dimensions (Colquitt et al. 2001).

More research is needed using this context—mediation—looking deeper into the reasons that parties exhibit various levels of satisfaction. Future work using experimental methods similar to work by Norma Riccucci et al. would provide insight into the dynamics between AR and SR. Such a design would allow for evaluating expectations at different phases of the mediation process, as well as provide the opportunity to ask why participants give certain scores.

Further exploration is needed, as well, into comparative effects of representation in mediation between sectors of employment and geographically. As noted in Mary A. Konovsky’s (2000) work, notions of organizational justice vary along several dimensions. Investigating this question in other industries and locations would provide not only more generalizable results, but also might offer further insight into the phenomenon.

Funding

This article was supported in part by Grant R24DA051950 from the National Institute on Drug Abuse, National Institutes of Health (NIH). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Human Subject Research Statement

Data used in this study were collected for other purposes, and all identifying information has been anonymized. Therefore, this study is exempt from IRB review.

Data Availability

The data underlying this article are available in the OpenICPSR repository, at https://www.openicpsr.org/openicpsr/project/152461/version/V1/view.