James J. Tucker, III WIDENER UNIVERSITY
and
Frank C. Lordi WIDENER UNIVERSITY

EARLY EFFORTS OF THE U.S. PUBLIC ACCOUNTING PROFESSION TO INVESTIGATE THE USE OF STATISTICAL SAMPLING INITIAL AICPA RESEARCH EFFORTS, LEGAL REVIEWS, AND FIRM SPECIFIC EXPERIMENTATION

Abstract: The study suggests that the AICPA’s efforts to investigate the use of statistical sampling appear to have been catalyzed by the confluence of the growing dissatisfaction with the traditional approach to sampling and the widespread recognition of the benefits of statistical sampling which were being realized in other professions and in industry. The fact that large corporations had begun to use statistical sampling in accounting and auditing lent additional urgency to the need to address the issue. In addition, some doubted the tenability of traditional sampling if challenged in court by a statistician..

The study also examines the research efforts of the AICPA which include case studies, legal opinions, a review of the mathematical underpinning of statistical sampling, and the initiation of an interdisciplinaiy collaborative effort with the American Statistical Association. The research suggests that from the very beginning, the ATCPA was most cognizant of the legal implications of statistical sampling and that the legal opinions obtained played a key role in the profession’s eventual acceptance of this new technology. The study also indicates that the AICPA’s research led it to reexamine basic auditing concepts and challenge existing standards.

The national CPA firms shared information regarding their individual inhouse experimentation with statistical sampling which suggests that a spirit of collcgial cooperation existed rather than competitionrelated secrecy. Lastly, the research demonstrates the importance of individual leadership in the profession’s efforts to adopt new technology; here, the study provides a glimpse of two Accounting Hall of Fame members, Robert M. Trueblood and Oscar Gellein, and the first recipient of the Distinguished Service in Auditing Award, Kenneth W. Stringer, making one of their many profes sional contributions. The study suggests that without the strong leadership provided by these three individuals, the profession’s adoption of this technology would have taken significantly longer.

WANTED: EXAMPLES OF USE OF STATISTICAL SAMPLING

The American Institute’s committee on statistical sampling which is exploring the possible usefulness of statistical sampling techniques in accounting and audits by independent certified public accountants, desires to obtain information as to experimentation with these techniques that may have been carried out by our readers. In particular, the committee is interested in information regarding experiments that have been conducted by smaller accounting firms or individual practitioners in their audits of small businesses. It would also be interested in having information as to any use that has been made of statistical sampling in the accounting or internal control procedures.
If any of our readers has information of this kind . . . [Journal of Accountancy, July 1957]

RESEARCH OBJECTIVES AND IMPORTANCE OF THE STUDY

The purpose of this study is twofold. The study first seeks to provide an overview of the issues and conditions that provided the impetus for the U.S. public accounting profession’s efforts to investigate the use of statistical sampling. Second, the study attempts to provide insight into the initial research efforts taken by the AICPA to determine the feasibility of using statistical sampling to improve the practice of accounting and auditing. Such insight may increase our understanding of the manner in which the accounting profession reacts to emerging issues and problems.

Specifically, the study first examines the issues and conditions which led to the formation of the AICPA Committee on Statistical Sampling (AICPACSS). Second, the study focuses upon the AICPA’s initial efforts to educate itself in a discipline in which it had little or no educational background. Insights provided by the interdisciplinary aspect of these educational efforts appear especially useful to the profession as it seeks to adapt to the ongoing fusion of traditional accounting, information systems, and computer science.

Third, the research examines the concerns of the AICPA regarding the legal implications of the use of statistical sampling, as well as its efforts to resolve these concerns. This focus seems particularly appropriate given the current litigation crisis. Also, the research attempts to provide insight regarding the profession’s level of “legal awareness” in this “prelitigation” era.

Fourth, the study presents an overview of the earliest efforts of major accounting firms regarding the use of statistical sampling as described by Oscar Gellein in 1959. This analysis is undertaken to gain a greater understanding of how individual firms react to emerging issues, as well as the magnitude of diversity of these reactions. Such insights may be useful to the profession in its efforts to encourage or require the adoption of technological innovations by its membership.

Lastly, the study documents the importance of leadership in initiating and pursuing technological innovation. Here, the research provides insight into the pioneering efforts of Robert Trueblood, Oscar Gellein and Kenneth W. Stringer, to harness new technology for the improvement of accounting and auditing practice. (Trueblood and Gellein have been inducted into the Accounting Hall of Fame and Stringer was the first recipient of the American Accounting Association’s Distinguished Service in Auditing Award.) The research also documents and recognizes these individuals contributions with the hope of facilitating a feeling of pride and continuity in a profession characterized by significant turnover; such research also may provide young professionals with role models.

RESEARCH METHODOLOGY

To provide an overview of the issues and conditions which contributed to the AICPA’s interest in the use of statistical sampling, all articles regarding statistical sampling appearing in the Journal of Accountancy for the 10year period prior to the formation of the AICPACSS and five years after its formation were examined (1947 to 1961). Second, telephone interviews were conducted with Oscar Gellein and Kenneth W. Stringer who were heavily involved in the profession’s investigation of the use of statistical sampling. (Their involvement is described later.) Third, a number of studies were examined which focus upon the history of individual firms [Allen and McDermott,

1993; O’Malley, 1990; Wise, 1982; Swanson, 1972] as well as a history of the accounting profession [Carey, 1969a and 1969b].

A special focus is placed upon two books which provide insights into the thoughts, efforts and convictions of Robert M. Trueblood [Trueblood and Cyert, 1957; Bryson, 1976]. Trueblood was the first chairman of the AICPACSS and also was the primary author of what is probably the first book written by an accountant regarding the use of statistical sampling in accounting and auditing (see Trueblood and Cyert, 1957). Trueblood’s efforts to apply statistical sampling to resolve auditing problems in the early 1950s, while he was a partner at Touche, Niven, Bailey & Smart, constitute one of the earliest experiments with and applications of statistical sampling of any major firm [Bryson, 1976, pp.109128].
Surviving documents of the earliest AICPA statistical sampling committees and subcommittees were examined at the AICPA library. One part of the study is organized around a key report delivered orally by Robert M. Trueblood, as recorded in the minutes of an AICPA Council meeting.

The authors also acquired and are in possession of the personal business files of Kenneth W. Stringer. Stringer was a member of the AICPACSS for four years (19611965) and chaired the Committee for three of those years (19621965). Stringer’s files contain all documents which he accumulated during his 24year career (19571981) at the Executive Office of Haskins & Sells. A search of Stringer’s extensive files produced several of the earliest reports prepared by the AICPACSS. One of the reports contains the first legal opinions obtained by the AICPA concerning the use of statistical sampling. The authors have been unable to find any evidence that suggests that these reports were ever published or made available to the public.

THE AICPA AND STATISTICAL SAMPLING

A number of factors provided the impetus for the AICPA’s efforts to investigate the use of statistical sampling. One important factor was the successful application of statistical sampling in other fields. In 1957, Trueblood and Cyert observed that “Perhaps the greatest impetus for a considered evaluation of the possible application of statistical sampling to auditing problems comes indirectly from the success of scientific sampling techniques in other professional fields, many of which are no less judgmentbased in character than auditing.” [1957, p.61]

To illustrate, Trueblood and Cyert note that the decision to inoculate millions of children with Salk polio vaccine during 1955 was based upon conclusions reached on statistical grounds and that testing and inspection procedures in many critical manufacturing processes involve “scientific sampling”, as do many problems of engineering design and military logistics [p. 61].

In addition, the value of statistical sampling as a management tool received broad recognition during World War II when statistical sampling become integrated into manufacturing quality control programs in the nation’s effort to supply the Allies; as Trueblood and Cyert note: “In the manufacturing field, statistical quality control came of age during World War II” [1957, p.v].

Trueblood and Cyert also provide insight regarding the manner in which the widespread use of statistical sampling in World War II “carriedover” from the factory to the office:
A considerable amount of material has been published over the years with respect to the adaptability of statistical quality control procedures to the control of clerical accuracy, as it relates to the daytoday workload of a business enterprise. This type of application has been particularly attractive to statisticians, since the regular flow of paper work can be regarded as reasonably comparable to the continuously flowing manufacturing production line. … It can be accepted that, as a general rule, the carryover of statistical control methods from the factory into the purely clerical function of accounting is feasible. There are, however, many situations in which tests for clerical accuracy are somewhat more complex than the measurement of physical characteristics in a straightforward manufacturing process. Accordingly, the forthright application of statistical quality control methods should not be accepted blindly, [p. v.]

The use of statistical sampling in the war effort was a manifestation of a broader movement toward the increased use of quantitative methods in management decisionmaking. As an early observer notes:

The post World War II period has been characterized by the development of “scientific” management techniques. The extension of operations research from military to commercial application is a prime example of the utilization of mathematical models and statistical probability theory for management decisionmaking. Certain statistical techniques offer advantages to professional accountants in their decision function, the opinion regarding fairness of the financial statement presentations. [Brown, 1961, p. 46]

Another factor which had an impact upon the development of quantitative management techniques was the growth of big business during World War II. As McQuaid observers:
The production miracle that turned the United States into the arsenal of democracy and made it the source of half the manufacturing output of the planet by 1945 was impossible without big business. . . . Big business grew during war emergency because of their comparative economic, technological, and organizational advantages. . . . Big firms were much more likely to have corporate research laboratories to develop new technologies quickly [1994, p. 15].

STATISTICAL SAMPLING AND THE INDEPENDENT AUDITOR

Another major factor which contributed to the AICPA’s investigation of the use of statistical sampling was the perceived need for a more objective and scientific approach to determining the number of items to be tested when performing various audit procedures. Traditionally, the number of items to be tested was usually determined by each individual auditor on a subjective nonstatistical basis. In 1949, Vance commented that “Accountants have long been aware of the need for an objective criterion for determining the size of the sample of accounting detail they examine in an audit.” [p. 214] One early writer observed that “there has been no method for specifying the size of sample to be taken. Therefore, the feeling that any such specification was certain to be arbitrary was inevitable.” [Shartle, 1952] In a more graphic comment, Allen and McDermott observe: “Traditional sampling had been done by an accountant’s ‘horse sense’.” [1993, p. 113] Trueblood and Cyert describe the traditional approach below:

… it must be agreed that sample selection techniques are completely based on judgement in present auditing practice. Auditors use their own opinions, checked by whatever experience and facts they may muster for each occasion, in order to determine the size of the samples with which they work, the method of their selection, and the interpretation of sampling results. [1957, p. 58]

Regarding the perceived benefits of statistical sampling, Trueblood and Cyert note that statistical sampling may be used to reduce the area in which purely subjective judgements are required and to improve the quality of, or to serve as a substitute for, the “judgmental testcheck” procedures used so widely in both accounting and auditing [1957, p. iv]. (During this era, “judgmental” was the term frequently used in place of the term “nonstatistical”; also, “testchecking” was a phrase used interchangeably with the term “sampling”.)
Dissatisfaction with the subjective nature of traditional nonstatistical sampling proved to be the catalyst which drew the attention and galvanized the efforts of Kenneth W. Stringer to develop one of the most frequently used statistical sampling approaches, probabilityproportionaltosize sampling [Tucker, 1994, p. 238]. Since his early years as an auditor, Stringer had been dissatisfied with the lack of authoritative guidelines regarding sample selection and believed that in similar audit situations there existed an unjustifiably wide variation in the extent of testing prescribed by various auditors [Tucker, 1989, p. 29].

Stringer’s dissatisfaction with the traditional “judgmental” or nonstatistical approach to sampling provided the impetus for what is probably the earliest foray into behavioral research in auditing [Tucker, 1994, p. 247]. In 1959, Stringer conducted a laboratory experiment in which 92 senior auditors were asked to determine the sample size that they would test in four different cases. The results revealed a very wide distribution of sample sizes selected by these senior auditors in each of the four cases. When the senior partners of his firm were presented with the results of his experiment, Stringer stated that they were “shocked and dismayed at the disparity that the survey showed” [Tucker, 1994, p. 248]. Both Gellein [1994] and Stringer [1994] believe that the desire for a more objective method of determining sample size was the primary reason that the U.S. public accounting profession pursued the possibility of using statistical sampling in audit practice.

Another issue that focused the AICPA’s attention upon statistical sampling was the growing use of statistical sampling by major corporations. By 1958, organizations that were using statistical sampling in accounting and auditing included Esso (Exxon) Research, MinneapolisHoneywell, United Air Lines, Lockheed Aircraft, Western Electric, Reader’s Digest and the U.S. Air Force [Hill, 1958, p. 57]. The use of statistical sampling by corporate clients required that independent auditors acquire at least a general understanding of statistical sampling and the related accounting and auditing implications. This issue will be examined in greater detail later.

LEGAL TENABILITY OF THE TRADITIONAL SAMPLING APPROACH

Since the traditional approach to determining the extent of testing was viewed by many auditors as too subjective, these auditors may have had doubts concerning tenability of the nonstatistical “judgmental” approach if challenged in litigation. This concern is reflected below:

Should auditors’ present methods of testchecking prove inadequate in a particular case, would it not be difficult for the profession to justify its failure to use a technique found to be of such material help in other professional fields? Is it possible to argue, at the present time, that the profession has adequately tested the practicability of scientific sampling and mathematical probability? What would happen if in a court proceeding involving accountants’ liability, a competent statistician were to demonstrate mathematically that the auditor’s sampling procedures or conclusions were not statistically justifiable? [Trueblood and Cyert, 1957, p.61].

Litigationrelated concerns also appear to have fueled the efforts of Price Waterhouse (PW) to develop a more objective sampling approach [Allen and McDermott, 1993, p. 113]. PW’s liability insurance had increased dramatically in the early 1950s after another national accounting firm had been sued and a $1.4 million settlement had been obtained by plaintiffs. Spurred by the threat of litigation, PW increased its efforts to improve the effectiveness of its auditing procedures. One of the results of these efforts was the development of a statistical sampling plan which PW implemented in 1958 [McDermott and Allen, 1993, p. 113]. The statistician who helped develop PW’s plan commented that statistical sampling “will provide a sampling technique which is not only objective but defensible from a legal viewpoint.” [Arkin, 1958, p. 67]

POTENTIAL FOR REDUCING THE EXTENT OF TESTING

A few early writers note that in some situations, statistical sampling “can considerably reduce the presently required volume of work” for accountants [Trueblood and Cyert, 1957, p. iii] and “reduce the amount of work done by the auditor while maintaining the risks at the present level.” [Neter, 1949, p. 397] Another author presented the results of a case study in which the time to conduct the taking of a physical inventory was reduced from 560 hours which was required for a 100% physical count to 260 hours using statistical sampling [Obrock, 1958, p. 53]. However, based upon the research conducted for this study, the authors believe that the potential value of statistical sampling to reduce testing was viewed by most as a “point of interest” rather that a primary goal.

STATISTICAL SAMPLING RESEARCH
EFFORTS OF THE AICPA
CIRCA 1956

By November 1956, the accounting profession’s interest in statistical sampling had progressed to the point that a special AICPA committee, the Committee on Statistical Sampling (AICPACSS), was formed to research the issue. The AICPACSS advised and reported to its senior committee, the Committee on Auditing Procedure. (The Committee on Auditing Procedure was the authoritative auditing standardsetting body during this period.) The newly appointed special committee was chaired by Robert M. Trueblood, who had just coauthored a book on the applications of statistical sampling to accounting and auditing [Trueblood and Cyert, 1957]. Oscar Gellein represented Haskins & Sells (H&S) on the Committee. (Gellein also was serving on the New York Society of CPAs Advisory Committee on Application of Statistical Sampling to Accounting and Auditing, which was formed in the same year.) [A listing of the AICPACSS members for the years 195665 is presented in the Appendix.]

A report presented orally by Trueblood at the Council Meeting of the AICPA on October 11, 1958 (as recorded in the minutes of the Meeting) lends insight to the AICPACSS’s early activities [AICPA, 1958a]. The first page of the report (presented below) reveals the paucity of knowledge regarding statistical sampling that existed within the profession at this time:

Mr. President, members of the Council. The Committee on Statistical Sampling has held five two day meetings since it was first organized in October or November, 1956. It is, of course, regarded as a satellite committee to the Committee on Auditing Procedures and our purpose and our function is investigative and exploratory in a rather new area.
For this reason, during the first year, our meetings were almost totally of a selfeducational nature. They were devoted to exploring the subject of statistical sampling both from the accountant’s point of view and from the statistician’s point of view. They were also devoted to studying problems involved in the possible or ultimate utilization of statistical sampling techniques as an auditing tool.

During our second year we have gone into a slightly more productive type of program. Our production is modest at best. First, we have developed a glossary of statistical terms which, in a sense, is a layman’s dictionary of such terms and a bibliography of literature on the general subject of statistical sampling. This glossary and bibliography is now in the process of production and will shortly be available to members on request. It is believed that this glossary may be of some help to those who are interested in exploring the subject.
[Regarding the glossary noted above, see reference: AICPA, 1958b.]

AICPACSS SUBCOMMITTEES

First Subcommittee

Trueblood’s report also noted that three subcommittees had been appointed within the AICPACSS to investigate specific issues related to statistical sampling. The first subcommittee was charged with examining the use of random selection without statistical evaluation of results. Two related issues addressed by the first subcommittee included the implications of using only a part of a statistical approach to sampling and “the question as to whether randomization is, in fact, desirable in sampling for audit purposes.” Concerning the forthcoming report of the first subcommittee, Trueblood noted that “this report has been informally polled and is now being formally polled” and would be submitted to the Committee on Auditing Procedure upon completion.

A search of Stringer’s personal business files produced a copy of the subcommittee’s unpublished report entitled Report of Subcommittee on Applicability of Random Sampling in Auditing, dated October 7, 1958 [AICPA, 1958c]. The report first presented a general discussion of advances in sampling by statisticians, terminology frequently used in sampling, and the concept of stratification. An important question raised in the report was whether “the use of random number tables carries with it an obligation to use statistical measurement.” The portion of the report which addresses this issue and the related legal ramifications is reproduced below [AICPA, 1958c]:

When we turn to the legal questions we should, of course, lean on the opinion of the legal profession. Mr. Fontaine Bradley said in his letter of November 15, 1957 to the Executive Director of the A.I.C.P.A.:

Testimony of a statistician to the effect that sampling techniques generally used by the profession are at variance with statistical techniques should not, standing alone, be sufficient to increase the danger of liability.

Mr. Bradley’s letter, a copy of which is attached, [the authors were unable to locate this letter] points out that the statistician would have to demonstrate to the satisfaction of a judge and jury that the sampling techniques used by the accounting profession are outmoded and immature and would have long since been discarded by any reasonableminded man. The introduction into our discussion of the conclusions of a reasonableminded man means that our legal question will not stand or fall by itself and must depend for its answer upon the logic of the conclusion otherwise reached. We assume, of course, that any conclusion reached by a committee of eleven practitioners after deliberations covering a period of years would be “reasonable” in the eyes of a judge and jury. It is the view of the committee, therefore, that our conclusions should be based on our own best judgment arrived at after sufficient study. As Mr. Bradley expressed it:

. . . the risk of increased liability resulting from giving the statistician such an additional argument is too remote to warrant its being a deterrent to further study directed toward determining the extent to which statistical techniques should be used in auditing sampling.

Specifically, we conclude that we can consider statistical selection solely on its own merits and not be obligated thereby to link our conclusion to the merits of statistical measurement. . . . the committee endorses the use of random selection as a permitted, but not required, auditing technique.

The authors believe that the results of the above legal opinion and a favorable legal opinion rendered by Saul Levy (described later) were critically important. Given the AICPA’s keen awareness of the legal implications, it is likely that if these opinions had been unfavorable, the AICPA would have ceased its investigation of statistical sampling.

Second Subcommittee

In his oral report [AICPA, 1958a], Trueblood stated that the second subcommittee of the AICPACSS had been given the responsibility of:
…. studying the literature on the broad subject of the purposes of audit sampling which question is necessarily raised when you get into the analysis of the statistical sampling movement. Some of the questions that are involved are these kinds of things:

Should the sample, the audit sample, be representative of the universe as is implied in a great deal of the audit literature?
Does the auditor, in fact, make inferences in the mathematical sense about the universe on the basis of the sample results?

How does one appraise relatively the purpose of sampling as a means of testing internal control and as a means of establishing the bonafides of accounts?

Another issue addressed by this subcommittee was the clarification of the circumstances in which a sample “should be selfsufficient for a particular purpose as distinguished from those cases in which using a sample is just one step in a long chain of events on the basis of which we come to an audit conclusion.” [AICPA, 1958a] Trueblood stated that by the nature of issues under investigation, much more work needed to be done, more studies needed to be conducted, and no report was yet available from the second subcommittee. (The authors have not been able to find any reports or other documents related to the efforts of this subcommittee.)

Third Subcommittee

Trueblood reported that a third subcommittee of the AICPACSS was charged with “. . . . examining the use of statistical sampling for the development of prime accounting data.” [AICPA, 1958a]. This subcommittee conducted two case studies (one manufacturing firm and one merchandising firm) which documented the determination of ending inventory through the use of statistical sampling rather than a 100% physical count.

The subcommittee issued a report of its findings which had already been received by the Committee on Audit Procedure. As Trueblood notes in his report, the subcommittee’s report regarding the case studies “. . . . raises certain questions to the Committee on Auditing Procedure.” [AICPA, 1958a]. Although Trueblood did not elaborate on the nature of these “certain questions”, a copy of the subcommittee’s unpublished report, dated June 6, 1958 [AICPA, 1958d], reveals four questions which were addressed by the subcommittee. These questions and the related conclusions reached by the subcommittee are examined below.

Mathematical underpinnings. The report first questioned “whether the statistical sampling method . . . furnishes a reasonable basis . . . for computing an inventory figure that is directly determinative of financial position and operating results. The Committee believes that it does.” The Committee reached its conclusion observing that statistical sampling “is designed to give results with specified degrees of accuracy and confidence and … is supported by a welldeveloped mathematical theory.” [AICPA, 1958d]

Physical Inventory. The second question addressed by the subcommittee was “whether in these cases there was a physical inventory in the sense that ‘Extensions of Auditing Procedures’ and other AICPA literature call for observation of the taking of the physical inventory.” The report noted that existing authoritative literature “implied that every inventory item should be counted at least once each year.” However, the report then observed that it was not uncommon to use various methods of estimation in taking physical inventories. The subcommittee then concluded that in each of the two case studies, a physical inventory was taken in a manner which satisfied existing authoritative guidelines. However, the subcommittee recommended that “the Committee on Auditing Procedure should consider revising the Codification to remove any implication that estimates, including those based on statistical sampling, might not be acceptable in particular circumstances.” [AICPA, 1958d].

Competence. The subcommittee then addressed the question of “what the auditor must do and what competence he must possess to satisfy himself in connection with inventories based upon statistical sampling.” The subcommittee concluded that “if a company applies statistical sampling, the auditor must in one way or the other learn about its limitations and the significance of its results. In some cases he may have to engage the services of an independent statistician to satisfy himself.” [AICPA, 1958d]
Error and Risk. Lastly, the subcommittee noted: “An essential question for the auditor is what limits of error and what degree of risk are acceptable. How much accuracy and how much confidence is the question.” The subcommittee stated that when statistical sampling is used to estimate “a prime figure in the balance sheet and that enters directly into the determination of income, . . . the auditor probably will be mainly concerned that the maximum error … is within the range of what he considers to be immaterial in the circumstances.” [AICPA, 1958d]

Concerning the question of materiality the report observes that “the maximum error that is acceptable in particular circumstances … is not for statistics to decide. This is a matter first for the company to resolve, and then for the auditor to consider in the light of the accounting tests for materiality in relation to fairness of the overall presentation.” The level of risk was addressed next by the subcommittee. The subcommittee noted that “the statistician thinks of slight risk in terms of 95% or more confidence. . . . There is no categorical answer. . . . It is like the question of what is the dividing line between materiality and immateriality. Manifestly, circumstances alter cases.” [AICPA, 1958d]

Impact and Reaction to the Subcommittees’ Reports

The research also suggests that the profession’s investigation of emerging technology forced it to reexamine and sometimes challenge basic auditing concepts and standards. Issues addressed by the first subcommittee led it to reexamine the fundamental purpose of sampling and evaluate the benefits and shortcomings of both existing methods and the proposed method. Likewise, the third subcommittee formally questioned what level of error and risk is acceptable to the auditor, what degree of competence in statistics an auditor should possess, and to what extent sampling can be used in the taking of a physical inventory. Regarding the third subcommittee’s report, Gellein stated in a H&S internal memo dated June 8, 1959: “Initially, all members of our committee concurred in the views expressed in the case studies. Later, . . . two dissents were filed (by Peat, Marwick, Mitchell & Co. and Price Waterhouse & Company). I understand that the Institute committee on auditing procedures is finding it difficult to formulate the position it will take.” [Gellein, 1959]

INTERDISCIPLINARY EFFORT

Lastly, Trueblood commented on the efforts of the AICPACSS to develop an “informal cooperative arrangement” with the American Statistical Association (ASA) “for the purpose of exchanging viewpoints.” [AICPA, 1958a] Trueblood noted that the AICPACSS had recently communicated this desire to the ASA, which responded enthusiastically with a list of five statisticians with whom the AICPACSS might seek consultation. He concluded, “At the last meeting of our committee in September it was agreed that such an informal meeting should be arranged with the American Statistical Association to further explore our mutual problems and questions.” [AICPA, 1958a].

PROGRESS OF INDIVIDUAL FIRMS CIRCA 1959

Since Gellein had been a member of the AICPACSS since its inception, he prepared a memo for the management of H&S in which he commented on that committee’s progress and any findings that might be relevant to H&S’s efforts regarding statistical sampling. The memo is dated June 8, 1959, and is the source of Gellein’s observations as described below [Gellein, 1959]. (Gellein was an AICPACSS member for five years (19561960) including one year as Chairman (1958).)

Gellein first observed that the AICPACSS had been organized for three years and that all of the national firms were currently represented except Ernst & Ernst and Touche, Niven, Bailey & Smart. The membership also included one small practitioner, two accounting educators, and a representative of the U.S. Army Audit Agency. He also noted that Robert M. Trueblood had been the Committee’s chairman for two years.

Gellein next observed that during the first year the members spent most of their time attempting to catch up with developments that had taken place in the application of statistical sampling to accounting. The Committee perceived its mission to be that of keeping abreast of developments in the field and keeping the profession informed. He noted that the Committee had compiled a comprehensive bibliography and “a glossary of technical terms cast in a form purporting to be helpful to an accountant seeking to learn more about statistical sampling.” In addition, from the beginning of the Committee’s activity, members had exchanged information about applications of statistics to accounting and auditing that had come to their attention.

Gellein then addressed the legal implications of statistical sampling. This segment of his report is reproduced below:

LEGAL HAZARD An early point of concern to the members of the committee was whether a new legal hazard might develop if auditors made use of some, but perhaps not all of the statisticians’ methods, in testing the records as a part of an examination of financial statements. For instance, if items to be tested were selected randomly (in the statistician’s sense of randomness), but the results of the sample were not evaluated in the way that the statistician would draw his inferences and the area tested later should be a trouble area, there is the question of whether conditions are created in which a statistician could qualify as a expert in judging the reasonableness of an auditing procedure. The Committee obtained an opinion of the Institute’s legal counsel on this point and several of the members discussed the matter with Mr. Saul Levy. The conclusion was that any new risk that might be created would generally be more than offset by the advantage of being able to show that the testing was done objectively and scientifically, assuming the audit conditions called for testing on such a basis. Mr. Levy seemed not to be concerned about the creation of a new risk. As a result of consideration of this matter, the Committee has submitted to its senior committee, Committee on Auditing Procedure, a statement on the applicability of random sampling to auditing.

Activities and developments of other firms regarding statistical sampling also were discussed by Gellein. He noted that his observations were based on his conversations with other members of the Committee. Touche, Niven, Bailey, and Smart (TNB&S) (the firm in which Trueblood was a partner) had experimented extensively with statistical applications in accounting and had assisted a number of their clients in developing applications. They also were making applications to auditing where “the conditions for statistical sampling are clearly present.” TNB&S was still experimenting with applications to transactions work and “to those areas of the audit intended to give satisfaction about internal control.” TNB&S took a “dim view of acceptance sampling” for the purpose of transactions testing but believed that estimation sampling could be used for a number of applications.

Price Waterhouse & Company (PW) had experimented considerably with the application of statistical sampling to auditing, with the assistance of Dr. Arkin who had served as an outside consulting statistician for several years. Dr. Arkin developed for PW an approach termed “discovery sampling” which was designed to assist in the selection of a sample with the use of especially constructed tables. This method gives measurable assurance that at least one error (accounting deficiency or variation) will be included in the test when the population from which the sample is drawn contains a designated number of errors. Gellein also noted that it was his understanding that PW was presently preparing a manual on statistical sampling for internal staff use. The manual was intended to serve as a training aid and provide the guidance needed to make applications. Sampling tables, descriptions of conditions of table applicability, instructions as to their use, and random number tables were included in the manual.

Gellein observed that Arthur Andersen & Company (AA) had only recently started a firmwide study of the application of statistical sampling and were drawing on their operations research personnel for technical advice. Gellein had reviewed three case studies that AA prepared concerning audit applications. The studies focused upon the testing of stock transfer transactions, accounts receivable confirmation, and voucher selection. It was his opinion that AA had not yet taken a position on the applicability of statistical sampling to auditing.
Lastly, concerning the other national firms, Gellein commented, “As far as I can tell, the other firms have not done more than to consider ways in which they might approach the study of the subject. Lybrand, Ross Bros. & Montgomery may have gone a little beyond this point.

Gellein’s memo reveals that large CPA firms shared information regarding their individual inhouse experimentation with statistical sampling as well as their progresstodate. This suggests that a spirit of collegia! cooperation existed among the firms rather than one of competitionrelated secrecy. The research also reveals significant variation in the efforts and progress achieved by the various firms regarding the use of statistical sampling.

CONCLUSION

The AICPA’s efforts to investigate the use of statistical sampling appear to have been catalyzed by the conQuence of the growing dissatisfaction with the traditional approach to sampling and the widespread recognition of the benefits of statistical sampling which resulted from U.S. efforts to supply the Allies during World War II. Many auditors appear to have been very dissatisfied with the traditional approach to determining the extent of testing, which they perceived as too subjective. Due to the subjective “unscientific” nature of the traditional approach, some doubted the tenability of the traditional approach if challenged in court by a statistician. The use of statistical sampling in quality control programs during World War II increased the accounting profession’s awareness of the potential benefits of statistical sampling to provide a more objective approach. Lastly, the fact that large corporations had begun to use statistical sampling in accounting and auditing lent additional urgency to the need to address the issue.

The study reveals that until the mid1950s, there was an almost complete lack of knowledge in the accounting profession regarding statistical sampling. The study also reveals the twopronged manner in which the profession attempted to educate itself. First, the AICPA appointed a committee which identified the principal issues and concepts regarding statistical sampling. The committee assembled a series of subcommittees to address these individual issues. Their research efforts included case studies, a legal opinion, a review of the mathematical underpinning of statistical sampling, a glossary of statistical sampling concepts and terms, and a bibliography. Second, the AICPACSS contacted the American Statistical Association to initiate an interdisciplinary collaborative effort. The American Statistical Association responded enthusiastically.

The research also suggests that from the very beginning of its investigation of statistical sampling, the AICPA was very cognizant of the legal implications. The AICPACSS obtained a written legal opinion from Fontaine Bradley and discussed the issue with Saul Levy. Bradley concluded that the risk of a statistician being able to successfully attack the auditor’s use of statistical sampling was too remote to deter the profession’s further study of the use of statistical sampling in auditing. Levy concluded that this risk was “more than offset by the advantage of being able to show that the testing was done objectively and scientifically.” In the authors’ opinion, the results of these legal opinions were critically important. Given the AICPA’s keen awareness of the legal implications, it is likely that if these opinions had been unfavorable, the AICPA would have ceased its investigation of statistical sampling.

The research also suggests that the profession’s investigation of emerging technology forced it to reexamine and sometimes challenge basic auditing concepts and standards. Issues addressed by the first subcommittee led it to reexamine the fundamental purpose of sampling and evaluate the benefits and shortcomings of both existing methods and the proposed method. Likewise, the third subcommittee formally questioned what level of error and risk is acceptable to the auditor, what degree of competence in statistics an auditor should possess, and to what extent sampling can be used in the taking of a physical inventory. Regarding this last issue, the third subcommittee recommended that the CAP “should revise the Codification to remove any implication that estimates, including those based on statistical sampling, might not be acceptable in particular circumstances.” [AICPA, 1958d] Concerning the resolution of these questions, Gellein stated that the CAP “is finding it difficult to formulate the position it will take.” [Gellein, 1959]

Gellein’s memo reveals that the large CPA firms shared information regarding their individual inhouse experimentation with statistical sampling as well as their progresstodatc. This suggests that a spirit of collegial cooperation existed among the firms rather than one of competitionrelated secrecy. The research also reveals significant variation in the efforts and progress achieved by the various firms regarding the use of statistical sampling.

Lastly, the study demonstrates the importance of individual leadership in the profession’s efforts to adopt new technology. Trueblood was one of the earliest accountants to identify the potential use of statistical sampling. Trueblood coauthored one of the earliest books which described applications of statistical sampling to the practice of accounting and auditing. He then spearheaded the AICPA’s initial effort to address the issue by chairing the AICPACSS for the first two years (1956 and 1957). The fact that the statistical sampling issue was brought to the attention of the profession by a person of Trueblood’s professional stature and credibility lent additional urgency to the issue and increased the visibility of statistical sampling as an issue to be addressed.

The longterm commitment and leadership provided by Oscar Gellein and Kenneth Stringer also were crucial. Gellein served as a member of the AICPACSS for the first five years and served as chairman in the third year (the year after Trueblood stepped down as chairman.) Gellein was involved from the very beginning and was very influential as the Director of Research in H&S’s Executive Office. He encouraged and supported the research efforts of Kenneth Stringer who developed probabilityproportionaltosize sampling which was integrated into H&S’s audit practice in 1962 (Tucker, 1989 & 1994). Kenneth Stringer became a member of the AICPACSS for four years (196162 to 196465), two of which he served as chairman (196162 and 196263). Stringer also was the primary author of a committee report that described what is now refer to as the audit risk model (Tucker, 1989). Without the strong leadership provided by these three individuals, the harnessing of this technology to improve the practice of accounting and auditing certainly would have taken significantly longer.

APPENDIX

Listing of the AICPA Committees on Statistical Sampling for Fiscal Years 195665
( * ) New Appointee
195657
Robert M. Trueblood, Chairman (Touche, Niven, Bailey & Smart)
Jerome Abrams (Jerome Abrams & Co.)
Carl J. Bonne, Jr. (Authur Andersen & Co.)
Thomas J. Cogan (Lybrand, Ross Bros. & Montgomery)
Russell B. Driver (Russell B. Driver & Co.)

Tucker: Early Efforts of the U.S. Public Accounting Profession to Investigate the Use of Statistical Sampling
Oscar S. Gellein (Haskins & Sells)
Henry P. Hill (Price Waterhouse & Co.)
J. Robert Malone (Peat Marwick Mitchell & Co.)
Howard F. Stettler (University of Kansas)
Lawrence L. Vance (University of California)
Frank T. Weston (Authur Young & Co.)

195758
Robert M. Trueblood, Chairman
Carl J. Bohne, Jr.
Thomas J. Cogan
Russell B. Driver
Oscar S. Gellein
Henry P. Hill
William W. Jones*
J. Robert Malone
Howard F. Stettler
Lawrence L. Vance
Frank T. Weston
195859
Oscar S. Gellein, Chairman Carl J. Bohne, Jr. Thomas J. Cogan Russell B. Driver Henry P. Hill William W. Jones J. Robert Malone William A. Newman, Jr.*
Howard F. Stettler Lawrence L. Vance Frank T. Weston
195960
Thomas J. Cogan, Chairman
Carl J. Bohne, Jr.
R. Gene Brown*
Harold Cohan*
John B. Craig, Jr.*
Russell B. Driver
Oscar S. Gellein
Henry P. Hill
Robert W. Johnson*
Dominic J. Morris*
William A. Newman, Jr.
Arch Rounsaville* Howard F. Stettler

(Touche, Niven, Bailey & Smart)
(Authur Andersen & Co.)
(Lybrand, Ross Bros. & Montgomery)
(Russell B. Driver & Co.)
(Haskins & Sells)
(Price Waterhouse & Co.)
(Ernst & Ernst)
(Peat Marwick Mitchell & Co.)
(University of Kansas)
(University of California)
(Arthur Young & Co.)
(Haskins & Sells)
(Arthur Andersen & Co.)
(Lybrand, Ross Bros. & Montgomery)
(Russell B. Driver & Co.)
(Price Waterhouse & Co.)
(Ernst & Ernst)
(Peat Marwick Mitchell & Co.)
Deputy Director, Defense Accounting
and Auditing Division, GAO (University of Kansas) (University of California) (Arthur Young & Co.)
(Lybrand, Ross Bros. & Montgomery)
(Arthur Andersen & Co.)
(Eberhart & Brown & Co.)
(S.D. Leidesdorf & Co.)
Eli Lily & Co.
(Russell B. Driver & Co.)
(Haskins & Sells)
(Price Waterhouse & Co.)
(Touche, Niven, Bailey & Smart)
US Air Force, Auditor General
(Deputy Director, Defense Accounting
and Auditing Division, GAO) Commodity Credit Corporation (University of Kansas)

114
Lawrence L. Vance J. M. Waterman*
196061
Thomas J. Cogan, Chairman Morton Backer* Carl J. Bohne, Jr. R. Gene Brown Harold Cohan John B. Craig, Jr. Peter A. Firmin* Oscar S. Gellein Newman T. Halvorsen* Henry P. Hill Robert W. Johnson Dominic J. Morris William A. Newman, Jr.
Lawrence L. Vance J. M. Waterman
196162
Robert W. Johnson, Chairman
Morton Backer
Roy Gene Brown
John B. Craig, Jr.
Marvin M. Deupree*
Peter A. Firmin
Dominic J. Morris
J. L. Roth*
Kenneth W. Stringer*
Lawrence L. Vance
196263
Kenneth W. Stringer, Chairman
Morton Backer
R. Gene Brown
Harold Cohan
H. Justin Davidson*
Marvin M. Deupree
Robert G. Ettelson*
Peter A. Firmin
J. Robert Malone

The Accounting Historians Journal, June 1997
(University of California) (Authur Young)
(Lybrand, Ross Bros. & Montgomery)
New York University
(Arthur Andersen & Co.)
(Eberhart & Brown)
(S.D. Leidesdorf & Co.)
(Eli Lily & Co.)
(Tulane University)
(Haskins & Sells)
(Ernst & Ernst)
(Price Waterhouse & Co.)
(Touche, Ross, Bailey, & Smart)
(U.S. Air Force, Auditor General)
(Deputy Director, Defense Accounting
and Auditing Division, GAO) (University of California) (Arthur Young & Co.)
(Touche, Ross, Bailey & Smart) New York University Harvard University (Eli Lily & Co.) (Authur Andersen & Co.) Tulane University International Cooperation
Administration (Price Waterhouse & Co.) (Haskins & Sells) (University of California)
(Haskins & Sells) (New York University) (Harvard University) (S.D. Leidesdorf & Co.) (Touche, Ross, Bailey & Smart) (Arthur Andersen & Co.) (Arthur Young & Co.) (Tulane University) (Peat Marwick & Mitchell & Co.)

196364
Kenneth W. Stringer, Chairman R. Gene Brown Harold Cohan

(Haskins & Sells) Stanford University (S.D. Leidesdorf & Co.)

Tucker: Early Efforts of the U.S. Public Accounting Profession to Investigate the Use of Statistical Sampling
H. Justin Davidson (louche, Ross, Bailey & Smart)
Marvin M. Deupree (Arthur Andersen & Co.)
J. Robert Malone (Peat Marwick Mitchell & Co.)
William Piercy* (Main, Lafrentz & Co.)
Marvin Tummins* (University of Virginia)

115

196465
Kenneth W. Stringer, Chairman
Saul Beldock*
H. Justin Davidson
William D. Hall*
Harry W. Kirchheimer*
William Piercy
Morton J. Rossman*
Marvin Tummins
196566
Morton J. Rossman, Chairman
Saul Beldock
William D. Hall
Harry W. Kirchheimer
C. W. Maurer*
James M. Owen*
Anton Petran*
William Piercy
A. Karl Scharrf*
Charles G. Steele*
Marvin Tummins

(Haskins & Sells) (S.D. Leidesdorf & Co.) (Touche, Ross, Bailey & Smart) (Arthur Andersen & Co.) (Arthur Young & Co.) (Main Lafrentz & Co.) (Peat Marwick Mitchell & Co.) (University of Virginia)
(Peat Marwick Mitchell & Co.)
(S.D. Leidesdorf & Co.)
(Arthur Andersen & Co.)
(Arthur Young & Co.)
(Alexander Grant & Co.)
(Louisiana State University)
(Touche, Ross, Bailey & Smart)
(Main, Lafrentz & Co.)
(Lybrand, Ross Bros. & Montgomery)
(Haskins & Sells)
(University of Virginia)

REFERENCES

AICPA, (1958a) Robert M. Trucblood, Report of Committee on Statistical Sampling, as recorded in the minutes of the American Institute of Certified Public Accountants Council Meeting, SheratonCadillac Hotel, Detroit Michigan, Saturday, October 11, 1958. (AICPA Library)
AICPA, (1958b) Committee on Statistical Sampling, Lprepared by Vance, L.L., R.M. Trueblood, and R.J. Monteverdc], Glossary of Statistical Terms for Accountants and Bibliography on the Application of Statistical Methods to Accounting, Auditing and Management Control, pp. 130 (AICPA Library)
AICPA, (1958c) AICPA Committee on Statistical Sampling, Report of Subcommittee on Applicability of Random Sampling in Auditing (October 7, 1958) pp. 13. Obtained from the personal business files of Kenneth W. Stringer which are in possession of the authors.
AICPA, (1958d) AICPA Committee on Statistical Sampling, Physical Inventory Based Upon a Statistical Sample Auditing Implications (June 6, 1958) pp. 115. Obtained from Lhe personal business files of Kenneth W. Stinger which are in possession of Lhe authors.
Allen, D. G., and McDermott, K., Accounting for Success A History of Price Waterhouse in America 18901990 (Harvard Business School Press, Boston Massachusetts 1993)
Arkin, H. A., “A Statistician Looks at Accounting,” Journal of Accountancy
(April 1958): 6669.
Brown, R. G., “Statistical Sampling Tables for Auditors,” Journal of Accountancy (May 1961): 4654. Bryson, R. E., (1976) “Robert M. Trueblood The Consummate Professional,”
(Georgia State University Press, 1976). Carey, J. L., The Rise of the Accounting profession: 18961936, New York:
American Institute of Certified Public Accountants (1969a). Carey, J. L., The Rise of the Accounting Profession: 19371969, New York:
American Institute of Certified Public Accountants (1969b). Gellein, 0., “AICPA Committee on Statistical Sampling Some Comments
About Its Work,” Internal Haskins & Sells Memo dated June 8, 1959. pp. 1
5. In possession of authors. Gellein, O., Telephone interview, April 4, 1994. Hill, H. P., “An Accountant Looks at Statistics,” Journal of Accountancy (April
1958): 5765. McQuaid, K., Uneasy Partners Big Business in American Politics 19451990,
Baltimore and London: The John Hopkins University Press (1994). Neter, J., “An Investigation of the Usefulness Of Statistical Sampling Methods
in Auditing,” Journal of Accountancy (May 1949): 390398. Obrock, R. F., “A Case Study of Statistical Sampling,” Journal of Accountancy
(March 1958): 5359. O’Malley, S. O., Price Waterhouse. 100 Years of Service in the United States
(Princeton University Press 1990). Shartle, R. B., “How a System of Continuous Scientific Sampling Is Used To
Control Accuracy in LargeScale Invoicing,” Journal of Accountancy (August 1952): 167172.
Swanson, T., Touche Ross: A Biography (New York: Touche Ross & Co., 1972). Stringer, K. S., Telephone interview, April 7, 1994. Trueblood, R. M., and Cyert, R. M., Sampling Techniques in Accounting
(PrenticeHall, 1957). Tucker, J. J., “Initial Efforts of Kenneth W. Stringer to Develop a Statistical
Sampling Plan,” The Accounting Historians Journal (in press June 1994):
233254. Tucker, J. J., “An Early Contribution of Kenneth W. Stringer: Development and
Dissemination of the Audit Risk Model,” Accounting Horizons (June 1989):
2837. Vance, L. L., “Auditing Uses of Probabilities in Selecting And Interpreting Test
Checks,” Journal of Accountancy (September 1949): 214218. Wise, T. A., Peat Marwick, Mitchell & Co. 85 Years, New York: Privately
Printed (1982).