Abstract
This project looks at the 1994-2006 years of the NFL Draft. These years were selected because 1994 was the first year of the 7 round draft and 2006 is the most current year where almost all the players are now retired. It would be expected that the talent drafted would decrease mostly linearly across the picks. Specifically, one would expect the last pick of the first round and the first pick of the second round to be similar. To analyze whether this is the case a regression discontinuity was used. Overall draft pick number, a dummy variable for the seven rounds, and five separate variables for the dependent variable will be used to run the regression. The draft pick number will control for the talent decrease and the round variables are used to pick up any of the discontinuities. The five dependent variables chosen were games played, Career Approximate Value, seasons started, probowls selected, and career length. The results from these regressions showed that there are large significant negative discontinuities at the first to second round change for all five output variables and in the second to third round change in four of the five. For career length there is also negative discontinuity at the fifth to sixth round change. As the initial investment of the pick decrease in the later rounds the decrease becomes more linear. These results show that General Managers and Coaches make decisions partially based upon sunk costs and round drafted and not solely upon talent.
Sports and Society Statement
My project is on regression discontinuities in The NFL Draft. This project is significant in the fact that it provides proof that GM's and coaches base decisions on who to play partially on sunk costs and the selected round of a player and not solely talent.