ERROR BOUNDS FOR COLLECTIVE APPROXIMATIONS OF STOP-LOSS PREMIUMS