New York closing-statement data provide unique insight into settlement and selection. The distributions of settlements and adjudicated damages are remarkably similar, and the average settlement is very close to the average judgment. One interpretation is that selection effects may be small or nonexistent. Because existing litigation models all predict selection bias, we develop a simple, no-selection-bias model that is consistent with the data. Nevertheless, we show that the data can also be explained by generalized versions of screening, signaling, and Priest-Klein models.