Let's Not Make a Deal: An Empirical Study of Decision Making in Unsuccessful Settlement Negotiations


  • We thank Jeffrey Rachlinski, Theodore Eisenberg, and an anonymous review for their insightful comments on previous versions of this article.

*Randall L. Kaiser, DecisionSet, 550 Hamilton Ave., Ste. 300, Palo Alto, CA 94301; email: rkiser@decisionset.com. Asher is Director, Research and Scholars Programs, Wharton Undergraduate Division, and Adjunct Professor of Finance at The Wharton School, University of Pennsylvania; McShane is a graduate student in the Department of Statistics, The Wharton School, University of Pennsylvania.


This study quantitatively evaluates the incidence and magnitude of errors made by attorneys and their clients in unsuccessful settlement negotiations. The primary study analyzes 2,054 contested litigation cases in which the plaintiffs and defendants conducted settlement negotiations, decided to reject the adverse party's settlement proposal, and proceeded to arbitration or trial. The parties' settlement positions are compared with the ultimate award or verdict, revealing a high incidence of decision-making error by both plaintiffs and defendants. This study updates and enhances three prior studies of attorney/litigant decision making, increasing the number of cases in the primary data sets more than threefold, adding 72 explanatory variables from 19 classes, applying a multivariate analysis, presenting an historical review of error rates during the 1964–2004 period, and comparing the primary study error rates with error rates in cases where the parties are represented by attorney-mediators. Notwithstanding these enhancements, the incidence and relative cost of the decision-making errors in this study are generally consistent with the three prior empirical studies, demonstrating the robustness of the earlier works by Samuel Gross and Kent Syverud, and Jeffrey Rachlinski. The multivariate analysis, moreover, shows that the incidence of decision-making error is more significantly affected by “context” variables (e.g., case type and forum) than by “actor” variables (e.g., attorney gender and experience level).