The prediction of car ownership and use revisited: the beginning of the end?