An Empirical Study of Employment Arbitration: Case Outcomes and Processes

Abstract

Using data from reports filed by the American Arbitration Association (AAA) pursuant to California Code requirements, this article examines outcomes of employment arbitration. The study analyzes 3,945 arbitration cases, of which 1,213 were decided by an award after a hearing, filed and reaching disposition between January 1, 2003 and December 31, 2007. This includes all the employment arbitration cases administered nationally by the AAA during this time period that derived from employer-promulgated arbitration procedures. Key findings include: (1) the employee win rate among the cases was 21.4 percent, which is lower than employee win rates reported in employment litigation trials; (2) in cases won by employees, the median award amount was $36,500 and the mean was $109,858, both of which are substantially lower than award amounts reported in employment litigation; (3) mean time to disposition in arbitration was 284.4 days for cases that settled and 361.5 days for cases decided after a hearing, which is substantially shorter than times to disposition in litigation; (4) mean arbitration fees were $6,340 per case overall, $11,070 for cases disposed of by an award following a hearing, and in 97 percent of these cases the employer paid 100 percent of the arbitration fees beyond a small filing fee, pursuant to AAA procedures; (5) in 82.4 percent of the cases, the employees involved made less than $100,000 per year; and (6) the mean amount claimed was $844,814 and 75 percent of all claims were greater than $36,000. The study also analyzes whether there is a repeat player effect in employer arbitration. The results provide strong evidence of a repeat employer effect in which employee win rates and award amounts are significantly lower where the employer is involved in multiple arbitration cases, which could be explained by various advantages accruing to larger organizations with greater resources and expertise in dispute resolution procedures. The results also indicate the existence of a significant repeat-employerarbitrator pairing effect in which employees on average have lower win rates and receive smaller damage awards where the same arbitrator is involved in more than one case with the same employer, a finding supporting some of the fairness criticisms directed at mandatory employment arbitration.

Full text

Page 1

An Empirical Study of Employment
Arbitration: Case Outcomes and Processesjels_12001..23
Alexander J. S. Colvin*
Using data from reports filed by the American Arbitration Association (AAA) pursuant to
California Code requirements, this article examines outcomes of employment arbitration.
The study analyzes 3,945 arbitration cases, of which 1,213 were decided by an award after
a hearing, filed and reaching disposition between January 1, 2003 and December 31, 2007.
This includes all the employment arbitration cases administered nationally by the AAA
during this time period that derived from employer-promulgated arbitration procedures.
Key findings include: (1) the employee win rate among the cases was 21.4 percent, which
is lower than employee win rates reported in employment litigation trials; (2) in cases won
by employees, the median award amount was $36,500 and the mean was $109,858, both of
which are substantially lower than award amounts reported in employment litigation; (3)
mean time to disposition in arbitration was 284.4 days for cases that settled and 361.5 days
for cases decided after a hearing, which is substantially shorter than times to disposition
in litigation; (4) mean arbitration fees were $6,340 per case overall, $11,070 for cases
disposed of by an award following a hearing, and in 97 percent of these cases the
employer paid 100 percent of the arbitration fees beyond a small filing fee, pursuant to
AAA procedures; (5) in 82.4 percent of the cases, the employees involved made less than
$100,000 per year; and (6) the mean amount claimed was $844,814 and 75 percent of all
claims were greater than $36,000. The study also analyzes whether there is a repeat player
effect in employer arbitration. The results provide strong evidence of a repeat employer
effect in which employee win rates and award amounts are significantly lower where the
employer is involved in multiple arbitration cases, which could be explained by various
advantages accruing to larger organizations with greater resources and expertise in
dispute resolution procedures. The results also indicate the existence of a significant
repeat-employer–arbitrator pairing effect in which employees on average have lower win
rates and receive smaller damage awards where the same arbitrator is involved in more
than one case with the same employer, a finding supporting some of the fairness criticisms
directed at mandatory employment arbitration.
*408 Dolgen Hall, ILR School, Cornell University, Ithaca, NY 14853-3901; email: ajc22@cornell.edu. Colvin is
Associate Professor of Labor Relations and Conflict Resolution, ILR School, Cornell University.
I thank participants in the Research Conference on Access to Justice: Empirical Perspectives, held at the NYU
School of Law, November 2008 and the NYU 62nd Annual Labor Conference, June 2009, for their helpful comments.
I also thank my graduate research assistants, Lauren Beauseigneur, Mark Gough, Ying Lu, and Paul Schwartz, for their
hard work on this research project. The analysis in this article particularly benefited from valuable suggestions from
Professor Martin Wells, JELS co-editor, and from an anonymous reviewer.
Journal of Empirical Legal Studies
Volume 8, Issue 1, 1–23, March 2011
1

Page 2

I. Introduction
The expansion of arbitration of employment disputes represents a revolutionary change in
the landscape of dispute resolution. The combination of rising levels of litigation in the
employment area and the Supreme Court’s 1980’s reversal of its earlier rejection of the use
of arbitration to resolve statutory claims produced a perfect storm of incentives for employ-
ers to adopt arbitration agreements as mandatory terms and conditions of employment.
Advocates and critics have fiercely debated the question of whether the growth of arbitra-
tion represents a more efficient and equitable mechanism for resolving disputes that would
otherwise clog the courts, or an imposition of a mandatory arbitration mechanism on
unwitting employees that undermines the statutory rights designed to protect them (e.g.,
Estreicher 2001; Schwartz 2009; Sherwyn et al. 2005; Stone 1996). These debates have
achieved a new intensity with the consideration by Congress of the proposed Arbitration
Fairness Act (AFA), which would effectively bar mandatory predispute agreements to
arbitrate employment and consumer claims.
Despite the intensity of focus on public policy issues relating to employment arbitra-
tion, solid empirical data on this topic have proven slow and difficult to gather. Part of the
reason for this is the lack of publicly available data on arbitration. Most empirical research
has had to rely on cases or files for which individual arbitration service provider organiza-
tions have provided access. The resulting data sets have tended to be relatively small in size
and potentially lacking representativeness of the broader population of arbitration cases. In
this article, I will describe the results of an analysis of a new large-scale data set based on
publicly available information about employment arbitration cases administered by the
American Arbitration Association (AAA). I will note at the outset that a significant limita-
tion of existing empirical knowledge on arbitration is that this study, and much of the past
empirical research, have been based on AAA data. Although the AAA is arguably the
leading and largest player in the arbitration field, it also may be somewhat unrepresentative
in its willingness to sign onto and monitor compliance with due process protocols on
arbitration.
Employment arbitration grew dramatically in the wake of the 1991 Gilmer decision.1
The proportion of workers covered by nonunion employment arbitration procedures now
likely exceeds those covered by union representation (Colvin 2008). Indeed, recent esti-
mates suggest that for perhaps a third or more of nonunion employees, arbitration, not
litigation, is the primary mechanism of access to justice in the employment law realm
(Lewin 2008). Yet our empirical knowledge of the nature of this system remains minimal at
best. Basic questions such as the typical characteristics and outcomes of cases in employ-
ment arbitration remain to be definitively answered. Here, I move beyond past research by
analyzing employment arbitration outcomes using a representative data set on 1,213 AAA
cases from 2003–2007 that derive from employer-promulgated arbitration procedures. I
present outcome statistics on key measures such as employee win rates, award amounts,
arbitrator fees, and length of time to process cases. I then turn to the issue of whether there
1Gilmer v. Interstate/Johnson Lane, 500 U.S. 20 (1991).
2
Colvin

Page 3

are repeat player effects in arbitration and, if so, whether we can identify possible expla-
nations for them. Lastly, I examine the issue of self-representation by employees in employ-
ment arbitration.
II. The Data
The data on employment arbitration in this study are based on arbitrator service provider
filings required under California state law. Under the California Civil Procedure Code,
organizations that provide arbitration services within the state are required to make avail-
able to the public certain prescribed information on arbitration cases administered by the
service provider that involve consumers.2This provision applies to employment arbitration
cases that are initiated under employer-promulgated agreements, as opposed to under
individually negotiated agreements. The effect of this law is to override contracts that
protect the presumptively private nature of arbitration and allow public access to informa-
tion on arbitration outcomes. The provision prescribes what types of information need to
be filed, including: the name of the employer; the name of the arbitrator; filing and
disposition dates; amounts of claims; amounts awarded; and fees charged. At the same time,
many other pieces of information, notably the name of the employee and the basis for the
claim, are not included. More generally, the arbitration service providers are not required
to provide the complete arbitration decision accompanying the award. Despite these sig-
nificant limitations, the California Code filings provide a major new source of data on
employment arbitration outcomes, which allows us to analyze a number of questions
regarding this dispute resolution system.
In the present study, I analyze cases administered by the American Arbitration
Association (AAA). The reason for focusing on the AAA is that it is the largest of the
arbitration service providers in the employment arbitration field and has provided the most
complete filings in this area. An additional advantage is that to comply with the California
Code requirements, the AAA has included in its filings all employment arbitration cases
under employer-promulgated procedures that it administers nationally. As a result, the
AAA filings provide a much larger national data set that is not restricted to cases heard in
California. Based on a comparison of arbitration service provider filings, those compiled by
the AAA appear relatively comprehensive. A general problem with all filings in this area is
that they contain some degree of missing data on particular variables. For example,
although the California Code provision requires the service provider to include informa-
tion on the employee’s salary level, in many cases the parties decline to provide this
information. Although some degree of missing data exists in all the service provider filings,
the AAA filings include substantially fewer instances of missing data than those of other
service providers. The nature of the data set also limits analysis to only those items that the
California Code provision requires arbitration providers to report. For example, the data
set does not include the state where the case occurred or the legal basis for the claim.
2Cal. Civ. Proc. Code § 1281.96 (West 2007).
An Empirical Study of Employment Arbitration
3

Page 4

Despite these limitations, the filings provide a more extensive source of data on employ-
ment arbitration than has previously been available to researchers.
The data set analyzed in this article includes all employment cases from the AAA
California Code filings (what I will refer to henceforth as the AAA-CC filings) for the period
January 1, 2003 through December 21, 2007. This produced a total of 5,592 cases. Of these,
1,647 were employment mediation cases administered by the AAA. For purposes of this
analysis, I focused on the remaining 3,945 employment arbitration cases in the data set.
Data on the individual cases were compiled from the filings by a team of four graduate
students working under my supervision. I also separately rechecked the data for typographi-
cal and other errors. For many of the analyses conducted in this study, I focus on the 1,213
cases that resulted in awards, with the remainder of the cases being settled or withdrawn
prior to the award stage. Cases were relatively evenly distributed across the five-year period
of the study, though with some slow growth over this time period.3In the analysis below, I
control for year-specific effects to check for possible changes over time in outcomes.
III. Arbitration Outcomes
Given the relatively limited extent of existing information on employment arbitration,
some of the most interesting questions relate to the basic descriptive outcomes from
arbitration. Knowing what the mean and median outcomes of arbitration are will allow us
to develop a general portrait of how this dispute resolution system operates. They also
provide an initial basis for moving toward comparisons of litigation and arbitration out-
comes. Whereas there have been increasingly sophisticated analyses of litigation and its
outcomes in recent years (e.g., Clermont & Schwab 2004; Eisenberg & Schlanger 2003;
Oppenheimer 2003), our understanding of arbitration has lagged behind. Although the
present data do not allow a comparison of systematically matched cases in litigation and
arbitration, to begin to compare across systems it is initially necessary to establish what the
arbitration outcomes are.
A. Win Rates
One of the most basic questions in arbitration is who wins? Past research in this area has
mostly used convenience samples of arbitration awards maintained by organizations like the
AAA or the securities industry service providers. These studies tended to show relatively
high employee win rates in arbitration. For example, early studies by Bingham (1998a),
Maltby (1998), and Howard (1995) found employee win rates in the 65–75 percent range.
More recent studies, including those by Bingham and Sarraf (2000), and by Hill (2003),
found lower, though still substantial, employee win rates in the 40–45 percent range.
Examining securities industry employment arbitration cases, Delikat and Kleiner (2003)
3The numbers of total cases reaching disposition by year were: 2003—294 cases; 2004—803 cases; 2005—906 cases;
2006—957 cases; 2007—982 cases. The reason for the smaller number in 2003 is that the data set includes only cases
filed after January 1, 2003, so that many cases filed in 2003 did not reach disposition until 2004.
4
Colvin

Page 5

found a similarly high 46 percent employee win rate. These employee win rates compare
favorably to employee win rates found in litigation (see Table 1), ranging from the 33 and
36 percent employee win rates in federal court employment discrimination trials reported
in studies by Delikat and Kleiner (2003) and by Eisenberg and Hill (2003), to the employee
win rates in the 50–60 percent range found in studies of state court trials (Oppenheimer
2003). A note of caution in interpreting these findings, however, is that studies by both
Eisenberg and Hill (2003) and by Bingham and Sharaff (2000) found that employee win
rates were lower in cases based on employer-promulgated procedures than in cases based
on individually negotiated contracts. Earlier studies tended to include both types of cases in
the same sample, which would be likely to produce higher overall employee win rates given
the greater likelihood of success under individually negotiated contracts, which often
involve contractual claims by senior business executives.
What are the employee win rates in the AAA-CC filings data? To answer this question,
itisnecessarytomakedecisionsabouthowtoclassifyanemployee“win.”Mostgenerally,any
caseinwhichtheemployeereceivessomeawardrepresentsacaseinwhichthearbitratorhas
ruled in the employee’s favor on at least some aspect of his or her claim. On the other hand,
if the employee receives an award, but the amount is relatively small and/or the award is
much lower than the amount claimed, the employee might view the outcome of the case as
unsuccessful. Taking a narrow view of an employee win as cases in which the employee
receivesalloratleastsomesubstantialportionoftheamountclaimedwouldproducealower
estimateoftheemployeewinrateinarbitration.Bycontrast,usingabroaderdefinitionofan
employee win will increase the estimated win rate. To take a more arbitration-favorable
approach in this study, I use a broad definition of an employee win as including any case in
which some award of damages, however small, is made in favor of the employee. Using this
Table 1:Descriptive Statistics on Employment Arbitration and Litigation
Employment
Arbitration
(AAA-CC Data)
Federal Court
Employment
Discrimination
(Eisenberg & Hill)
State Court
Non-Civil Rights
(Eisenberg & Hill)
California State
Court Common-Law
Discharge (Oppenheimer)
Employee win rate21.4%
(n = 1,213)
$36,500
(2005 $)
$109,858
36.4%
(n = 1,430)
$150,500
(2005 $176,426)
$336,291
(2005 $394,223)
57% 59%
(n = 145)
$68,737
(2005 $85,560)
$462,307
(2005 $575,453)
$1,291,020
$328,008
(n = 117)
$296,991
(2005 $355,843)
Median damages
Mean damages
SD
Mean including zeros
(2005 $)
Mean time to trial (days)
$238,227
$23,548 $143,497
361.5709723
Notes: Conversions of award statistics to standard 2005 $ amounts are provided to facilitate comparison. AAA-CC
Data: Data set on all employment arbitration cases based on employer-promulgated procedures administered by the
American Arbitration Association during the period January 1, 2003 to December 31, 2007. Data assembled by the
author from reports filed by the AAA under California Code arbitration service provider reporting requirements.
Eisenberg & Hill: Litigation statistics reported in Eisenberg & Hill (2003). Oppenheimer: Litigation statistics reported
in Oppenheimer (2003).
An Empirical Study of Employment Arbitration
5