Estimating the Value of 'Going for It' (When No One Does)

This paper estimates success rates and the value of "going for it" on fourth down using three alternative approaches. First, the paper uses actual fourth down attempts from all regular season NFL games played in 1998, 1999 and 2000. It uses early quarter data and functional form assumptions to estimate success rates at various to go distances. It also uses all fourth down attempts and a Heckman selection model to account for the selection bias in the data. Second, the paper estimates success rates at various positions and to go distances from simulated data using Madden NFL 07. Third, the paper estimates a game theoretic structural model of fourth down attempts using actual third down data and estimated expected points. Romer claims to demonstrate that American football teams systematically err in their choice of whether to punt or "go for it" on fourth down, and uses those results to argue that firms do not maximize profits. The analysis below calls into question whether Romer indeed successfully demonstrated that football teams punt too often, which in turn casts doubt on the more general conclusions he drew from his results.