Because at the core of the point is a comparison between a very large data set (auto fatalities) and a very small data set (large scale terrorist attacks). Since - thank god - there is a very limited data set available in the second case it's worth asking if the expected value is higher than the observed value up til now.