This paper analyzes data on union and employer rankings of different panels of arbitrators in an actual arbitration system. A random utility model of bargainer preferences is developed and estimated. The estimates indicate that unions and employers have similar preferences: in favor of lawyers, more experienced arbitrators, and arbitrators who seem to have previously favored their side. Tests of whether bargainers rank arbitrators strategically reveal no evidence of such behavior.

This study analyzes a new set of data on the decisions of conventional arbitrators.The main goal is to draw inferences about the extent to which conventional arbitration decisions are fashioned as mechanical compromises of • the parties' final offers, without reference to the exogenous facts involved, in different disputes.The results of the analysis are remarkably clear: conventional arbitrators tend to split-the-difference between the parties' final offers with virtually no evidence of systematic reference to the facts of the cases.However, since there is a substantial amount of unexplained variance in the arbitration decisions, this evidence of mechanical compromise behavior should be viewed as characterizing the overall operation of conventional arbitration mechanisms and not the behavior of individual arbitrators in any particular case.Indeed, the results are consistent with the view that individual arbitrators pay close attention to the facts of the cases, but that there is considerable variation in the structure of different arbitrators' preference functions.

This paper analyses data on union and employer rankings of different panels of arbitrators in an actual arbitration system. A random utility model of bargainer preferences is developed and estimated. The estimates indicate that unions and employers have similar preferences, in favor of lawyers, more experienced arbitrators, and arbitrators who seem to have previously favored their side. Alternative rankings models, which are estimated to test whether bargainers rank arbitrators strategically, reveal no evidence of strategic behavior.