Sunk costs in the NBA: the salary cap and free agents

The previous research on the role sunk costs play in the National Basketball Association (NBA) has come to mixed conclusions. We use a direct measure of sunk costs, player compensation, and a quasi-experimental method that considers the endogenous relationship between compensation and productivity. Using the spike in salary cap for the 2016-2017 season, from new television broadcasting contracts, we find compensation has a significant effect on playing time. Specifically, we use players signing free agent contracts in 2015-2016 or 2016-2017. For both groups, we analyze the change in salary from the pre-contract season to the post-contract season. Using the difference-in-differences, we find the inflated salary cap, from the new television contracts, increased player compensation by 81.7%, on average. Instrumental variables estimations show the increase in compensation significantly affects playing time. The change in salary yields an additional 1.93 min played per game, which is approximately equal to the effect of a one-standard deviation increase in contemporaneous productivity. Furthermore, we find that our conclusions are robust to the use of various advanced analytical measures or traditional box-score statistics.