Models of Arbitrator Behavior: Theory and Evidence

Our purpose in this paper is to open up the empirical analysis of some simple models of arbitrator behavior under alternative mechanisms and in different economic environments. We do this by studying the outcomes of the first two years of operation of a New Jersey statute that mandates the arbitration of unsettled pay disputes by New Jersey police officers and the municipalities that employ them. This remarkable statute provides for conventional arbitration of pay disputes if the two parties can agree to this, but requires the use of final-offer arbitration if they cannot. Con- sequently, the results of both mechanisms may be analyzed and compared. Although further evidence from the New Jersey experience with an arbitration statute is necessary before firm conclusions should be drawn, several preliminary results of the empirical analysis are worth empha- sizing. First, the New Jersey system does seem to confront the parties with considerable uncertainty about the arbitration decisions they can expect. The mean of arbitrator's preferred awards is apparently closely related to alternative local wage settlements, but these decisions still contain considerable variability around this mean. Second, the union final offers have thus far been very conservative relative to the distribution of arbitrator's preferred settlements. This is demonstrated both by (a) the low values of union and employer final offers relative to the mean of con- ventionally arbitrated settlements, and by (b) the high correlation between the incidence of employer victories and the mean of union and employer final offers. The result is that a high proportion of union offers are being accepted by arbitrators and this proportion is predictable from the data on conventionally arbitrated outcomes alone. Whether this state of affairs simply reflects the risk aversion of the parties or a mistaken view by the parties of what arbitrators will allow is an important subject for further research. If the latter is the case the central tendency of future arbi- trated settlements may be considerably greater than has heretofore been seen, but employer victories are also likely to become considerably more numerous. In either case, our empirical results suggest that the form of the arbitration system adopted may have a considerable impact on the size of arbitration awards.