There are a few 2013 incidents, but I don't think our dataset covers this time period. This gives the incorrect impression that collisions are "way down" for 2013.
Maybe it's a timezone issue? E.g. collisions that happened just before midnight on new years eve.