Sunk Costs In The NBA

In recent years, social scientists have started performing more research on sports organizations. For example, the Journal of Sports Economics was started in February 2000 with the intent of publishing academic research on sports by economists. One of my favorite sports economics articles was written by Barry Staw and Ha Hoang titled “Sunk Costs in the NBA: Why Draft Order Affects Playing Time and Survival in Professional Basketball.” The article appeared in the September 1995 edition of Administrative Science Quarterly, a leading social science research journal. I enjoy the article because it investigates a favorite topic of economists â€" sunk costs â€" in the setting of my favorite professional sport: basketball.

“Sunk costs” is a term used by economists to describe costs that have been incurred and are irrecoverable. Once these costs are made, they cannot be “unmade”, and so future decisions should not take into account the size of the initial cost. Economists often complain that companies take actions that result from the sunk cost fallacy: managers often appear to act under the mistaken impression that if a plant cost a lot of money to build, then it should be operated even if it is losing money. In the context of the NBA, former draft picks can be thought of as sunk in the following sense. Certainly future draft picks are valuable, and often function as a form of currency to help trades go through. However, once the pick is used to select a player in the draft, the pick is irrecoverable. There is no way to “do over” the pick once it has been made. Hence, draft picks represent a sunk cost.

Staw and Hoang ask an interesting question. Given that we can think of draft picks as resources that teams spend on acquiring new players, do we see any evidence of sunk cost fallacy? In other words, do we see teams playing high draft pick players more than they should, simply because they are high draft picks? The statistical tests the authors use show that players who were selected earlier in the draft are given more playing time than players who were selected later in the draft, after controlling for a number of factors including player position, on the court productivity, etc. In other words, the authors use statistical methodology to show that if there were two identical players of exactly the same quality, but one was picked early in the draft and one was picked late in the draft, the one picked earlier would get more playing time. This does not make rational sense; clearly the two players should get the same amount of playing time.

The authors’ finding is dependent on the validity of their statistical tests, and it seems that there may be two additional variables that should be controlled for. They used data from the 1980-1986 NBA drafts, so this was before the era of high school players jumping straight to the pros and before the era of salary caps and luxury taxes. However, players could still be selected after one to four years of college, and arguably the amount of time a player spent in college is an important variable. Players that enter the NBA draft after one or two years of college are those considered to be of high potential talent, but sometimes that talent takes years to mature. The Celtics’ own Rajon Rondo might be an example of such a player. On the other hand, players that enter the draft after a full four years of college are those that have already developed into solid players and are ready to contribute right away. The Celtics’ Ryan Gomes is an example of this type of player, as is the Knicks’ David Lee. It is conceivable that managers and coaches may want to play the less experienced players more so as to allow them to develop. This does not seem to be the case with the Celtics’ own Doc Rivers, but nevertheless years in college may be an important variable to add to the statistical tests.

A more subtle but perhaps more important effect that should be controlled for is the effect of a player on the play of his teammates. Statisticians already account for easily measurable things like points, rebounds, blocks, steals and assists, but do not yet account for things like “missed defensive assignment” or “movement off the ball”. These things are hard to quantify, of course, but can have a positive or negative effect on teammates and game outcome. Thankfully, there are now various composite measures that help capture these types of variables. The Roland Rating (found at 82games.com) calculates a score for each player based on how the team does when that player is on the floor versus off the floor. The inclusion of such a variable in Staw and Hoang’s analysis would help account for these difficult to measure effects on team play.

Despite the authors not including the two variables mentioned above, they make a valid point. Namely, past decisions should not influence future decisions. In summary, consider the case of Mark Blount as an example of how this concept relates to the Celtics. Following a productive year Mark Blount was signed to a large contract. Soon after, it became clear that Blount was not a good fit in Boston. If Danny Ainge was more concerned about his ego instead of the health of the Celtics, he might have tried to keep Blount in green, pushing Doc to give him more playing time, and doing his best to show critics that the signing was not a mistake. Instead, Ainge decided to trade Blount to Minnesota (with a package of other players), thereby tacitly acknowledging that he made a mistake, but that the mistake was not going to hinder future decision making.