Discussant's response to sampling risk vs. nonsampling risk in the auditor's logic process;

              Discussant's Response to
Sampling Risk vs. Nonsampling Risk in the
Auditor's Logic Process
Robert K. Elliott
Peat, Marwick, Mitchell & Co.
Professor Felix' paper is ostensibly a statistical sampling paper. However, statistical and nonstatistical sampling are essentially identical, except that sampling error is quantified in the former but not the latter:
• the concepts of sampling error and nonsampling error apply equally,
• any inference must be confined to the population subject to sampling,
• if the test objective is substantive, stratification must be employed (at a minimum, all individually significant items must be audited),
• all selected items must be audited (the only rigorous alternative— "supplementary sampling"—being impractical in auditing),
• the sample findings must be projected to the population sampled, and
• sampling is highly effective against overstatement, relatively weak against understatement.
The only apparent way that statistical sampling is relevant to the paper hinges on the author's allegation that "based on discussions with a number of staff and supervisory personnel" the mere fact of quantifying the sampling error causes auditors to overlook other sources of audit risk. (Felix' sampling methods to reach this conclusion were apparently informal, but the conclusion is plausible.)
Misinterpretation of Sampling Conclusions
The possible misinterpretation of sampling conclusions that Felix is con­cerned
with was anticipated by the AICPA Statistical Sampling Committee in its 1964 Statement (Journal of Accountancy, July, 1964; now included as Section 320A of Statement on Auditing Standards No. 1):
The competence of evidential matter as referred to in the third standard of field work is solely a matter of auditing judgment that is not comprehended in the statistical design and evaluation of an audit sample. In a strict sense, the statistical evaluation relates only to the probability that items having certain characteristics in terms of monetary amounts, quantities, errors, or other features of interest will be included in the sample—not the auditor's treatment of such items. Consequently, the use of statistical sampling does not directly affect the auditor's de­cisions
as to the auditing procedures to be performed, the acceptability of the evidential matter obtained with respect to individual items in the
57