An Empirical Investigation of Diversity in U.S. Arbitration

Abstract

For decades, the United States system of arbitration has been subject to nearly constant public criticism. Calling arbitration a rigged judicial system, consumer and employee rights groups have voiced opposition to the practice of “forced arbitration” whereby millions of Americans are contractually required to resolve disputes in arbitration rather than in litigation. On top of the concerns over the unfairness of forced arbitration itself, recent attention has been drawn to the lack of racial and gender diversity within the arbitrator profession. When women and racially marginalized plaintiffs are forced to arbitrate their employment discrimination or consumer-based claims in the arbitral forum, that they may have no meaningful access to arbitrators that look like them seems additionally problematic.

Scholars in the field have argued back and forth about the root of the diversity problem. Is it a labor supply problem? In other words, are parties to arbitration open to hiring marginalized arbitrators but there are just not enough to choose from? Or is it a labor demand problem? In other words, when women and arbitrators of color are available, are they chosen at rates consistent with their white male counterparts? Or, are both supply and demand problems at work? Because much of the scholarly diversity conversation has been based on anecdotal information and survey data which don't cover the full population of U.S. arbitrators, these basic questions are still unanswered.

This paper contributes to the literature by using an originally-collected data set of arbitrator race, ethnicity and gender from the two largest arbitration firms in the U.S., Judicial Arbitration and Mediation Services (“JAMS”) and the American Arbitration Association (“AAA”). The data were collected using public data sources and cutting-edge machine learning techniques. This is the first-ever scholarly effort to empirically estimate the race and ethnicity of arbitrators for both the JAMS and AAA populations. The analysis presents estimates of the demographic profile of the supply of U.S. arbitrators and the demographic profile of the subset of arbitrators that are actually selected to arbitrate—with a special focus on the extent to which under-selection is happening.

The study has four main findings. First, along the supply dimension, women and people of color are underrepresented amongst JAMS arbitrators, both relative to the U.S. population and relative to the population of American lawyers and judges. The extent of the underrepresentation for both groups is significant, though it is more severe for arbitrators of color than for female arbitrators. For AAA arbitrators, I find an even greater degree of under-representation for Black arbitrators.

Second, along the demand dimension, I find different results for JAMS and AAA. For JAMS, I find that, conditional on being selected to arbitrate at least once in the sample period, Asian and Black arbitrators receive fewer cases than their proportional share, and female arbitrators receive slightly more cases than their proportional share. Moreover, arbitrators that were formerly judges receive more cases than their proportional share. For AAA, the selection analysis is hampered by limited data availability. However, the data that I do have suggest that diverse neutrals are selected for cases at a rate that is at or above their proportional share.

Third, given the first two results, my data suggest that diversity issues exist both along the labor supply dimension and the labor demand dimension within U.S. arbitration.

Fourth and finally, I find that future empirical diversity work in arbitration will be severely hindered unless more and better data are available to researchers.

The study concludes by offering concrete and specific recommendations for how and why better data should be collected and made available to the public.

Keywords: machine learning, race and the law, arbitration, empirical legal studies, alternative dispute resolution, diversity, empirical, AAA, American Arbitration Association, JAMS

JEL Classification: K00, J7, J52, C55