A note on comparing the unrestricted and restricted least-squares estimators