DRAFT: Comments Welcome

Individual Employment Rights Arbitration in the U.S.:

Actors and Outcomes

August, 2012

Alexander J.S. Colvin and Mark Gough ILR School, Cornell University

Ithaca, NY Email: ajc22@cornell.edu

1

Abstract

This study examines disposition statistics from nine years of employment arbitration

cases administered by the American Arbitration Association (AAA) to investigate the impact

institutional characteristics of a case have on employee win rates and award amounts. We find

these institutional variables have mixed impacts on employee outcomes. Employee win rates are

higher in California and lower in Texas compared to those filed in all other states. Female

arbitrators and members of the National Academy of Arbitrators (NAA) render awards in favor

of employees less often than do male arbitrators and non-NAA members. And former judges

award higher damages, on average, than arbitrators without judicial experience. Finally, this

study provides evidence of a significant repeat employer-arbitrator pair effect; employers that

use the same arbitrator on multiple occasions win more often relative to employers appearing

before an arbitrator for the first time.

2

Introduction

With the decline of collective representation, contemporary employment relations is

increasingly focused on individualized determination of terms and conditions of employment.

Statutory rights, not unions, are becoming the primary check on managerial prerogatives in the

workplace (Piore and Safford, 2006; Colvin, 2012). However, despite the reduction in collective

action and industrial conflict, this individual rights era is not characterized by an absence of

conflict, but rather has seen a growth in conflicts around individual rights in the workplace. The

resolution and outcomes of these individual rights conflicts are central to determining the nature

of the employment relationship in this era.

In this study, we examine the operation and outcomes of a major new institutional

structure for the resolution of individual rights conflicts in the United States - the arbitration of

individual employment rights claims under mandatory employer-promulgated procedures.

Mandatory arbitration procedures allow employers to require all nonunion employees to agree as

a condition of employment to resolution of individual statutory rights claims, such as Title VII of

the Civil Rights Act, through private arbitration, barring recourse to the public court system. The

impetus for the rise of mandatory arbitration of employment disputes was two key Supreme

Court decisions. In the 1991 case Gilmer v. Interstate/Johnson Lane (500 U.S. 20), the majority

of the court held that claims based on employment statutes, the Age Discrimination in

Employment Act in that case, were arbitrable. In a subsequent 2001 case, Circuit City v. Adams,

a majority of the court held that arbitration agreements incorporated into employment contracts

were enforceable under the Federal Arbitration Act. Following these decisions, the possibility of

avoiding employment lawsuits in the courts with their high litigation costs and the danger of

large jury awards has encouraged increasing numbers of employers to adopt mandatory

3

arbitration procedures for their workforces. Although precise figures on the extent of mandatory

arbitration are difficult to come by, estimates suggest a quarter or more of all nonunion

employees in the U.S. are covered by these procedures (Colvin, 2008; Lewin, 2008). This

suggests mandatory arbitration is now a much more widespread institution in employment

relations than union representation, which had shrunk to only 13.0 percent of the workforce by

2011 (BLS, 2012).

How does this new institution operate and what is its impact on employment relations?

The Supreme Court stated in Mitsubishi Motors v. Soler Chrysler-Plymouth, Inc. (1985), and

later reiterated in Gilmer that: “By agreeing to arbitrate a statutory claim, a party does not forgo

the substantive rights afforded by the statute; it only submits to their resolution in an arbitral,

rather than a judicial, forum.” This study investigates and questions this assumption that the

universally-applicable rights afforded by federal statutes are resolved uniformly in the forum of

mandatory arbitration. In particular, we analyze how variation in who is deciding the case in

arbitration affects outcomes and how arbitration outcomes vary regionally across the country.

Theory and Literature Review

Arbitrator Characteristics and Arbitrator Selection Processes

Parties in arbitration directly influence who will arbitrate their case. 1 No equivalent

power is provided for in the judicial system. While some degree of forum shopping occurs,

parties have little sway over the judge selected to preside over their case. If arbitrator

characteristics affect case outcomes, then parties with the resources and institutional knowledge

to capitalize on this information can directly influence their expected remedies through arbitrator 1 Parties can designate an arbitrator in the original agreement, otherwise, the parties are encouraged to mutually select an arbitrator from a list provided by the AAA. See http://www.adr.org/aaa/ShowProperty?nodeId=/UCM/ADRSTG_004362&revision=latestreleased

4

selection. Evoking the adage “there is no right without a remedy,” an agreement to arbitrate may

not be a mere forum substitution; it may be a forced sacrifice of statutory rights.

There is a body of research documenting that parties rely on arbitrator background

characteristics prior to selecting a arbitrator in labor disputes (Bingham and Mesch, 2002;

Bemmels 1990; Thorton and Zirkel, 1990; Lawson, 1981; Bloom and Cavanagh, 1986). As

Dworkin (1974, p. 200) states: “it is well known that companies and unions conduct an extensive

preliminary inquiry into an arbitrator’s background and reported decisions prior to agreeing upon

his selection.” Employment arbitration lacks the countervailing force unions exert in arbitrator

selection, potentially leaving employees at a distinct disadvantage relative to large or repeat

employers that can benefit from background information to select a predictably biased arbitrator.

As a first point of analysis, we can find evidence that superior arbitrator selection by one

of the parties may affect outcomes if the results of arbitral proceedings vary with the

characteristics of the arbitrator. Put alternatively, if we see systematic variation in outcomes by

arbitrator characteristics, then parties can potentially exploit these differences through selection

of arbitrators contingent on these characteristics. A second indication of the impact of arbitrator

selection would be if we see repeat players having a systematic advantage in obtaining more

favorable outcomes in arbitration. We begin by discussing the potential effects of arbitrator

characteristics and then turn to the issue of repeat player effects.

Gender

Researchers from varied fields have acknowledged certain aspects of social behavior,

personalities, and abilities differ between the sexes. Research psychologists have contended that

these differences extend to decision-making, with woman basing decisions on moral concerns

5

with priorities given to social relationships, while male decision-making is dominated by “rights

[and] formal reasoning that uses a universalizable, abstract, and [an] impersonal style (Gilligan,

1982).” And while debates about the origin of these differences continue—whether they are

immutable biological differences resulting from evolution (maximalist perspective) or responses

to societal structures of gendered proscriptions and prescriptions (minimalist perspective)—their

existence is not in dispute (Eagley and Wood, 1999).

Numerous studies have empirically tested what effect, if any, an arbitrator’s gender has

on their decision-making. Zirkel (1983), looking at 369 labor arbitration decision, and Bigoness

and Dubose (1985), using 80 students as mock arbitrators, found no difference between male and

female arbitrator outcomes. Bemmels (1988a) reports male arbitrators treat female grievants

more leniently with no such preference displayed by female arbitrators. Clyde Scott and

Elizabeth Shadoan (1989) tested to see if female arbitrators issue less severe judgments than

male arbitrators, but their empirical findings failed to identify any significant differences

between the genders. In analyzing 177 arbitrators rulings on six hypothetical cases, Thorton and

Zerkil (1990) found that arbitrator gender cannot account for any differences in the awards

granted to grievants. Bemmels (1990) found that women arbitrators issued judgments slightly

more favorable to grievants than male arbitrators in a sample of 2,001 labor arbitrator awards.

Finally, Caudill and Oswald (1993) asked 146 arbitrators to rule on a hypothetical drug-testing

case and found that women arbitrators are harsher than their male counterparts. They found that

women arbitrators were less likely to issue full reinstatements, but also found that gender did not

have any explanatory power in length of suspensions.

From the field of criminology, Steffensmeier and Hebert (1999), using data from 1991-

1993 judicial sentences in Pennsylvania and additional archival data, examined whether a judge's

6

gender affected criminal defendants’ outcomes in court. The authors concluded that female

judges were 10 percent more likely to incarcerate and their sentences were 5 months longer on

average than those of male judges. The authors impressed, however, that in the majority of the

sentencing-related variables analyzed, gender effects were non-existent or negligible.

Corroborating this study, Kulik, Perry and Pepper (2003) fail to find a relationship between a

judge’s gender and case outcomes in a sample of 786 sexual harassment cases decided between

1981 and 1996.

Unlike judges and labor arbitrators, an employment arbitrator wields extraordinary

discretion in case disposition and is unencumbered by precedent, formal sentencing guidelines,

rules of evidence, or detailed collective bargaining agreements. In addition, they are not subject

to the uniform experiences associated with law school and the career trajectory necessary for a

judicial role, which would moderate the influence such personal characteristics like gender

would have on decision making. In the absence of such constraints and socialization, differences

in outcomes based on the decision-maker’s gender should be more pronounced in employment

arbitration relative to the other forums. In addition to possible maximalist difference between the

genders, life experiences of female arbitrators will likely differ in fundamental respects from

those of male arbitrators, leading to gendered differences in case outcomes.

The weak and inconsistent findings in the literature are likely the result of small sample

sizes of female actors. Still today, and to a greater extent historically, the arbitrator and judge

professions were predominantly male. The 1982 labor arbitrator roster at the Federal Mediation

and Conciliation Service (FMCS) comprised 0.33 percent women (47 out of 14,075) (Bemmels,

1990). In 1993, the percentage of women in the federal judiciary was around 12 percent

(American Bar Association, 2006).

7

Professional Background and Status

Does professional background and status affect arbitral decision-making? While

employment arbitration is a relatively recent phenomenon, there is a long-standing system of

labor arbitration used to resolve disputes in unionized workplaces. The cadre of professional

neutral labor arbitrators has won a well-deserved reputation for respect for due process and

fairness in decision-making. Some of these same labor arbitrators are now among the arbitrators

deciding employment cases in mandatory arbitration proceedings. Do employment arbitrators

who also practice labor arbitration differ in their decision-making? We investigate the potential

impact of this type of professional background and status by looking at the effect of arbitrator

membership in the National Academy of Arbitrators (NAA), the leading professional

organization of prominent labor arbitrators.

Membership in the NAA is a prestigious credential desired by labor arbitrators, reflecting

general acceptability by institutions from both management and labor. Membership requires

substantial experience as an impartial arbitrator of labor-management disputes: at least five years

of experience and a minimum of 60 written decisions over a six-year period (National Academy

of Arbitrators, 2012). The induction criteria to the AAA roster of employment arbitrators, by

comparison, specify only that a potential member needs 10 or more years of experience in

employment law (American Arbitration Association, 2012).

The cumulative decisions and professional norms built over the past 75 years have given

labor arbitrators a far-reaching and unique role in industrial relations. A labor arbitrator’s broad

authority is predicated on the view that collective bargaining agreements are imperfect, or at

least incomplete, charters for enforcing the “common law of the shop.” Rather than interpreting

8

collective bargaining agreements with a rigid adherence to the black letter of the text, decisions

are rendered to comport with basic standards of due process and unwritten rules, practices, and

customs. Notions of just cause, progressive discipline, and the allocation of the burden of proof

are not binding rules enumerated in most collective bargaining agreements; rather, they are

routine expressions of discretionary arbitrator power.

Unlike the common law experienced by judges and lawyers, the common law of the shop

is fluid, informal, and not subject to the principle of stare decisis. Employment arbitrators, being

closely associated with norms of the judiciary, will likely adhere to these narrower standards

when deciding cases. In contrast, NAA members acting in an employment arbitrator capacity

will likely continue taking a liberal view of their role and powers, culminating in more

employee-favorable outcomes. Employer fears that labor arbitrators would be too sympathetic to

employees manifested during the constitution of the AAA Employment Disputes Arbitrators

Panel. In 1996, the AAA committee designated to establish the new panel virtually excluded

labor arbitrators from it (Bingham and Mesch, 2002).

Practical support for arbitrator characteristics influencing outcomes can be found in a

2006 Klaas et al policy-capturing study looking at decisions of 140 AAA employment

arbitrators, 82 labor arbitrators from the National Academy of Arbitrators (NAA), and 83 jurors

who had served in employment discrimination cases over the last five years. The study analyzed

participant responses on 32 hypothetical termination cases and concluded that labor arbitrators

are most likely to rule in favor of the employee, followed by jurors, and then employment

arbitrators.

Another aspect of professional background and status that may be associated with

differences in arbitral decision-making is if the arbitrator is a former judge. Increasing numbers

9

of judges have developed arbitration practices after retirement from the Bench. Indeed, the

relatively low judicial salaries and potential to earn substantial arbitration fees has prompted

growing numbers of judges to take early retirement in order to pursue this new career path. This

phenomenon has become so widespread in the state of California that it prompted a warning

from the Chief Justice of the California Supreme Court that it was causing significant shortages

in staffing of the courts in that state.

Our review of the literature failed to produce a study looking directly at the differences

between judge and arbitrator decision making. However, in their policy-capturing study, Klaas et

al (2006) found that former jurors who had decided employment cases in the courts were more

likely to rule in favor of an employee plaintiff than were employment arbitrators reviewing the

same scenarios. At the same time, Clermont and Schwab (2004) analyzing outcomes in federal

court litigation found that judges deciding cases without a jury were also less likely to rule in

favor of employee plaintiffs than were juries. Taken together these studies suggest that both

judges and employment arbitrators are less likely to rule in favor of employee plaintiffs than are

juries. However, they do not indicate between judges and employment arbitrators which are

more likely to rule in favor of employee plaintiffs. Past experience as a judge may produce

differences in decision-making and we will investigate its impact, though past research and

theory do not provide a clear a priori indication of what direction any effect will be in.

Repeat Player Effects

Repeat player concerns derive from a fear that arbitrators, eager to be selected in future

cases, will betray their neutrality and award favorable decisions to repeat players with hopes of

being selected to arbitrate future disputes. With regard to employment arbitration, employers are

10

the repeat players, leaving individual employees, only ever participating in a single case,

potentially at a distinct disadvantage. This is in contrast to labor arbitration, where management

and the union are seen as countervailing institutional forces equally as likely to be involved in

future arbitration selections.

Though there is a clear economic incentive for arbitrators to give biased rulings favoring

employers, doing so would be an obvious breach of ethical and professional codes. The current

body of literature, however, presents evidence that a repeat player effect does exist in the

arbitration environment but is careful to note many potential explanations may account for it

Bingham (1997) found that repeat players received favorable outcomes when compared

to non-repeat players in a study examining 270 consumer arbitration rulings decided under the

auspices of the AAA between 1993 and 1994. Whereas arbitrators awarded damages to an

employee grievant in 71 percent of the cases involving non-repeat employer players, when a

repeat employer was involved, arbitrators awarded damages to an employee grievant in a mere

16 percent of cases, a 55 percentage point difference. These findings were mirrored in a similar

study published a year later (Bingham 1998b).

Far from proving a repeat player bias in arbitration, these early studies merely established

that employers appearing in multiple cases achieved more favorable results in arbitration (Hill,

2003; Sherwyn et. al. 2005). Advantages can accrue to repeat players relative to one-time

participants for legitimate reasons such as those enumerated by Professor Marc Galanter:

“(1) experience leading to changes in how the repeat player structures the next

similar transaction; (2) expertise, economies of scale, and access to specialist

advocates; (3) informal continuing relationships with institutional incumbents; (4)

bargaining reputation and credibility; (5) long-term strategies facilitating risk-

11

taking in appropriate cases; (6) influencing rules through lobbying and other use

of resources; (7) playing for precedent and favorable future rules; (8)

distinguishing between symbolic and actual defeats; (9) investing resources in

getting rules favorable to them implemented.

One-shotters, on the other hand: (1) have more at stake in a given case; (2) are

more risk averse; (3) are more interested in immediate over long-term gain; (4)

are less interested in precedent and favorable rules; (5) are not able to form

continuing relationships with courts or institutional representatives; (6) cannot use

the experience to structure future similar transactions; and (7) have limited access

to specialist advocates”iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii

(Bingham,1999, p. 5-6, citing Galanter, 1975)

To test more directly for a repeat-player bias, researchers have looked at repeat employer-

arbitrator pairs. Focusing analysis on cases where an employer has used a given arbitrator at

least once controls for most of the legitimate advantages expressed above. Bingham (1998)

shows employees lose more frequently when a decision is made involving a repeat employer-

arbitrator pair. In such cases, arbitrators gave no relief 75 percent of the time, which is

significantly higher than the 45 percent of no relief rulings in cases where employers had not

previously used the arbitrator. In the most thorough study to date looking at the repeat-player

bias, Colvin (2011) presents significant differences in win rates and award amounts for repeat

employer-arbitrator pairs compared to other repeat employers. He finds that employees win 18.6

percent of cases involving repeat employers but only 12 percent of cases if it involves a repeat

employer-arbitrator coupling. Mean award amounts are reported as $19,146 for repeat

employers encountering an arbitrator for the first time and $7,451 for repeat employer-arbitrator

12

pairs. The explanatory power of these variables remained when conducting a multivariate

regression analysis.

Overall, existing theory and literature suggest repeat employers and employers involved

in repeat employer-arbitrator pairs experience advantages in arbitration relative to their one-shot

counterparts. Whether these advantages are a result of arbitrator bias or reflect specialized

knowledge is a question that remains to be answered.

Regional Variation

So far our focus has been on how variation in the characteristics and selection of arbitrators

may affect outcomes. Another factor that may affect outcomes is variation in where arbitration

occurs. Patterns of employment relations vary widely across the United States at the regional and

state level. This is seen dramatically in union membership rates, which range from a low of 2.9

percent in North Carolina to a high of 24.1 percent in New York State. There are distinct

regional patterns to this variation with unions tending to be weaker in the Southern, Prairie, and

Mountain states and stronger in the Northeast, Midwest, and on the West coast. Some of this

relates to differences in culture, industrial structure and economic history between these regions.

However they have also become embedded in the institutional structure of labor relations, most

famously through the “right to work” laws that have been enacted in many southern and central

states.

Should we expect similar variation in individual employment rights between regions and

states? States do have the authority to enact employment laws, which can produce variation

depending on the political and economic climate of the state. At the same time, Federal

employment statutes, such as Title VII, FLSA, ERISA, the ADEA, the ADA, and the FMLA,

13

have national coverage and are designed to provide a universal floor of employment rights for

employees across the county. Most famously, the Civil Rights Act of 1964 was explicitly

adopted as Federal legislation to override the segregationist policies and laws of state

governments in the South of that era. Similarly, the Federal Arbitration Act (FAA) has been

interpreted by the Supreme Court as reflecting a strong Federal policy that preempts most efforts

to regulate mandatory arbitration at the state level. Indeed a series of state laws and state court

rulings that would limit mandatory arbitration have been explicitly held to be pre-empted by the

FAA’s liberal federal policy favoring enforcement of arbitration agreements. What is unclear is

whether mandatory arbitration in practice reflects the idea of universal national coverage

suggested by both the FAA’s liberal federal policy and the Federal employment statutes or if

there is in practice more substantial variation at the state level akin to what we observe in labor

relations.

If there is state level variation in mandatory arbitration, where would we expect to see it? The

state that is often considered to provide the most employee favorable employment law

environment is California.2 California is known for having relatively extensive state level

employment laws that supplement the Federal statutes. Its state courts are considered relatively

favorable to employees, with a high rate of employee success in lawsuits and relatively large

damage awards (Oppenheimer, 2003). This may be in part driven by the relatively strong and

well organized plaintiff employment lawyers bar in California. In the specific area of arbitration

law, California courts have been willing to require, to the degree that Federal pre-emption

doctrines allow, that arbitration agreements need to meet certain basic due process standards to

be enforceable. By contrast, most of the Southern states have generally been considered to have

2 Montana would be the other logical candidate for most employee-favorable status as the only state that has statutorily modified the employment-at-will rule.

14

more employer favorable state laws, courts, and legal environments. We will focus our analysis

on the particular Southern state example of Texas, which is both the largest state by population

after California and has had an especially high number of mandatory arbitration cases. The status

of California and Texas as exemplars of relatively employee favorable and unfavorable states,

respectively, is supported by Block and Roberts’ (2000) index ranking of labor standards

amongst American states and Canadian provinces, where they rank California as having the 15th

highest labor standards out of the 63 jurisdictions and Texas as only the 57th highest.

Data and Methods

The data used in this study was published by the American Arbitration Association in

compliance with Section 1281.96 of the California Code of Civil Procedure. This California

state law mandates that private arbitration service providers publish select characteristics on all

consumer arbitrations administered within the past five years. In the employment context, the

burden of this provision is limited to cases where employees were required to accept a pre-

dispute arbitration agreement drafted by their employer (i.e., employer-promulgated

agreements). Specifically, the California Code requires providers to publish such basic case

information as the: name of the employer, name of the arbitrator, salary level of employee,

prevailing party, date of filing, date of disposition; type of disposition, amount claimed, amount

awarded, total arbitrator fee, fee allocation between parties, and whether the employee was self-

represented. Information regarding employee characteristics, types of claims being filed (e.g.,

FLSA, contract, Title VII), the arbitrator’s complete written award, among many others, are

conspicuously absent from these disclosure requirements. And while limited in scope, simply by

affording a glimpse into the privately-administered types of cases implicated in the Gilmer and

15

Circuit City decisions, this reporting provision is a wellspring among a dearth of publically

available information.

The American Arbitration Association (AAA), the nation’s largest provider of

employment arbitration services, publishes quarterly consumer arbitration reports (hereafter

referred to as “AAA C-filings”) containing information on all consumer disputes it administers

nationwide pursuant to a broad interpretation of the Section 1281.96. This provides a

comprehensive national set of data on all mandatory employment arbitration cases administered

by the AAA. By virtue of its size and liberal disclosure policy, the AAA C-filings represent the

best publically available dataset for analyzing disposition trends in employment arbitration. The

data used in this article include all employment disputes arising under employer-promulgated

arbitration agreements published in these AAA C-filings.3 Covering claims from all 50 states,

the data include 8,211 employment cases filed and terminated between January 1, 2003 and

December 31, 2011.4 Of these 8,211 cases, 2,211 were adjudicated by an arbitrator, while the

remaining 6,000 were settled, withdrawn, or otherwise disposed of prior to the award stage.

All variables used in the proceeding analysis section, with the exception of National

Academy of Arbitrators (NAA) membership status, were imputed using content found in the

AAA C-filings. Descriptive statistics for the variables described below are shown in Table 1.

Dependent Variables

3 The current (as of June 2012) AAA C-filing can be found here: http://www.adr.org/aaa/ShowPDF?doc=ADRSTG_015407. 4 For the purpose of this analysis, a “case” is defined as an individual arbitration proceeding. Consequently, each “case” can refer to a single plaintiff naming multiple parties (such as a claim against both Citibank and Citigroup) or a claim with multiple plaintiffs naming a single defendant (if two coworkers were to file a single claim against their employer).

16

We analyze two dependent variables representing different aspects of an arbitrator’s

decision: employee win rates and award amounts. Defining what constitutes an employee win in

arbitration is mildly provocative. It seems apparent that cases resulting in a monetary award or

compensation of any kind are, to some degree, employee victories. However, if an arbitrator

were to render a de minimis, albeit non-zero, financial award for an employee, it may seem

disingenuous to interpret such an outcome as a win. For the purpose of this study an employee

win (Employee Win = “1”) is an adjudicated case where the employee receives any form of

monetary or non-monetary relief. Conversely, an employee loss (Employee Win = “0”) is

attributed to cases where an award provides no compensation for the employee. Award amounts

were positive in 442 out of 2,211 awarded cases, producing an employee win rate of roughly 20

percent. Award Amount reflects the nominal monetary amount, if any, awarded on a case. It is

published by the AAA for all awarded cases. The mean and median amounts awarded to

successful employee claimants are $112,855 and $40,312, respectively.

Independent Variables – Claim Characteristics

Repeat Employer measures the total number of instances an employer appears in our

database. It is calculated using all disposition types (settled, dismissed, withdrawn, etc.), not just

awards, as it is meant to measure employer size and familiarity with the arbitration procedure.5

3,167 unique employers appear over the period covered in the data; during this nine year span, an

employer is, on average, implicated in 3 arbitration proceedings, with one employer being named

in 416 separate cases. Repeat Employer-Arbitrator Pairing is a binary variable coded as “0” for

the first instance an employer engages a particular arbitrator and “1” for all subsequent

5 Including separate measures for employer size and employer experience would introduce problem of co-linearity.

17

interactions between the two parties. 23 percent of the cases and 15 percent of all awards

involved repeat employer-arbitrator pairs.

The variable Self-Represented indicates whether an employee was professionally

represented (“0”) in arbitration or represented themselves pro se (“1”). 1,966 cases, representing

nearly 25 percent of all disputes, involved self-represented employees. The AAA publishes the

variable “salary range,” categorizing employee income as between $0 and $100,000, $100,001

and $250,000, and $250,000 and above. Despite being subject to the mandatory disclosure

requirements enumerated in Section 1281.96, salary range data was missing for nearly two-thirds

of all observations. We therefore capture salary data using the binary variable Salary > $100k,

which is coded as “1” if salary is reported to be above $100,000 and “0” if salary is reported to b

between $0 and $100,000 or was unreported.

Finally, the California and Texas binary variables capture the filing state of each case.

They are coded as “1” if a case was filed in the state of the variable’s namesake and “0”

otherwise. Other States is coded as “1” if the case originated in any state other than California or

Texas, and “0” otherwise; it is the reference category for our regression models. These state

locations were imputed from the hearing address published in the AAA C-filing.

Independent Variables – Arbitrator Characteristics

We ascertained the National Academy of Arbitrators (NAA) membership status of

arbitrators by referencing the roster maintained at the NAA’s website. 98 arbitrators, responsible

for 14 percent of all arbitrations conducted, yielded a positive match, corresponding to a NAA

Member code of “1.” 6 The remaining 958 unmatched arbitrators were coded as “0”. We should

note that not all members agree to be listed on the NAA website; however, the 658 unique names 6See: http://naarb.org/member_list.asp (accessed July 12, 2012).

18

published provide a large enough sample to be confident in our measure. Next, we determined

Male Arbitrator (“1” = Male, “0” = Female) using arbitrator first names. When ruling on

ambiguously gendered names, we assessed middle names and resolved any remaining

uncertainty using an online search. 30 percent of hearings in the entire database and 35 percent

of cases disposed of through an award were conducted under the auspices of a female arbitrator.

Former Judge was created based on the presence (“1”) or absence (“0”) of any of the following

titles in arbitrator names: “Hon.,” “Judge,” “Honorable,” and “Justice.” Approximately 10

percent of arbitrators had this indicator of judicial experience. This may miss some cases of

former judges who no longer use these honorifics, however in most instances former judges do

tend to use these titles given the potential attractiveness of ex-judge status for the arbitrator

selection process and the ability to charge higher fees attendant to this status.

Results

Descriptive statistics and simple bivariate analyses are presented in Table 1. Overall,

employees won in 442 out of 2,211 awarded cases or nearly 20 percent of the time. However,

the employee win rate was 14.9 percent for cases involving repeat employers compared to 29.6

percent in cases involving first time employers. This 14.7 percentage point difference is

statistically significant (p<0.01). There is also a difference when comparing the employee win

rates for the 329 cases involving repeat employer-arbitrator pairs, 11.3 percent, to employee win

rates for the 1,882 cases involving first time employer-arbitrator pairs, 21.5 percent. Again, this

10.2 percentage point difference is highly statistically significant (p<0.01). Taken together, these

results suggest a strong repeat player effect and equally as strong repeat employer-arbitrator

pairing effect. The AAA C-filing data include 172 observations where employee salary is

19

reported as being above $100,000, or in 7.8 percent of all cases. The employee win rate is 32.6

percent for employees with high salaries, significantly greater (p<0.01) than the win rate of 18.9

percent for cases from all other salary categories. Thirty-three percent of employees in the

sample were self-represented and experienced a 10.5 percentage point reduction, on average, in

win rates relative to attorney-represented employees. Claims filed in Texas have an employee

win rate of 15.6 percent, significantly lower (p<0.05) than the 20.0 percent employee win rate

found for claims filed in all other states. Claims filed in California experience a win rate of 30.4

percent, which is 10.4 percentage points above the rate for claims filed in all other states. This

difference is significant at the .001 level. These regional differences are aligned with our

expectations. Arbitrator characteristics produce significant differences in terms of gender

(p<0.01), NAA membership status (p<0.01), and judicial experience (p<0.10). Employee win

rates are higher when the arbitrator is male, lower when the arbitrator is a member of the NAA,

and higher when the arbitrator is a former judge. These findings are evidence that outcomes vary

significantly by arbitrator characteristics.

The average award amount for the 442 cases where the plaintiff received an award of

damages was $112,855, with a median award of $40,312, indicating the distribution of award

amounts is right-skewed with a number of relatively large awards. Employees representing

themselves received significantly lower (p<0.01) awards relative to cases where employees had

representation. The average award amount in a self-represented case was $60,963, but was over

twice that amount, $126,683, in cases where the employee had a lawyer. Employees with high

salaries were awarded, on average, $202,077 compared to $99,911 in cases with all other salary

levels, a significant difference (p<0.01). Average award amounts were not significantly different

between the categories represented by the remaining claim characteristic variables. The final

20

significant difference (p<0.01) within cases won by employees appears between average

arbitrator awards rendered by former judges, $184,435, and arbitrators without judicial

experience, $103,930. Again, significant differences were not found between NAA members and

non-members and male and female arbitrators.

The rightmost column of Table 1 presents mean award amounts for all awarded cases,

reflecting differences in both win rates and award amounts. It can be interpreted as the expected

value of a case, taking into account both the chance of winning and the likely damages. The

expected value of cases involving repeat employers is $17,236 but is 91 percent higher, $32,885,

for cases involving first time employers. This difference is highly significant (p<0.01).

Differences are similarly significant (p<0.05) between repeat employer-arbitrator couplings and

first time associations. The average award is $12,146 for repeat employer-arbitrator pairs and

$24,473 for cases involving first time associations between an employer and arbitrator.

Employees representing themselves received, on average, $7,896 compared to an average award

of $29,753 for employees represented by an attorney. This $21,857 difference is significant

(p<0.01). Average award amounts for claims filed in Texas cannot be said to differ between

average awards from the remaining 49 states; however, average award amounts for claims filed

in California, $35,252, are significantly different (p<0.10) from the $21,993 average award in

other states. All three arbitrator characteristic variables produce significant differences. NAA

members award $10,347 on average, while non-members award significantly (p<0.05) more,

$24,568, on average. Male arbitrators award $7,194 more (p<0.10) than female arbitrators. And

the expected value of a case involving arbitrators with judicial experience award is $45,187 and

$20,382, significantly (p<0.01) less, where arbitrators did not have judicial experience.

21

To analyze the impact of simultaneous influences upon arbitration outcomes, in Table 2

we estimate regression models for employee win rates and award amounts controlling for the

variables described above. Model 1 looks at the independent effects each control variable has on

employee win rate. As employee win is a dichotomous variable, we employ a logit model.7

Models 2 and 3 estimate Poisson regression models for award amount.8 Award amount is right-

skewed, including a large number of observations where zero damages were awarded and several

high value awards. Poisson regression models the natural log of the dependent variable, thereby

normalizing the distribution of award amount to better fit the underlying assumptions of the

model. In Model 2 we estimate a Poisson regression for award amount constrained to cases

determined to be employee wins. Next, in Model 3, we use a Poisson regression for award

amount from all cases resolved through adjudication, even those where zero damages were

awarded (i.e., an employee loss). This third regression functions to model the expected value of

a case, incorporating both an employee’s chance of success and damages awarded.

The first model in Table 2 tests the effects arbitrator and claim characteristics have on the

odds employees will win their case. The coefficients (β) reported in model 1 imply that for every

unit increase in an independent variable, the likelihood of an employee win will be multiplied by

e^β. This transformation is presented as the odds ratio. For example, each appearance by an

employer multiples the odds an employee will win by e^-0.003= .997. In other words, for every

unit change in repeat employer, the odds an employee will win their case decreases by 0.3

percent. This is statistically significant (p<0.001) and practically significant considering the

mean value for repeat employer is 63, implying an average decrease of 21 percent in the

employee win rate relative to employers appearing only once. An odds ratio of 0.712 (p<0.10)

7 For a discussion of logit models, see Cramer (2003, chapter 9) 8 For a discussion of Poisson models, see Long, (1997, chapter 8).

22

for repeat employer-arbitrator pair suggests, on average, an employee’s likelihood of winning

decreases by 28.8 percent when faced with a repeat employer-arbitrator pair as opposed to pairs

with no repeat encounters. An employee who is self represented is 44.6 percent less likely to be

successful in arbitration relative to an employee with professional representation, a difference

significant at the .001 level. The odds ratio for Salary > $100K is above 1, meaning high-salaried

employees are 70.4 percent more likely to succeed in arbitration than employees with salaries

below $100,000 or in cases where salary was kept confidential. The model also suggests

employee win rates have a significant positive relationship (p<0.10) with claims filed in

California and a significant negative association (p<0.05) with claims filed in Texas relative to

claims filed in the remaining 49 states. These findings support the contention that claim

characteristics affect arbitration outcomes. With regard to arbitrator characteristics, the model

speaks with equal force that such characteristics are related arbitration outcomes, though judicial

experience was not found to be significant. However, if an arbitrator is a member of the NAA,

there is a significant (p<0.05) drop of 46.3 percent in the odds an employee will win. Finally, the

odds an employee will win their case increases by 24.9 percent if the presiding arbitrator is male

as opposed to female.

Just as logit regression models the log odds of an event, Poisson regression models the

natural log of the expected count of the dependent variable. Since we are again modeling the

natural log of a variable in Models 2 and 3, it is easiest to interpret the exponentiated

coefficients, reported as the award ratio corresponding to a one unit change in the predictor

variable. Model 2 measures the effects predictor variables have on employee award amount,

restricted to cases where employees won. Of the claim characteristic variables, only self-

represented (p<0.10) and salary > $100K (p<0.05) significantly affect the amount awarded to

23

employees. Even when they receive damages, a self-represented employee can expect a 50.2

percent decrease in award amount relative to an employee with representation. And an awarded

case involving an employee with a high salary level can expect a 93.6 percent increase in award

amount compared to cases involving all other salary amounts. Of the arbitrator characteristic

variables, only former judge is significant (p<0.05); if arbitrators were formerly judges, their

awards are 63.2 percent larger than awards from arbitrators without judicial experience.

The third model in Table 2 includes all adjudicated cases, whether the plaintiff wins or

loses. Consequently, it models the expected value to the plaintiff of an adjudicated case. Given a

unit change in repeat employer, the average damages awarded in a case decreases by 0.3 percent,

which is statistically significant (p<0.05). The mean value for repeat employer is 63, implying an

average decrease at this mean value of 21 percent in the size of the award relative to awards

rendered in cases with one-shot employers. Damages awarded to self-represented employees are

only 32.9 percent the size of damages awarded to their represented counterparts, a significant

difference (p<0.001). An employee plaintiff with salary > $100K is associated with a significant

(p<0.001) increase of 174.6 percent in the amount awarded. Finally, the model suggests a

significant association (p<0.001) between awarded damages and NAA membership status. An

arbitrator member of the NAA will, on average, award damages that are only 49.1 percent of

amounts awarded by non-members. Finally, the reported ratio of 1.605 (p<0.10) for Former

Judge indicates award amounts from former judges are 60.5 percent higher than amounts from

arbitrators without judicial experience.

Analysis

24

Our analysis provides strong evidence of a repeat player effect and, of more concern,

evidence of a repeat employer-arbitrator pair effect. Both repeat players and repeat employer-

arbitrator pairs are associated with lower employee win rates relative to their one-shot

counterparts. The average award amount for all adjudicated cases is further diminished in

observations involving repeat players. Employees trying to vindicate their rights against repeat

employers may experience lower win rates and award amounts because of arbitrator bias, but it

strikes us as plausible that repeat employers accrue legitimate advantages by virtue of their larger

size, greater resources, and experience which could also explain these results. These alternative

explanations hold less currency when interpreting the repeat employer-arbitrator pair results.

Even when controlling for employer size and experience in arbitration, employers involved with

arbitrators they have selected in the past realize more favorable outcomes than employers

involved in first time encounters. This alarming relationship suggests that arbitrators may be

responding to economic incentives and issuing relatively favorable awards to repeat clients.

Here again, however, there are alternative explanations for the reported relationship. Past

experience with an arbitrator may confer advantages that can be applied to subsequent hearings.

For example, employers involved in repeat pairs may cater their arguments to known arbitrator

preferences. Alternatively, employers may adjust their settlement strategies according to prior

experience with a given arbitrator. Although not arbitrator bias, this would still constitute an

employer advantage. In labor arbitration, the opposing institutional force of unions inoculates

labor arbitration against such advantages systematically favoring one side. Although the

plaintiffs’ bar could theoretically fulfill a similar role in employment arbitration, our results

suggests that as yet they do not have a sufficient presence to eliminate repeat employer-arbitrator

pairing advantages from mandatory employment arbitration. Future studies should control for

25

these varied explanations and explore how other institutions such as the plaintiffs’ bar could

more effectively reduce the repeat player advantages currently possessed by employers in

mandatory employment arbitration.

Proponents of employment arbitration proclaim its accessibility, particularly for

claimants unable to find representation in courts of law (Estreicher, 2001). Our findings,

however, show that employee self-representation is common and clearly associated with adverse

impacts in terms of win rates and size of damages awarded to employees. Employees are 50

percent more likely to represent themselves in arbitration relative to federal courts, so in a

technical sense the arbitral forum may be more accessible.9 But if a 50 percent reduction in both

the odds of winning and award amounts are attendant to self-representation in arbitration, such

employees may not be getting access to an institution of justice. Such a drastic reduction is

especially poignant considering Clermont and Schwab (2004, page 445) assert: “pretrial and trial

win rates [in federal courts] are similar across types of discrimination cases…despite the

differences in pro se representation…” The data do not allow a standardized comparison of cases

between those with professional and those with self-representation. Therefore, we do not know

whether self-representation is the cause of lower win rates and award amounts, or whether self-

represented claims are simply less meritorious.

Our results indicate that there are substantial differences between states in the outcomes

of mandatory arbitration. As expected, employees do much better in California than in other

states. This confirms the popular assumption that California is a particularly employee-friendly

state regime. What our results do not allow us to determine is the degree to which this is due to a

more employee friendly set of substantive employment laws in California, greater state court

9 Clermont and Schwab (2004) report that 19.8 percent of plaintiffs were represented pro se in federal courts, while 29.5 percent of claimants in our sample, 32 percent of adjudicated cases, were self-represented.

26

recognition of rights to due process in arbitration, or the strength of the California plaintiffs’ bar.

Conversely, Texas appears to be a relatively employer friendly state for mandatory arbitration

based on the lower employee success rates we find in that state. These findings indicate that

mandatory arbitration does not operate as a uniform national institution, but rather exhibits

substantial regional variation at the state level. This raises the question of whether it makes sense

to assume a uniform federal policy on arbitration and, particularly, to assume that arbitration

uniformly provides across the country a system that is merely an alternative set of procedures for

enforcing the same substantive rights. By contrast, what we find is a distinct lack of uniformity

in the outcomes of attempts to enforce substantive employment rights through mandatory

arbitration.

Turning to arbitrator characteristics, the multivariate models provide nuanced and

reinforcing results to the bivariate analysis, indicating variation in arbitrator characteristics is

related to outcomes. This result raises potential concerns over the ability of sophisticated

employers to exploit this variation to obtain more favorable outcomes. Unlike litigation, parties

in arbitration have direct influence over the party who decides their case. If such an asymmetric

benefit is conferred on employers, being forced into an arbitral forum will directly affect the

resolution and determination of universally-applicable substantive rights. And this would be in

stark contrast to the Supreme Court’s current stance of assuming a different process but similar

outcomes. These results, however, do not allow us to distinguish between whether arbitrator

characteristics are the cause of these differences, or whether the types and merits of cases

presented to certain types of arbitrators systematically differ. Addressing this uncertainty should

be a priority for future researchers.

27

One unexpected result is that NAA members are associated with employee unfriendly

outcomes. Coming from a collective bargaining environment where labor arbitrators enforce

standards that are tied to the actual contract language, especially the just cause standard for

discipline and dismissal, we predicted NAA members would apply relatively employee-

favorable standards in employment arbitration. However, the employee win rate is lower if the

arbitrator is an NAA member and there is also a reduction in the expected value of a case. This

may be the result of NAA member arbitrators being more skeptical of employment law based

claims, as compared to the labor arbitration claims they more typically adjudicate. Or

alternatively it may be due to employers selecting NAA member arbitrators in cases where

employees have weaker claims. The finding should certainly be troubling for NAA members

participating in employment arbitration and suggests that they should guard against any

systematic tendency to be skeptical of employee claims.

An employee’s odds of winning a case are higher when the arbitrator is male relative to

female. This may be evidence that male arbitrators are more severe to employers (or, stated

differently, less severe to employees) or, simply indicate that female arbitrators are selected for

less meritorious cases. In addition, former judges tend to make higher awards than other

arbitrators, providing further evidence of significant variation in outcomes depending on

arbitrator characteristics. This raises the policy concern that sophisticated parties, in this area

particularly larger more experienced employers, will be able to obtain advantages through the

arbitrator selection process by choosing arbitrators based on critical arbitrator characteristics.

We want to recognize some limitations of our data. First, we investigate disposition data

from a single arbitration provider and there is no guarantee the data found in AAA C-filings is

representative of arbitrations generally. Future studies should include data from multiple

28

arbitration providers to determine the generalizability of our results. Of particular concern is that

ad hoc arbitration cases where there is no administering agency enforcing standard procedures

may have very different characteristics and potentially more substantial due process defects.

Second, the rudimentary reporting requirements found in Section 1281.96 of the California Code

limit our ability to draw strong casual conclusions. Without specific case information, we cannot

be certain that types or merits of cases and our predictor variables are independent. For example,

do former judges award larger damages to employees because of their past careers, or because

they are selected in higher-value cases? Lastly, we employ just one of a multitude of potential

measures for employer size, experience in arbitration, and past experience with a given

arbitrator. Future studies could, for example, model employer experience in arbitration using

data from multiple service providers.

Conclusion

The rise of mandatory arbitration represents the emergence of a major new institution for

the governance of employment relations in the individual rights era. Alternative dispute

resolution procedures such as arbitration and mediation are often viewed as relatively neutral

technique focused interventions in labor and employment relations. By contrast, we have sought

to show how the institutional structure of these procedures is a key determinant of their operation

and impact. Mandatory arbitration’s endorsement by the U.S. Supreme Court was premised on

the idea that it simply involved an alternative set of procedures for enforcing the same set of

substantive rights. In our study, we have analyzed how the outcomes of efforts to enforce

substantive employment rights in fact vary widely depending on who the decision-makers are

and what the institutional context is. Justice in mandatory arbitration is not universal in that its

29

outcomes vary both regionally across the country and depending on the characteristics of the

arbitrator deciding the case. Justice is also not blind if an employer is able to gain an advantage

from selecting an arbitrator with desirable characteristics and especially if there are gains from

doing repeat business with the same arbitrator.

30

Table One

Plaintiff Win Rates and Award Amounts

Category Counts Employee Win Rate

Award Amount ($) (Wins Only)

Award Amount ($)

(All Rulings) N (%) % Mean (Median) Mean All cases 2,211 (100) 19.99 112,855 (40,312) 22,632 Repeat employer 1,447 (65.5) 14.93** 115,228 (37,701) 17,236** Non-repeat employer

764 (34.5) 29.58 110,588 (43,060) 32,885

Repeat employer-arbitrator pairing

329 (14.8) 11.25** 107,997 (40,000) 12,146*

Non-repeat pairing 1,882 (85.2) 21.52 113,299 (40,624) 24,473 Self-represented 719 (32.5) 12.93** 60,963** (12,745) 7,896** Attorney rep’d 1,492 (67.5) 23.39 126,683 (61,488) 29,753 Salary > $100K 172 (7.8) 32.56** 202,077** (69,751) 65,793** Other Salary 2,039 (92.2) 18.93 99,911 (37,109) 18,979 NAA member 301 (13.6) 13.62** 75,710 (42,620) 10,347* Non-NAA member 1,910 (86.4) 20.99 116,653 (39,867) 24,568 Male arbitrator 1,441 (65.2) 21.65** 115,706 (39,934) 25,139+ Female arbitrator 770 (34.8) 16.88 106,013 (41,700) 17,945 Former Judge 200 (9.0) 24.50+ 184,435** (75,000) 45,187** Non-judge 2,011 (91.0) 19.54 103,930 (38,617) 20,382 California 125 (5.6) 30.40** 115,032 (46,750) 35,252+ Texas 289 (13.1) 15.57* 135,034 (65,000) 21,173 Other states 1,797 (81.3) 19.95 109,845 (37,219) 21,993

Note: Within category differences significance levels: + p<.10; * p<.05; ** p<.01.

31

Table Two Plaintiff Win and Award Amount Models

Model 1 Model 2 Model 3 Plaintiff Wins Award

Amounts (Employee wins only)

Award Amounts

(All cases)

Logit Poisson Poisson [Odds Ratio] [Award Ratio] [Award Ratio]

Repeat Employer -0.003*** -0.0003 -0.003* (0.00) (0.0007) (0.001) [0.997] [0.9997] [0.997]

Repeat Emp-Arb Pair -0.34+ 0.085 -0.175 (0.196) (0.266) (0.320) [0.712] [1.089] [0.839]

Self-represented -0.59*** -0.695+ -1.11*** (0.130) (0.377) (0.418) [0.554] [0.499] [0.329]

Salary > $100K 0.53** 0.661* 1.01*** (0.178) (0.263) (0.291) [1.704] [1.936] [2.746]

NAA member -0.45* -0.305 -0.711*** (0.183) (0.225) (0.269) [0.637] [0.737] [0.491]

Male arbitrator 0.22+ 0.019 0.163 (0.120) (0.211) (0.228) [1.249] [1.020] [1.177]

Former Judge 0.04 0.490* 0.473+ (0.181) (0.214) (0.250) [1.044] [1.632] [1.605]

California 0.36+ -0.174 0.026 (0.211) (0.276) (0.325) [1.436] [0.840] [1.027]

Texas -0.37* 0.102 -0.127 (0.177) (0.209) (0.251) [0.693] [1.107] [0.881]

Constant -1.17*** 11.57*** 10.16*** (0.118) (0.232) (0.256)

F [Chi-sq] 99.94*** 40.83*** 93.56*** Pseudo R2 0.0452 0.0993 0.1134

N 2211 442 2211 Note: + p<.10; * p<.05; ** p<.01; *** p<.001.

32

References American Arbitration Association (2012). “Qualification Criteria and Responsibilities for

members of the AAA Panel of Employment Arbitrators” http://www.adr.org/aaa/ShowPDF?doc=ADRSTG_003881. Accessed on 6/12/2012

American Bar Association (2006). “Charting Our Progress: The Status of Women in the

Profession Today,” http://www.americanbar.org/content/dam/aba/migrated/marketresearch/PublicDocuments/ChartingOurProgress.authcheckdam.pdf. Accessed on 6/12/2012

Bemmels, Brian (1990). “Arbitrator characteristics and arbitrator decisions.” Journal of Labor

Research 11: 181-91. Bemmels, Brian (1988). “Gender Effects in Discharge Arbitration.“ Industrial and Labor

Relations Review 42: 63-76. Bemmels, Brian (1988a). “The Effect of Grievants' Gender on Arbitration Decisions” Industrial

and Labor Relations Review 41: 251-62. Bemmels, Brian (1988b). “Gender Effects in Discipline Arbitration: Evidence from British

Columbia.” Academy of Management Journal 31: 699-706. Bigoness, W. J. & DuBose, P. B. (1985). “Effects of Gender on Arbitrators' Decisions.”

Academy of Management Journal 28: 485-91. Bingham, Lisa B (1999). “Employment Arbitration: The Repeat Player Effect.” Employee Rights

and Employment Policy Journal 1(1): 1-38. Bingham, Lisa B. (1998). “On Repeat Players, Adhesive Contracts, and the Use of Statistics in

Judicial Review of Employment Arbitration Awards.” McGeorge Law Review 29(2): 223–259.

Bingham, Lisa B. (1998b). “An Overview of Employment Arbitration in the United States: Law,

Public Policy and Data.” New Zealand Journal of Industrial Relations 23(2): 5-19. Bingham, Lisa B. (1997). “Employment Arbitration: The Repeat Player Effect,” 1 Employee

Rights & Employment Policy J. 189. Bingham, Lisa B. & Mesch, Debra J. (2002). “Decision Making in Employment and Labor

Arbitration” Industrial Relations, 39(4): 671-694. Block, Richard & Roberts, Karen (2000). “A Comparison of labour standards in the United

States and Canada” Relations Industrielles, 55(2): 273-307.

33

Bloom, D.E. & Cavanagh C.L. (1986). “The Impact of Attorneys and Arbitrators on Arbitrator Awards.” Industrial and Labor Relations Review 40(1): 543-555.

Bureau of Labor Statistics (2012). “Union Members Summary,”

http://www.bls.gov/news.release/union2.nr0.htm. Accessed 6/12/2012. Caudill, Steven B. & Oswald, Sharon L. (1993). “A Sequential Selectivity Model of the

Decisions of Arbitrators,” Managerial and Decision Economics, 14(3): 261-267. Clermont, Kevin J. & Schwab, Stewart J. (2004). ‘How Employment Discrimination Plaintiffs

Fare in Federal Courts,” J. of Empirical Legal Studies 1(2): 429-458. Colvin, Alexander J.S. (2012). “American workplace dispute resolution in the individual rights

era,” International Journal of Human Resource Management, 23(3):459-475. Colvin, Alexander J.S. (2011). “An Empirical Study of Employment Arbitration: Case

Outcomes and Processes,” J. of Empirical Legal Studies, 3(1): 1-23. Colvin, Alexander J.S. (2008). “Empirical Research on Employment Arbitration: Clarity Amidst

the Sound and Fury?” Employee Rights and Employment Policy Journal,11(2): 405-446. Cramer, J.S. (2003). Logit Models from Economics and Other Fields. Cambridge: Cambridge

University Press. Dworkin, Harry J. (1974). “How Arbitrators Decide Cases.” Labor Law Journal 24(2):200-210. Eagley, Alice H. & Wood, Wendy (1999). “The Origins of Sex Differences in Human Behavior:

Evolved Dispositions Versus Social Roles,” American Psychologist 54(6): 408-423. Eisenberg, Theodore & Hill, Elizabeth (2003). “Arbitration and Litigation of Employment

Claims: An Empirical Comparison.” Dispute Resolution Journal 58(4): 44-54. Estreicher, Samuel (2001). “Saturns for Rickshaws: The Stakes in the Debate over Pre-Dispute

Employment Arbitration Agreements,” Ohio State Journal on Dispute Resolution 16(6): 559-570

Galanter, Marc. (1975). “Why the "Haves" Come Out Ahead: Speculations on the Limits of

Legal Change,” L. & Soc'y Rev. 95(9). Gilligan, Carol. (1982) In A Different Voice. Psychological Theory and Women's Development.

Cambridge: Harvard University Press. Hill, Elizabeth. (2003). “AAA Employment Arbitration: A Fair Forum at Low Cost.” Dispute

Resolution Journal 58(2): 8-16.

34

Klaas, Brian S., Mahony, Douglas, and Wheeler, Hoyt N. (2006). “Decision-Making about workplace Disputes: A policy-Capturing Study of Employment Arbitrators, Labor Arbitrators, and Jurors.” Industrial Relations. 45(1):68-93

Kulik, Carol R., Perry, Elissa L. & Pepper, Molly B. (2003). “Here Comes the Judge: The

Influence of Judge Personal Characteristics on Federal Sexual Harassment Case Outcomes,” Law and Human Behavior, 27(1): 69-86.

Lawson, Eric W. (1981). “Arbitrator Acceptability: Factors Affecting Selection,” The Arbitration

Journal 36: 22-29. Lewin, David (2008). “Employee Voice and Mutual Gains,” Proceedings of the 60th Annual

Meeting of the Labor and Employment Relations Association (LERA), 61-83. Long, J.S. (1997). Regression Models for Categorical and Limited Dependent Variables.

Thousand Oaks, CA: Sage. National Academy of Arbitrators (2012). “NAA Membership Guidelines”

(http://naarb.org/member_guidelines.asp. Accessed on 6/12/201. Oppenheimer, David Benjamin (2003) “Verdicts Matter: An Empirical Study of California

Employment Discrimination and Wrongful Discharge Jury Verdicts Reveals Low Success Rates for Women and Minorities,” 37 U.C. Davis Law Rev. 535-549.

Piore, Michael & Safford, Sean (2006). “Changing Regimes of Workplace Governance, Shifting

Axes of Social Mobilization and the Challenge to Industrial Relations Theory,” Industrial Relations, 45(3): 299-325.

Scott, Clyde & Shadoan, Elizabeth (1989). “The Effect of Gender on Arbitration Decisions.”

Journal of Labor Research 10: 429. Sherwyn, David, Samuel Estreicher, and Michael Heise (2005). “Assessing the Case for

Employment Arbitration: A New Path for Empirical Research.” Stanford Law Review 57: 1557-1591.

Steffensmeier, Darrell, and Chris Herbert (1999). "Women and men policymakers: does the

judge's gender affect the sentencing of criminal defendants?" Social Forces. 77. 3: 1163(3). Opposing Viewpoints Resource Center. Gale. CIC Penn State University.

Thorton, Robert J and Zirkel, Perry A. (1990). “The Consistency and Predictability of Grievance

Arbitration Awards.” Industrial and Labor Relations Review 43(2): 294-307.

U.S. Supreme Court Gilmer v. Interstate/Johnson Lane Corp., 500 U.S. 20, (1991).

U.S. Supreme Court Circuit Cirty Stores, Inc. v. Adams, 532 U.S. 105 (2001).

35

Zirkel, Perry A. (1983). “A Profile of Grievance Arbitration Cases.” Arbitration Journal 38(1): 35-38