A series of experiments compares the bargaining behavior in a Rubinstein game under three different settings: i) no arbitration, ii) conventional arbitration, iii) final offer arbitration. In the no arbitration treatment, dispute rates with a zero payoff for the bargaining parties were under 10%, while the pie was equally splitted in less than 40% of the cases. Introducing the safety net of arbitration has a significant impact on bargaining behavior. In the conventional arbitration treatment dispute rates increased to over 40% of the cases while at the same time equal splits are observed in more than 50% of the cases. In the final offer arbitration there is only a small increase of dispute rates (compared to the no arbitration treatment) while agreements according to an equal division of the pie have doubled to almost 80%.

The first result - true for all kinds of arbitration - is that the average size of the pie was larger than in the no arbitration treatment because with arbitration the bargainers had not to face zero payoffs when they broke off negotiations.

The second result is that under conventional arbitration the dispute rates have sharply increased due to the chilling effect. This can be avoided when the arbitration is designed as a final offer arbitration. The dispute rates were about the same in the final offer arbitration as in the basic treatment without any arbitration. However, this advantage comes at certain cost leading to a trade-off between the two arbitration designs. The change in the bargaining behavior is highest under the final offer arbitration. Weak bargainers increase, strong bargainers decrease their first claims because the latter ones want to avoid disadvantageous awards by the arbitrator.

It is also important to know whether the arbitrators were influenced by the bargainers. In the conventional arbitration treatment the arbitrators awarded the pie in three out of for cases according to the split the difference notion. In every fourth case they solved the dispute by an equal split. For the conventional arbitration this means that there are two contradicting effects, one in favor of the weak (more equal splits when negotiations ended in agreement), and one in favor of the strong bargainer (unequal splits by the arbitrator).

In the final offer arbitration both effects are in favor of the weak bargainer. For the most arbitrators the more 'appropriate' award will be around the equal split. They prefer an equal to an unequal and a less unequal to a more unequal split. In this case it is going to be the weak bargainer who will profit from an arbitration award.

Due to its impact on bargaining behavior, the final offer arbitration is therefore the tightest safety net for weak bargainers while bargaining costs are ceteris paribus smallest under conventional arbitration.