Abstract A recent article in "The Accounting Review" makes a commendable effort to simplify statistical sampling procedures for use by accountants and auditors. Unfortunately it contains a number of mistakes and misunderstandings which seriously limit its usefulness. The suggestion that the tolerances or acceptable amount of error should be increased is questionable. While the original determination of tolerances may have been unduly strict, one should not abandon them merely to make his work lighter. The passage just quoted makes no distinction between tolerance limits and limits of immateriality. Tolerance limits should be set by considerations of both risks due to sampling variability and materiality. Statisticians have developed several improvements in sampling technique which increase the accuracy of a sample of a given size or increase the efficiency of sampling in other ways such as reducing the cost of attaining a desired degree of accuracy. The article utilizes one such technique, stratification, but here again there are several flaws. The gains from stratification come from the reduction of variation within parts by grouping into each part accounts that are similar in size.
Frederick F. Stephan (Fri,) studied this question.