Application of Statistical Sampling to Audit and Control

The traditional literature applying statistical sampling to auditing sometimes overlooks the special structure of audit populations. Much of the literature is based on techniques developed for sample surveys. Of late there is an increasing awareness to take note of the unique environment in which audit sampling takes place and to incorporate all available auxiliary information to improve the precision of estimators. The present paper begins with a brief historical review and then focuses on the special nature of audit populations. This is followed by the description of a class of auxiliary information estimators and the occasional problem caused by situations of low frequency of errors but with large magnitudes. Next, monetary unit sampling is reviewed and key unit sampling is proposed as an alternative when the former may not apply. An outline of a Bayesian formulation to use prior information is provided. Finally, guidelines are provided for a choice of procedure enumerating the major factors to be considered.