An Examination of the Status of Probability Sampling in the Courts
Brigham Young University
University of Oregon
Auditing researchers, staggering under an ever-increasing blizzard of pronouncements,
articles, and memoranda, might well be disposed to believe that complexity feeds on itself. For example, two themes which were of much concern—
perhaps even dominant—at the first two University of Kansas Auditing Symposia1 were the increase in litigation against auditors and the growing sophistication of statistical sampling techniques in auditing. This paper attempts to examine the interaction between the two.
Indeed, his paper arose directly as a result of the 1974 Symposium. At that conference, one of the present authors raised the question of whether statistical sampling (i.e., probability sampling) would be a better defense in the courts than judgment sampling. There appeared to be some division among practitioners
present at the conference.
Some felt that the use of probability sampling would ameliorate the position of the defense in a lawsuit, since probability sampling is viewed as more "scientific,"
as encompassing more up-to-date technology, and as more susceptible to mathematical "proof" of its validity.
Others were more skeptical or at least agnostic. They felt that expert witnesses
might debate the merits of a particular probability sampling plan to the ultimate utter confusion of jurors and jurists. Their contention was that the "expert judgment" of a highly trained professional, on the other hand, was less suspect and less susceptible to point-by-point rebuttal.
Our purpose in this paper is twofold. First, we will examine past court decisions to discover whether the courts have shown any preference for probability
sampling. Then we will examine cases in which probability sampling was used, either by one of the parties or by the courts themselves, to see what implicit standards for such sampling may be emerging from the judicial process.
For readers who are curious about the outcome of these questions, yet less than enthusiastic about wading through the details of this paper, we will admit at the outset that our conclusions are more equivocal and more tentative than