these player transactions were Pareto optimal, indicating that NBA GMs do not
generally employ information optimally. In other words, there exists a multibillion-
dollar industry, where decisions are made beneath the intense scrutiny of millions of
interested observers, where motives are clear and consequences for failure severe, and
yet information that could effect better decisions is employed incorrectly. Such a
finding begs an important question. If an industry, where information is abundant and
motives clear, does not utilize information correctly, does any industry make decisions
strictly according to the dictates of economic theory?
In addition to the tremendous support of the editors, we wish to also acknowledge the
comments of Alexandra Bernasek, Dennis Black, Nancy Jianakopolos, and Lynn Watson. Any
errors are, of course, our own.
The value of a player depends upon the motivation of the decision maker. If one assumes
the objective is to maximize profits, then the value of a player is the rents a team accrues from
his employment, where rents are the difference between the marginal revenue product of the
player and his salary. In contrast, if one assumes the objective is to maximize wins, then the
value of a player is simply the number of wins he produces. Because it is not known for certain
whether teams primarily seek to maximize profits or maximize wins, both values will be
The model utilized herein was also employed in "Do Coaches Coach to Win? Rationality,
Resource Allocation, and Professional Basketball." ( Berri, 1997, working paper).
The relationship between teams wins and the team's accumulation of points scored and
points surrendered was examined utilizing data from the 1991-1992 season through the 1995-1996 campaign. The estimation of this relationship, utilizing a fixed effects model,
produced the following result:
Adjusted R2 = .930.|
Observation = 137.
To capture only the variation within each season, each model estimated within this
discourse will follow the fixed effects methodology. Hence, the constant term in this regression
is A1m, where m runs from one to five, representing the five seasons examined in this study.
The article authored by Zak, et al. ( 1979, pp. 379-392) presents a model that has inspired
both the works of Scott, et al. ( 1985, pp. 50-59) and Hofler and
Payne ( 1997, pp. 293-299).
Each of these studies included the opponent's accumulation of each statistic. However, by
including the opponent's statistics, it is not possible to precisely determine the impact each
player has on team wins. The model employed in this chapter, through the utilization of two
equations, proposes to connect factors tabulated for each individual player to team wins. A
comparison of the explanatory power of the Zak, et al. double log model and the linear system
of equations presented herein demonstrates that the methods utilized here are able to explain
wins as well, if not better, than the model previously offered in the literature.