With less than three minutes to play in the Super Bowl, Tom Brady and the Patriots needed a touchdown to win the Super Bowl. Last year Brady and the Patriots overcame a 28-3 deficit to win their fifth Super Bowl. So many people watching were expecting yet another Brady comeback.
Two plays later, though, disaster struck. The Eagles sacked Brady and forced a fumble.
A few plays later the Eagles kicked a field goal to give them an eight-point lead. But again, Brady got the ball back with over a minute to play.
Despite being given another chance, though, Brady couldn’t find any magic. After not making much progress to move down the field a last second Hail Mary fell to the ground in the end zone. And with that, Brady and the Patriots went home losers.
So, what happened to Brady?
If you think you know, check the data. As the very last pages of Sports Economics reports, quarterbacks – no matter how you measure performance – are very inconsistent. At least, relative to athletes in basketball, what a quarterback does from season-to-season is hard to predict.
The reason for this is simple. A quarterback’s performance depends on his teammates. A quarterback requires linemen to block and receivers to catch for a pass play to work. It also helps to have a reliable running game. Plus the decisions of offensive coordinates – who often call the plays – also matters.
A few things could have happened on Brady’s fumble. Maybe his line could have blocked better. Maybe his receivers could have done better to get open. Maybe the play could have been different. Or maybe everyone else was playing well and Brady simply didn’t pass the ball quick enough and/or didn’t maintain ball security.
All of this means that researchers who wish to use American football data to answer questions in labor economics have a problem. Researchers wish to use data from sports to measure the marginal productivity of athletes. But because the stats we use to track individual performance in American football depend on the player’s teammates, this data really can’t be used to measure marginal productivity. So, researchers who need marginal productivity data to answer economic questions… well, they might want to look for another sport!