the F-statistics for the hitters' and pitchers' models are 1.25 and 1.80, respectively. Thus, the hypothesis that the arbitration-eligible dummy and interaction terms are equal to zero cannot be rejected at the 1% level. Hence, one can conclude that after controlling for experience, the relative weights placed on exogenous factors in the two groups are comparable. 6
The purpose of baseball's salary arbitration is to create an incentive for players and clubs to negotiate. Because negotiated and arbitrated settlements will be driven by the preferences of the arbitrators, it is important that the source of these preferences be investigated. Theory suggests that the arbitrators' notions of fairness will be largely based on the parties' own fairness assessments. Thus, given that baseball's collective bargaining agreement directs arbitrators to utilize comparative salaries in rendering decisions, it is likely that arbitrators would use the free agent salary determination model to infer fairness.
This study suggests that arbitrators try to replicate salary determination in the free agent market. However, they also have incorporated an experience factor, separate from performance, into their preferences. Thus, a first-year eligible player cannot expect to earn as much as a free agent with identical performance statistics. Given the well-documented gap between ineligible and free agent salaries, the experience factor may reflect an effort by arbitrators to compromise between the union's desire to accelerate salary growth and management's wish to restrain it.