On agglomeration in competitive location models