One imperfect, but perhaps worthwhile method would be to limit your sample to only the metropolitan areas that can be identified in all of the years from 2005 - 2016. You can see which metropolitan areas these are if you follow the procedure described in this answer. This will allow you to aggregate data along the geography variable METAREA, but only for those areas which are identifiable in all years your are analyzing.