Why might the first method give a closer average than the second method?