Development of a Win Production Function and Evaluation of Cross-Sectional Dependence

Abstract

We consider correlations between pairs of teams when deriving a win production function for professional baseball teams. Previous empirical studies derived win production estimates based on the assumption that the error terms were independent, but this is not reasonable given that a win is a relative output. The total number of wins per season is fixed: all teams share wins, in that a win for one team is a loss for the opponent. We applied spatial panel data models to a Major League Baseball dataset to control for cross-sectional dependence. A strong negative dependence was found; our estimates differ somewhat from previous estimates that assumed cross-sectional independence. JEL: L83, L80, Z20