This note reports the results of an experiment which was designed
to test Rubinstein's(1982) theory of bargaining. We were particularly
interested in how it would compare with the hypothesis that bargainers
tend to split a pie 50-50. We duplicated Binmore, Shaked and
Sutton's(1986) result that the equal split hypothesis is rejected in a two
round game with alternating offers. However, we show that in similar
games with more than two rounds Rubinstein's theory is also rejected.
Thus their conclusion, that subjects behave as "gamesmen" (i.e. in a
manner consistent with the predictions of game theory), was premature.
In experiments with varying numbers of rounds, our first players
consistently offered their opponents shares equal to the value of the
second round pie. In a two round game this behaviour by definition
yields offers consistent with Rubinstein's theory. In games with more
rounds it does not.
In each game, the majority of first players chose to make the same
offer. In fact, the regularity of their behaviour is perhaps our
strongest result. While neither Rubinstein's theory nor the equal split
model explain our findings, the regularity of our subjects' behaviour
suggests that there is hope of finding a model of bargaining which
does.

This paper reports the results of a systematic experimental comparison
of the effect of alternative arbitration systems on dispute rates. The key to
our experimental design is the use of a common underlying distribution of
arbitrator "fair" awards in the different arbitration systems. This allows
us to compare dispute rates across different arbitration procedures where we
hold fixed the amount of objective underlying uncertainty about the
arbitration awards.
There are three main findings. First, dispute rates are inversely
related to the monetary costs of disputes. Dispute rates were much lower in
cases where arbitration was not available so that the entire pie was lost in
the event of a dispute. This confirms the empirical importance of the
so-called "chilling effect" on bargaining that has been conjectured is
produced by the adoption of arbitration systems. Second, the dispute rate in
a final—offer arbitration system is at least as high as the dispute rate in a
comparable conventional arbitration system. Contrary to the usual argument,
we find no evidence that final-offer arbitration eliminates the chilling
effect. Third, dispute rates are inversely related to the uncertainty costs
of disputes. Dispute rates were lower in conventional arbitration treatments
where the variance of the arbitration award was higher and imposed greater
costs on risk-averse negotiators. Our results can also be interpreted as
providing tentative evidence that the negotiators were risk—averse on
average. Finally, we find general agreement between the dispute rates in our
experiment and dispute rates found in the field in comparable settings.