A.P.O. rules are asymptotically non deficient for estimation with squared error loss