Arbitration provides a private sector alternative to the civil court system. In addition, it may serve as a substitute for the strike in labor management bargaining. This is particularly important for workers in the public sector who may not have the legal right to strike. Under conventional arbitration (CA), the arbitrator may impose whatever settlement he sees fit. Many have argued that if arbitrators in labor disputes merely split the difference between the positions of labor and management, arbitration would have a chilling effect on negotiations. That is, each party would be reluctant to make bargaining concessions, because these concessions would lead to a worse outcome at the arbitration hearing. In response to this perceived problem, Stevens (1966) argued for the use of what came to be know as final offer arbitration (FOA). Under this procedure, each party submits a final offer and the arbitrator is forced to choose one of these two offers as the final outcome of arbitration.
In the wake of Stevens’ article, final offer arbitration was adopted by a variety of jurisdictions to govern the outcome of public sector labor disputes. A voluminous academic literature has followed. This literature has focused on comparing the two arbitration procedures, either to each other or to the strike mechanism. Some of the issues addressed in this literature include the previously mentioned chilling effect, the frequency of arbitration usage, and the quality of the settlements achieved through the arbitration process. There is a strong presumption in the literature that negotiated settlements are superior to settlements imposed by an arbitrator on the bargaining parties.
Farber (1981) pointed out that if arbitrators under CA really split the difference in a simplistic way, then parties would adopt very extreme positions in the arbitration hearing. Since this is counterfactual, he argues that arbitrators adopt a decision rule which places less weight on unreasonable offers. As a result, the offers of the parties will not diverge to extreme values. This suggests that the chilling effect of CA is not as serious as earlier commentators had claimed.
Farber and Bazerman (1986) and Bloom (1986) sent hypothetical wage disputes to actual arbitrators. This research is designed to understand arbitrator behavior under CA. Do the final positions of the parties matter, or do arbitrator’s ignore these final offers and make a decision based solely on the facts in question? Though the two studies differ in important details, both agree that the final positions of the parties to the dispute do affect the outcome of CA.
Uncertainty plays a big role in discussions of arbitration schemes. In the absence of any uncertainty about the value of the arbitrator’s most preferred settlement, FOA and CA lead to the same certain outcome. This outcome is equal to the arbitrator’s most preferred settlement (Crawford, 1979). If the arbitrator’s most preferred settlement is described by a symmetric distribution, which is common knowledge, then the final offers under FOA will be distributed symmetrically about the mean of this distribution (Farber, 1980).
Stevens argued that FOA was a strike like mechanism, due to the uncertainty it generated about the final outcome. In the presence of risk aversion and identical beliefs about the expected arbitration outcome, this uncertainty would lead to the establishment of a contract zone. Of course uncertainty about the arbitrator’s preferred settlement would also generate a contract zone under CA. Farber and Katz (1979) and Bloom (1981) analyze the effects of risk aversion and uncertainty on negotiated settlements under CA.
As Farber and Bazerman (1989) point out, under CA risk averse bargainers are exposed to the entire distribution of preferred settlements, while under FOA, they are exposed to two points, which are the offers each party submits to the arbitrator. Early discussions of the issue to the contrary, there should be no a priori expectation that FOA is riskier than CA. For a constant absolute risk aversion utility function, Farber and Bazerman report that the contract zone under CA is larger for most of the parameter combinations used in their numerical simulations. This calls into question the belief that FOA is riskier than CA.
Based on field data, Ashenfelter and Bloom (1984) report that police salaries in New Jersey have a lower variance under FOA than under CA. Though there are selection issues (the choice of procedure is endogenous), this casts further doubt on the notion that FOA is inherently riskier than CA.
There seems to be some tension in the literature regarding the quality of arbitration awards and the effects of arbitration procedures on bargaining. Most authors view it as important that the arbitration procedure lead to the creation of a sizable contract zone. If the contract zone is a single point, then there is no real bargaining; the outcome merely reflects an imposition of the arbitrator’s preferred settlement. If parties are risk averse, a large contract zone is generated through an uncertain arbitration outcome. However, a large degree of uncertainty in the arbitration outcome suggests that some of the imposed outcomes will be of low quality, i.e., very different from what the parties would negotiated in the absence of the arbitration procedure.
It is generally thought that a good arbitration procedure is one which is not used very often. As a result, comparisons of CA and FOA often focus on the expected dispute rates under the two procedures. This requires having a theory of disputes, and there is not a total consensus on what causes bargaining failure. The early literature on arbitration relied on exogenous divergent expectations about the arbitration outcome as an explanation of bargaining failure. Typically in this approach, each party to the dispute has exogenous beliefs about the likely outcome at trial which may differ from one another. If expectations diverge sufficiently, then a contract zone may fail to exist and a dispute results. As discussed above, based on simulation analysis, Farber and Bazerman (1989) find larger contract zones for CA than FOA. These contract zones are computed under the assumption of identical expectations. Farber and Bazerman’s results suggest that if divergent expectations are the cause of disputes, there will be more disputes under FOA.
An alternative approach is to assume that disputes are caused by asymmetric information. Bebchuk (1984) is an important early example from the literature on civil litigation of a model in which asymmetric information leads to bargaining failure. While this paper led to an explosion of papers on civil litigation analyzing the role of asymmetric information, only a few papers have analyzed this issue in the arbitration context. Gibbons (1988) constructs a model in which the arbitrator learns about the state of the world from the offers submitted by the parties to the dispute. Samuelson (1991) derives optimal offers in the presence of asymmetric information. Neither of these papers is attempting to explain the cause of disputes. Risk preferences are not directly observable, and Curry and Pecorino (1993) argue that asymmetric information on risk preferences can be a cause of bargaining failure. While they develop this in a model of FOA, a similar mechanism would lead to disputes under CA as well.
An interesting aspect of FOA is that negotiations can take place after potentially binding offers have been submitted to the arbitrator. These offers have the potential to transmit privately held information held by the parties to the dispute. This in turn may lower the dispute rate. Farmer and Pecorino (1998) analyze a model of FOA with asymmetric information. In the absence of a pooling equilibrium, they find that allowing for negotiation after the submission of offers to the arbitrator lowers the dispute rate. However, the ability to negotiate after submission of offers to the arbitrator makes the existence of the pooling equilibrium, in which all cases settle, less likely. Pecorino and Van Boening (2001) conduct an experiment based on this model. The experiment is calibrated so that a pooling equilibrium is not predicted. They find that allowing for negotiation after the submission of offers to the arbitrator results in a substantial reduction in the dispute rate.
Experimental analysis is one important branch of the literature on arbitration. Farber et al. (1990) conduct an experimental analysis of CA in which they simulate labor-management wage negotiations. They find, as predicted by almost all theories of bargaining failure, that settlement is increasing in the cost of a dispute. They also induce risk aversion in some bargainers, by manipulating the relationship between the settlement and the participant’s cash payoff. Risk aversion increases the size of the contract zone, but the authors only find a weak positive relationship between induced risk aversion and the propensity to settle. They also report some settlement when no contract zone exists and some settlement outside of the contract zone when it does exist. These results are typical of experimental bargaining research in a certain sense; theory generates stark predictions on behavior, while the actual behavior of experimental subjects often generates anomalous results relative to these predictions.
Ashenfelter et al. (1992) conduct an experimental comparison of FOA and CA. In contrast to some earlier research on arbitration, participants were given considerable information about the arbitrator’s preferred settlement. Among their findings are that the dispute rate is inversely related to the cost of disputes and that disputes are more likely under CA when the variance of the arbitrated outcome is lower. The authors also report that dispute rates are higher in FOA than in the comparable treatment for CA. This work contrasts with some earlier experimental research (e.g., Neale and Bazerman, 1983) which had reported lower dispute rates under FOA.
There has also been a great deal of research which analyzes arbitration using field data. Delaney (1983) finds for teacher salaries that the ability to strike and the availability of arbitration have similar positive effects on teacher salaries. They each raise salaries by about 10%. On the other hand, he does not find a systematic relationship between the use of the strike or arbitration mechanisms and teacher salary. Using a large Canadian data set, Currie and McConnell (1991) find that the right to strike raises public sector wages 2% and that a switch to compulsory arbitration raises wages an additional 1-2%. Thus, in contrast to Delaney, they find that wages are higher under an arbitration mechanism than under the strike mechanism. They also report that the probability of a bargaining impasse is 13 percentage points lower in right to strike jurisdictions than in jurisdictions covered by mandatory arbitration. However, because the cost of a dispute is so much lower under arbitration, they conclude that total dispute costs are lower under the arbitration mechanism.
Using data on Canadian teachers, Currie (1994) finds that negotiated and arbitrated wages have a similar mean. This agrees with the findings of Delaney. However, she also finds that the variance of arbitrated wages are lower than the variance of negotiated wages. She interprets this to mean that negotiated wages reflect information held by the bargaining parties, but not known to the arbitrator. These results call into question the quality of arbitrated outcomes.
Olson and Rau (1997) examine how the bargaining parties learn about arbitrator preferences from a prior experience of having proceeded to arbitration. They find that the variance of the subsequent negotiated settlement falls by 30%. They attribute this to learning by the parties to the dispute. In addition, subsequent to an arbitrated outcome, negotiated settlements more closely reflect the decision criteria of the arbitrator.
Arbitration continues to be a rich area for both theoretical and empirical research. In addition, it is an area where academic writings, in particular Stevens (1966), have had a major impact on actual practice. While not as influential as Stevens’ call for the use of FOA, the literature has produced several other proposals for new arbitration procedures (e.g., Brams and Merrill, 1986).
