the clubs. This has led many to suspect that the arbitrators are not basing their
decisions on the merits of the case but, rather, that they are randomly
distributing victories between the two sides to avoid veto.
In fact, the nature of the arbitration game leads each side to submit a
reasonable offer. Therefore, the more closely the opposing offers are to the
arbitrators' preferred awards, the more it appears that the arbitrators are simply
Our results show that arbitrators do not flip coins, but rely on the facts of the
cases when rendering decisions. Further, although negotiated settlements appear
to be unbiased estimates of the preferred awards for pitchers, such does not
appear to be the case for hitters.
Another study by
McAllister ( 1992) attempted to model baseball salary
arbitration preferred awards. Although they claimed to have uncovered the arbitrators'
decision rule, they were unable to get their qualitative choice estimation to converge.
In the context of maximizing one's expected gains from arbitration, if FOA
decisions truly were random, the incentive for each side would be to submit more
extreme offers. In fact, given that baseball's opposing offers are highly correlated with
one another (see Appendix 8.1
), casual evidence suggests that the parties believe the
arbitrators consider the merits of each case.
From 1974 until 1987, two years of service were required for eligibility. After 1987, three years of service were necessary. Finally, beginning in 1990, all three-year
players were eligible, and of those with between two and three years of experience, the
17% with the most service were also eligible.
The exception is for those players with at least six years of experience. Although
some of these individuals may also be eligible for arbitration, the decision to invoke the
procedure must be mutually agreed on by the player and his club.
Marburger ( 1993) showed that arbitrators may think alike even if they do not
overtly attempt to replicate each other.
Prior to 1981, there is evidence of significant arbitrator learning in the context of Gibbons ( 1988). Specifically, the Bruce Sutter case in 1980 dramatically increased the
earning potential of arbitration-eligible players relative to ineligible players. The Steve
Kemp case one year later reinforced the role that free-agent salaries would play in
Our estimation included a set of annual dummy variables. The coefficients for the
constant term and the yearly dummy variables are not shown in the tables, but are
available on request from the authors.
The positive values of the log likelihood functions in Table 8.2
are not erroneous.
The log likelihood values of the unconstrained regression models for both hitters and
pitchers were positive. This is possible, albeit uncommon, when the value of d is less