James J. Tucker, III WIDENER UNIVERSITY
INITIAL EFFORTS OF KENNETH W.

STRINGER TO DEVELOP A STATISTICAL SAMPLING PLAN

Abstract: In 1981, the Auditing Section of the American Accounting Association selected Kenneth W. Stringer to become the first recipient of the Distinguished Service in Auditing Award. Stringer was a pioneer in the auditing research efforts of Haskins and Sells for nearly 25 years. One of Stringer’s many contributions was the development of a statistical sampling plan which was adopted by Haskins & Sells in 1962. The plan developed by Stringer is referred to in the literature as ProbabilityProportionaltoSize sampling. This study provides insight into facets of Stringer’s research efforts which include his evaluation of prior sampling plans, a behavioral laboratory experiment, the interdisciplinary process through which the plan was constructed, and the influence of Oscar Gellein, whom Stringer viewed as his mentor during his early career.

In 1981, Kenneth W. Stringer became the first recipient of the Distinguished Service in Auditing Award; this prestigious award is sponsored by the Auditing Section of the American Accounting Association. The award to Stringer was for his pioneering efforts in the auditing research at Haskins & Sells (H&S) for nearly 25 years (Haskins & Sells became Deloille & Touche after a merger.) One of the many contributions made by Stringer to the theory and practice of auditing is the development of a statistical sampling plan which was adopted by H&S in 1962. This plan was originally referred to as the Haskins & Sells Sampling Plan (hereafter referred to as the Plan). The Plan developed by Stringer is frequently referred to in the literature as ProbabilityProportionaltoSize sampling (PPS). Certain features of the Plan are still the subject of much auditing research

* The author is grateful to Gary John Previts for his efforts in the initiation of the study. The author also thanks William R. Kinney, Jr. for his helpful comments regarding the topical organization of the research. Lastly, the author is grateful to E. Judson Trueblood for his helpful comments and editorial assistance on earlier versions of this paper.

[Grimlund, 1988a and 1988b; Hall et ah, 1989; Hansen, 1993; Ko et al, 1988; Plantel et al, 1985]. This study documents Stringer’s efforts to develop the Plan from the initial research phases to the submission of his proposed Plan to the firm. A brief summary of Stringer’s personal background and early career is also presented.

Other professional contributions of Stringer include the Statistical Technique for Analytical Review (STAR), which is an application of regression analysis to the analytical review process [Stringer, 1975; Stringer and Stewart, 1986], one of the first generalized audit software packages — “AUDITAPE” — and the audit risk model which provides a systematic means of managing audit risk by structuring an orderly synthesis of the component parts.

RESEARCH OBJECTIVES AND IMPORTANCE OF THE STUDY

This study attempts to achieve five major objectives. First, the research seeks a greater understanding of the inherent benefits and limitations of the Plan by a) examining the issues and problems associated with prior sampling techniques which were instrumental in catalyzing Stringer’s effort to develop the Plan and b) describing the process through which the Plan was developed.

Second, the study examines research conducted by Stringer to gain insight into the extent of divergence in various auditors’ judgements regarding their choice of sample size and the extent of testing, in similar or identical audit situations. The study also examines the impact of his research results upon the firm’s policy regarding the need for the Plan.
Third, the research attempts to provide insight into the interdisciplinary process through which Stringer harnessed “outside” technology to improve the practice of auditing. Specifically, the study examines the circumstances which forced Stringer to form an interdisciplinary collaborative effort, the problems he experienced in forming the collaboration, and the achievements that resulted. Insights provided by his interdisciplinary efforts appear particularly useful to the accounting profession’s effort to adapt to the current ongoing fusion of traditional accounting, information systems, and computer science.

Fourth, the study examines the policies and attitude of top management regarding encouragement and funding of technological innovation and its impact upon Stringer’s success. Today, this issue is particularly important given the effect of competition and litigation upon the financial resources of many firms.

Lastly, the study also examines how Stringer was influenced by the protegementor relationship that developed between him and Oscar Gellein. The potential value of mentoring is an issue of current interest to the accounting profession and is the subject of recent studies [Dennis, 1993; Alter, 1991; Viator and Scandura, 1991; Pillsbury et at, 1989]. Insights provided by Stringer’s experience should contribute to an increased understanding of the potential value of these professional relationships.

RESEARCH METHODOLOGY

Research methods employed include a series of interviews with Stringer which were taped and transcribed resulting in 226 pages of conversation as noted in the reference section [Stringer, 1993]. The narrative portion of this study, including all direct quotes of Stringer, are taken from these transcriptions. The author also was permitted to examine the surviving written documents generated by Stringer. These include numerous memos, correspondence, reports, and manuals.

STRINGER’S BACKGROUND

Born in the small rural town of Birmingham, Kentucky (population, approximately 300) on February 23, 1918, Stringer was the only child of Amos and Elizabeth Allison Stringer. Amos was a barber who operated a local shop.

Advertisements for accounting courses offered by college correspondence schools such as La Salle Extension and International Business School provoked Stringer’s first curiosities and interest in the possibility of pursuing a degree in business. As to his decision to major in business he notes:

I suppose it was sort of the sense that business was exciting and a little bit different than the usual career choices. A degree in accounting was within the parameters of my horizons back then. It seemed somewhat challenging. As I look back, I never regretted the decision. I’m sure I made the right one.

Education and Early Career

In 1934, Stringer entered Bowling Green College of Commerce located in Bowling Green, Kentucky which was about 100 miles from his home. (This college later merged with Western Kentucky University.) While an undergraduate, Stringer served as president of a social fraternity and was a charter member of the local chapter of Beta Alpha Psi. He also was recipient of the Betty Austin Cup which was the school’s annual award for the most outstanding student.

Upon graduation in May, 1938, with a B.S. in Commerce (major in Accounting, minor in Business Administration), Stringer assumed the position of staff accountant at the Kentucky Public Service Commission, a state regulatory agency. He soon grew disenchanted with his assignments at the Public Service Commission due to what he describes as a lack of creative challenge.
While reading the employment section of the local newspaper, he read that Haskins & Sells (H&S) was sending someone from their Cincinnati office to open an office in Louisville, Kentucky. “I simply walked into the office and asked for an interview. That’s all there was to the recruitment process”, he recalls. In December, 1939, he joined the Louisville office of Haskins & Sells as a junior staff accountant.

As the country became engulfed in World War II, Stringer left H&S to serve in the war effort and was discharged in August 1946. During the war years, Stringer developed the notion that he would like to start his own practice for the independence and managerial discretion that is usually associated with selfemployment. Returning to civilian life, he decided to join Robert Killebrew who was the sole proprietor of a public accounting firm in Danville, Kentucky.

The first three years proved to be very challenging and provided many opportunities for creative activity as Stringer began building his clientele. He derived much satisfaction from managing the entire operation. However, during the fourth year, it was becoming clear that the challenge and excitement were beginning to wane. Though financially successful, he began to consider career alternatives.

Rejoining H&S

The diversity of assignments and professional challenge he had enjoyed earlier with H&S were still appealing. However, based on his prior experience with the firm and through subsequent discussions with H&S representatives, he knew that if he were to return to H&S, he initially would have to take a substantial cut in salary. He also was aware that if he eventually were admitted to the partnership, the financial rewards would be substantial.

After much thought and encouragement from his wife Catherine, he decided to return to H&S; as Stringer recounts, “I decided that I really should just go ahead and make the break and go for the brass ring.” Given the initial financial disincentives for his return to H&S, the accounting profession has been the beneficiary of the high value that Stringer placed upon professional challenge, intellectual stimulation, and a creative environment.

In January, 1952, Stringer resumed his career with H&S in the Cincinnati office. In 1954, he was promoted to manager. While working in the office of a client early in the summer of 1957, Stringer received a call from H&S’s Executive Office (EO) in Manhattan requesting that he accept a two year assignment inEO.

When EO requested that he accept a two year assignment, he replied that he first would like to meet in New York to discuss the assignment. He remembers, “I had some hesitancy in making the transfer to be very honest; I didn’t know how I’d like New York and I just didn’t know what to expect.” After a reassuring meeting with Oscar Gellein, Larry Walsh, Weldon Powell and John Queenan, Stringer decided to accept the assignment. These discussions indicated to him that he would be given relatively broad latitude to explore any ideas for possible improvement in the firm’s auditing policies and procedures. Stringer remembers being both excited and awed by the challenge of the assignment. Equally important, he felt comfortable with the people and the environment.

Since flying was not yet the most frequent mode of long distance transportation, the Stringers took a train to their new residence in New Jersey in October, 1957; as Stringer recalls, “I remember the day well since the Russians launched their first Sputnik the same day I launched my new career in EO.

STRINGER’S INITIAL EFFORTS TO CONSTRUCT THE PLAN

Oscar Gellein was hired by H&S in 1953 to direct the research efforts of the firm. It became apparent to Gellein that there was a need for a broad review of current developments in auditing. He expanded the firm’s efforts to provide guidance to those in the field with an emphasis upon increased internal communications. Gellein began to make inquiries to various practice offices concerning the identification of those who had a special talent for development and application of auditing procedures. He became aware of Stringer (who was then a manager) and met with him at the firm’s 1956 meeting of managers. When Stringer arrived at the EO, John Queenan was the managing partner of H&S and Weldon Powell was the senior technical partner and the second highest ranking partner in the firm.

Upon arrival in New York, Stringer spent a month outlining and gathering information on what he viewed as the “totality” of his assignment. The assignment, broad in its scope, was to conduct a review of H&S’s overall approach to the audit and to examine a number of technical procedures. While in practice at the Cincinnati office, Stringer had become convinced that two aspects of the audit process warranted immediate attention, even though he believed that his firm was already more sophisticated than most firms in these practice areas.
His first concern was the overall approach of evaluating a system of internal control. Second, he was dissatisfied with the process of determining the extent of testing based on the initial evaluation of internal control. He then decided that of all the auditing issues and problems identified, “the most pervasive was the process used to determine the extent of testing and the selection process of items to be tested.” Stringer pursued his interests in these two areas since he believed these issues to be generally compatible with his assignment as noted above.

EXISTING SAMPLING TECHNIQUES

After making a preliminary evaluation of the system of internal control, the auditor is faced with the task of determining the extent of testing. Typically, the auditor selects a sample of a population for examination rather than examining 100% of a population. When Stringer arrived at the EO, there were few, if any, authoritative guidelines concerning the extent of testing and the selection process. Since his early years as an auditor, he had been dissatisfied with the lack of authoritative guidelines. He believed that in similar audit situations there existed an unjustifiably wide variation in the extent of testing prescribed by various auditors.

One frequently encountered method of testing during this era was block testing. Using this approach, a period of time is selected and audit procedures are applied extensively to that specific period rather than applied to transactions through the year. Stringer perceived the process of selecting the time periods to be somewhat arbitrary and felt that there must be a more objective method to “link what you were trying to accomplish with the amount of work that needed to be done.

An article by Howard F. Stettler that appeared in the Journal of Accountancy (January, 1954) catalyzed Stringer’s interest and convictions to improve the existing methods although his research did not begin until his arrival at EO. Stringer was, however, illprepared to embark on research in this area having never taken a statistics course. Recalling his initial impression, “I began the project with the very naive notion that if I would just buy one or two good books on statistics and study them, that I could move from there right into suggesting ways and means to implement these techniques into our practice. I had no idea that I was opening up a ‘Pandora’s Box’ nor did I realize the great amount of time it was going to take.

Having identified the research area, Stringer read a number of statistics textbooks focusing primarily on the sections dealing with statistical sampling. In the context of accounting and auditing, early writers devoted a great deal of attention to acceptance sampling and estimation sampling.

Acceptance sampling was designed as a quality control evaluation procedure for use in manufacturing processes. This technique received broad exposure when the nation’s industries increased production to supply the allies during World War II. An important characteristic of acceptance sampling is its inherently dichotomous nature, i.e., the decision alternatives are either absolute acceptance or absolute rejection. Some expressed hope that acceptance sampling could be adopted for use in accounting and auditing “since the regular flow of paper work can be regarded as reasonably comparable to the continuously flowing manufacturing production line.” [Trueblood and Cyert, 1957, p. v.]

When Stringer began evaluating existing statistical sampling techniques, statistical validity and audit relevance were the two primary criteria he considered. Stringer believed that the auditor is concerned with the relative degree of accuracy of accounting data. Consequently, he viewed acceptance sampling as illsuited in the usual audit situation since it provides an “absolute” choice of having to either accept or reject a population. Ideally, he envisioned a method of evaluation that included some form of graduated scale to measure the relative significance of the sample results. Stringer viewed such a method to be much more realistic and appropriate in the audit process. Because of the acceptreject dichotomy of acceptance sampling, coupled with the fact that this technique was severely limited in its ability to evaluate dollar items, Stringer believed that acceptance sampling lacked audit relevance. The following comments are excerpts from a paper presented by Stringer at the 1963 annual meeting of the American Statistical Association. These comments reveal the deficiencies of the statistical sampling techniques as perceived by Stringer when he arrived a EO in 1957:

. . . The principal sampling plans that have been proposed for use in auditing will be considered next in relation to this formulation of the auditor’s problem. Some of these are designed to be evaluated in terms of particular attributes and others in terms of monetary amounts.

Among the former are the plans for acceptance sampling, discovery sampling, and estimation sampling for attributes. Although these plans may be useful to the auditor for testing compliance with internal control procedures and for some other purposes, they share the common deficiency of not being related directly to the area of his principal concern — namely, monetary amounts. In addition, these plans are subject to the following criticisms from the auditor’s viewpoint: the automatic decision rules of acceptance sampling are too rigid and extreme for his purposes; the tables developed for discovery sampling provide for no evaluation of the quantitative significance of any errors disclosed by the sample .. .

CLASSICAL THEORY AND HIGHLY SKEWED POPULATIONS

Another sampling method that was frequently encountered in the literature at that time was estimation sampling. Estimation sampling, also referred to at that time as “survey sampling”, had been used primarily in research surveys in the natural and social sciences. A number of sampling proponents had been advocating the use of estimation sampling in auditing. Estimation sampling is used to estimate quantitative variables of a population, such as a population mean, proportion, or aggregate. For example, a sample mean can be calculated to estimate the mean age of the total population of accounts receivable. The most frequently used sample statistic of estimation sampling in the audit function is the population aggregate. A sample aggregate value can be calculated to estimate the aggregate value of a population such as the total dollar balance of an account, along with the range or “precision” of the probable actual dollar balance.

Stringer felt that auditors traditionally viewed the primary audit interest to be the number and magnitude of errors rather than the estimation of an account balance. Since the client’s records usually provide the exact total of account balances, Stringer believed that the use of estimation sampling would not contribute significant new insight into the majority of audit situations. Consequently, he viewed estimation sampling to be lacking in audit relevance since it failed to focus on the principal item of audit interest, i.e., errors.

Another problem of estimation sampling was that this method assumes that the distribution of the sample means of errors is normal. Stringer believed that in many populations of accounting data the related distributions of sample means of errors were not normally distributed. When estimation sampling was applied to these types of populations, any inferences concerning the population could be invalid because of violations of this underlying assumption. Stringer (1963) states:

The plans for evaluating samples in monetary terms have been applied in two different ways. However, for the reasons to be mentioned briefly, I believe that neither of these approaches has been fully responsive to the auditor’s problem.

The first approach has been to estimate the aggregate monetary amount of the items in the population, using a mean, ratio, or regression estimate, and to calculate reliability and precision in accordance with normal distribution theory. Aside from any question concerning use of the normal approximation for samples from highly skewed populations, this approach may be criticized because the principal component of the sampling variance is attributable to variation in the amounts of the individual sample items rather than to the amounts or errors, if any, in such items. Assuming a population with no error in it, each of the possible distinct samples of a given size that could be selected from it would result in a different estimate and precision limit under this approach; however, from the viewpoint of the auditor, all samples which include no errors should result in identical evaluations. In other words, the auditor ordinarily is not interested in estimating the aggregate amount of items in an accounting population, since the purported aggregate amount usually is already known or is readily determinable by simpler means, but he is concerned primarily with the possibility of errors in the population. 1 want to point out that this criticism is not directed toward applications where the basic purpose of the sample actually is to estimate the aggregate amount of an inventory or something else for which detailed records are not available.

From his experience in the field, Stringer realized that in most audit situations relatively few errors are observed. If errors are the focus of testing, then the population to be estimated is the error population. Stringer viewed estimation sampling to be an inappropriate technique to estimate the amount of errors in a population for two reasons. First, estimation sampling techniques require a relatively large number of sample observations in contrast to the few error observations typically encountered in the field. In addition, Stringer intuitively felt that the sampling distribution of errors in accounting populations was not normal. As Stringer observes “I began to sense that where you have very few observations of your feature of interest (errors), then there is a violation of the statistical assumptions that underlie normal distribution theory.”
Sampling to estimate the amount of error in a population when few errors are anticipated is a sampling problem referred to as “sampling for rare items.” Very little guidance existed in the literature regarding this issue. As Trueblood and Cyert note,

Sampling for the Rare Item — The auditor is frequently concerned with the location of fraudulent items in the universe. Unfortunately, there is relatively little that statistical sampling can do if the proportion of fraudulent items is small. Except in certain special cases, the sample size necessary to give a high probability of finding the rare item would be exceedingly large [1957, p. 37].

The lack of guidance for dealing with the rare item problem may be attributed to the fact that most of the early literature concerning estimation sampling focused upon techniques designed for use in taking public opinion polls or for use in scientific surveys for which observations were usually numerous. As Stringer (1963) observes:
. .. [The] use of the normal approximation as comprehended in estimation sampling is questionable where samples include no errors or only a few .. .

Under the second approach [to evaluating samples in monetary terms], the foregoing objection has been recognized by designing samples to estimate the amount of monetary errors and to provide related precision limits. However, because of the extremely low rate of occurrence of such errors in many if not most accounting populations, I believe the use of the normal approximation in calculating reliability and precision for samples from such populations may be very misleading. This problem is manifested in its most extreme form in cases, which are by no means uncommon in auditing, in which a large sample includes no errors; in such cases, of course, the sample estimate and standard deviation become zero. If the sample estimate, and related reliability and precision, are computed from a sample that includes only a few errors, the results may be only slightly less misleading — in fact, they may be more so because they appear to be more meaningful than they are.

Stringer viewed the violation of normal theory, due to the issues noted above, as a particularly ominous problem with implications beyond the important issue of unreliable inferences. In 1957, skepticism as well as outright opposition to the use of statistical sampling were prevalent throughout the accounting profession. If auditors applied estimation sampling using errors as the feature of interest, Stringer feared that in litigation, critics could call competent, expert statisticians that would readily testify that these sampling techniques were unreliable due to the violation of the basic assumption of normal distribution theory. He believed that this valid criticism would, in turn, reinforce opponents and stifle any efforts or progress towards the development of reliable statistical techniques. Convinced that the inherent objective of testing was the determination of the magnitude of errors in a population, Stringer began to have grave

244 The Accounting Historians Journal, June 1994
doubts concerning his pursuit of sampling innovations through the classical statistical literature.

SEEKING CONSULTATION

By December, 1957, it was clear to Stringer that the violation of normal distribution theory caused by the “rare item” issue constituted a problem that would make his “error focus” impossible or at least far more difficult to achieve than he had originally anticipated. Seeking someone in the firm with whom to share his observations and ideas, he was directed to the head of what was then referred to as the Operations Research Department of H&S which was located in Chicago. He traveled to Chicago, and met with the head of the Department in the week preceding Christmas, 1957. They began their discussions early in the morning. By lunchtime, it was clear to Stringer that the Operations Research Department was unable to fully answer his questions and comment upon his ideas.

Although Stringer’s original assignment had been “broadly defined”, there had been no specific reference or mention of his undertaking extensive research on statistical sampling. Consequently, he viewed himself to be operating on the “fringe” of the assignment he had been given. He decided to approach the two key figures in the chain of command, Oscar Gellein and Weldon Powell, concerning the possibility of hiring an outside consultant to assist in developing alternative sampling techniques. This request would clearly identify Stringer’s desire to pursue the issue of sampling and also give his superiors the opportunity to confront the issue. Their response would either signal support of his efforts, or, conversely, challenge the propriety of his proposed research. Whether their response was positive or negative, Stringer needed to know their reactions since he felt he had been operating on the fringe long enough. In addition, as a practical matter, if he were to continue his efforts, he needed the help of someone with expertise in the area of statistics.

The response of top management was most positive; Stringer was given permission to engage an outside consultant as well as the responsibility of choosing the consultant. In receiving permission, he felt he had gained the support of Oscar Gellein and Weldon Powell as they openly concurred with his belief that research related to the use of statistical sampling was indeed a worthy undertaking. In the Spring of 1958, he began his search for a competent statistician with whom to collaborate.

First to be considered were the authors of the various texts that he had studied in the initial stages of his research. These included Deming, Cochran, and Hansen, Hurwitz and Madow. It was important that the consultant be located reasonably close to New York for the sake of accessibility.

Stringer soon acquired the proceedings of a seminar that had been conducted by The Society of Business Advisory Professions which was a group consisting of lawyers, accountants, statisticians and other professionals that serve in a consulting capacity. He acquired the proceedings from Weldon Powell who was active in the Society. The focus of the seminar was the use of statistical sampling in court cases, market research and, to a lesser extent, the possible use of statistical sampling in auditing.

The proceedings included a transcript of the discussions as well as the papers that were presented. As Stringer recalls:

I was very much impressed by what I thought was the common sense approach of one person who was a statistician. He seemed to realize that successful applications had to be a combined effort of the statistician and a person who is familiar with the subject matter. This person appeared to have an open and cooperative attitude.

This “cooperative attitude” was particularly important since H&S had once before consulted a statistician. This previous effort was terminated due principally to the condescending attitude of the statistician who alienated the senior management of H&S by conveying the notion that accountants were incapable of achieving an understanding of statistical theory and competence in the use of statistical methods. One of Stringer’s colleagues commented that this unpleasant experience had “set us back five years”; consequently, he was especially interested in finding a statistician who possessed a positive, constructive attitude and a cordial demeanor.
The statistician whose seminar comments had so impressed Stringer was Frederick F. Stephan, a professor of statistics at Princeton University and former President of the American Statistical Association. Stringer recalls:

I read a few of his articles and found he was also the author of a book which had not come to my attention up to that point. I read the book and made additional inquiries concerning his standing within the profession. Based on the attitude he had displayed as reflected in

246 The Accounting Historians Journal, June 1994
the proceedings, my review of his publications and his stature within the profession, I decided Fred was the first person I wanted to talk to. I called him and arranged a meeting at which Oscar Gellein was also present. At the end of the meeting, Oscar and I were both satisfied that Fred was the type of person we could work with. That was the beginning of a consulting relationship that was very constructive, pleasant and productive.

AN INTERDISCIPLINARY COLLABORATIVE EFFORT

The proximity of Princeton to New York enabled Stringer and Stephan to meet frequently; they wasted little time in pursuing the project. By the end of Spring, 1958, Stephan had reviewed the research that Stringer had done with regard to statistical sampling and they discussed the issues and problems that Stringer had identified. After these initial discussions, they decided to lay aside the statistical issues and concentrate their efforts on helping Stephan achieve a greater understanding of the nature and objectives of the audit process. This mutuallyagreed upon approach reinforced Stringer’s earlier conviction that the applications of statistical theory to auditing had to be the product of a synthesis of efforts emanating from both disciplines — auditing and statistics. The successful scenario he envisioned was an auditor with an understanding of statistics, collaborating with a statistician aware of the objectives of an audit and the audit techniques employed to meet those objectives.

Stringer views his efforts to develop and gain acceptance of a statistical sampling plan as a three step process: “exploration,” “development” and “selling.” In the exploratory phase, Stringer familiarized Stephan with the audit process. First, they undertook field trips to practice offices to examine workpapers and to converse with practitioners. Stringer explained the audit process using the workpapers of audits which had been performed without the use of statistical sampling. Stringer describes their activity at this stage as a “twoway process.” While they reviewed the workpapers, Stringer would suggest or point out auditing procedures that he hoped could benefit from the use of statistical sampling. At the same time, Stephan would be asking questions concerning the purpose of the procedures, what types of populations the data were drawn from, etc. They then visited a number of clients’ offices and were permitted to examine actual accounting records to increase Stephan’s understanding of the conditions experienced by the auditor in the field.

By midsummer of 1958, after considerable exposure to the audit process, Stephan agreed with Stringer’s conclusion concerning the significant limitations of existing statistical sampling methods. Stringer recounts, “… he agreed completely with my conclusion that something better than what was presently available was needed.” This realization marked the end of the exploratory phase and the beginning of their search for a theoretically sound method of statistical sampling which focused upon the amount of error in a population.
Our goal was to develop an approach that would not require us to rely on normal distribution theory, but at the same time, would allow us to express conclusions about errors in a population in the form of estimates and upper precision limits even though few or no errors were detected in the sample.

In addition, to facilitate the acceptance of his approach by the firm and the profession, comprehensibility and ease of application by those in the field were key attributes to consider; this was the beginning of the development phase.

THE DAWN OF BEHAVIORAL. RESEARCH IN AUDITING

As Stringer progressed through the development stage, the focus of his efforts turned increasingly towards the next phase which was achieving acceptance of the plan and its eventual implementation. Since no other national firm had yet developed and formally adopted the use of statistical sampling, these tasks proved to be equally challenging. In terms of time and effort, the “selling stage” equaled both the exploratory and the development stages combined.

In formulating his strategy, he knew the obvious strong point was that the statistical sampling plan would provide a more objective measure of the extent of testing and the evaluation of results. Although innovations in audit technology are often welcomed by the profession, the rampant misconceptions and pervasive lack of understanding of statistical sampling that existed at that time constituted a problem which Stringer viewed to be a serious threat to its eventual acceptance by the profession.

Intuitively, he felt that there would be wide variability in the amount of testing that various auditors would decide to perform if given an exact set of circumstances. He knew that if he could somehow measure and document the magnitude of variance in auditors’ judgment concerning the extent of testing, such documentation might provide persuasive evidence of the seriousness of the problem and contribute to establishing the need for the Plan. In an attempt to gain insight into the extent of variability in auditors’ judgment, he designed and performed a laboratory experiment. It is likely that this experiment constitutes the earliest documented effort in what is now referred to as behavioral research in auditing.
An annual training seminar for senior accountants was held every August by H&S, in Skytop, Pennsylvania. Prior to the meeting held in August, 1959, Stringer prepared a questionnaire consisting of four descriptions of audit situations that are encountered often in the field. Each senior was then asked to decide the extent of audit procedures he would perform in each of the four cases. The questionnaire was distributed to 92 senior accountants at the beginning of the meeting. (The questionnaire and results are reproduced in the Appendix.)
The results suggested a wide variation in the extent of testing that various auditors would perform given identical circumstances. Stringer did not attempt to release the results of the survey to the firm as a whole or make them available to anyone outside the firm since he viewed the results as “not something you would be particularly proud to have circulating in public.” He did present the results to various partners and managers who were aware of his work in the area of statistical sampling.

Regarding the reaction of partners who reviewed the results of the survey Stringer observes:

They were shocked and dismayed at the disparity that the survey showed. I can not say that the survey results were the deciding factor in the firm’s eventual adoption of the Plan, but I think it is fair to say that the results had a significant influence on the firm’s views concerning the existing disparity in the extent of testing and the need to improve the situation. However, there are two important points I always address in any public discussion of the survey results. First, given the lack of professional guidelines in this area, the results were not surprising. Second, the auditors surveyed were all employed and trained by the same firm. If the survey had been distributed to a group of auditors who had been selected randomly from throughout the profession, it is reasonable to assume that the disparity would have been even greater.

Regarding the impact of Stringer’s research, an H&S Committee which was later formed to evaluate the Plan, issued a progress report in 1960 which stated:
. . . The results of the survey at the 1959 meeting of incharge accountants concerning sample sizes in selected situations are disturbing and indicate that our accountants need better guides of selecting appropriate sample sizes [Stringer, 1993].

Collegial Support

While developing the Plan, Stringer continually kept his supervisors aware of his activities so that he might receive feedback on various issues as well as sustain their support. During this period, Stringer reported directly to Gellein with whom he developed a close personal and professional relationship. He kept Gellein particularly well informed of all his activities and progress. During his entire tenure in EO, Gellein was the source of much encouragement and support. Stringer viewed him as his “mentor, intermediary and confidant.”
Although neither Queenan nor Powell took a specific, personal interest in the project, they willingly permitted Stringer to devote the majority of his time to the project. Regarding the Plan, Stringer characterized them as “open minded”, “tolerant” and willing “to give me room to run.” Stringer recalls “I was receiving encouragement from Oscar, and was more or less, just being left alone by the other people.” In retrospect, their receptive attitude is especially significant given the magnitude of opposition that existed in the profession at that time concerning the use of statistical sampling.

The period during which he developed the Plan proved often to be frustrating and at times discouraging, but always was intellectually challenging. A constant burden was the requirement that the Plan not only had to be statistically sound, but practical in the sense that you could train people within a reasonable amount of time, and its application could not be economically prohibitive from the standpoint of the time and effort actually needed to apply it out in the field.

However, he never felt that the task was insurmountable; he always believed that he and Stephan would eventually develop a sound, workable Plan.

The next phase of the research will examine the remaining issues which Stringer needed to address and resolve before the Plan was officially adopted and fully implemented. These include the problem of relating the degree of reliance upon internal control to the extent of substantive testing, the field testing of the Plan, and the mathematical and legal reviews of the Plan.

CONCLUSION

First, this research suggests that Stringer’s questioning of sampling techniques then prevalent in auditing practice and his doubts concerning the statistical validity of existing statistical sampling techniques provided the impetus for the development of PPS. The study also reveals that the Stringer’s Plan, unlike existing plans, was tailored specifically to the needs of the auditor.

Second, the research reveals thait Stringer, in what was probably the first use of behavioral research in auditing, successfully used a laboratory experiment to document the wide variation in auditing testing. The results of his research convinced the firm’s management that a more objective method of testing needed to be developed. Stringer’s success in documenting the problem and thus reinforcing the firm’s commitment to his research efforts clearly illustrates the potential value of behavioral research.

Third, the study documents the manner in which Stringer engineered an interdisciplinary collaborative effort with Professor Stephan. This collaboration illustrates the potential of interdisciplinary efforts which attempt to adapt and apply scientific advances to the practice of accounting and auditing.

Fourth, the study illustrates the importance of creating within the firm an environment which is financially and philosophically supportive of research and experimentation. H&S’s progressive policy regarding technological innovation provided Stringer with the support which he needed to accomplish his achievements.

Fifth, the research provides a concrete example of the importance of mentorprotege relationships. Stringer, who viewed Gellein as his mentor during this period, believes that Gellein’s encouragement and support were important factors in his success during this period. Such encouragement and support are especially important when the protege is faced with the uncertainties and doubts inherent in any research undertaking.

Lastly, this study documents the pioneering efforts of Stringer as a means of providing younger members of the profession with a role model as well as instilling a feeling of professional pride and continuity.

REFERENCES

AICPA, Committee on Statistical Sampling, [prepared by Vance, L. L., R. M. Trueblood, and R. J. Monteverde], Glossary of Statistical Terms for Accountants and Bibliography on the Application of Statistical Methods to Accounting, Auditing and Management Control, (1958) [Available at the AICPA, New York]
Alter, J., “Retaining Women CPAs,” Journal of Accountancy (May 1991): 5055.
Dennis, A., “A Way to Build Future Leaders,” Journal of Accountancy (May 1993): 7075.
Grimlund, R. A., “Sample Size Planning for the Moment Method of MUS: Incorporating Audit Judgements,” Auditing: A Journal of Practice and Theory ( Spring 1988a): 77104.
Grimlund, R. A., and Schroeder, M. S., “On the Current Use of the Stringer Method of MUS: Some New Directions,” Auditing: A Journal of Practice and Theory (Fall 1988b): 5362.
Hall, T. W., Pierce, B. J., and Ross, W. R., “Planning Sample Sizes for StringerMethod Monetary Unit and SingleStage Attribute Sampling Plans,” Auditing: A Journal of Practice and Theory (Spring 1989): 6489.
Hansen, S. C, “Strategic Sampling, Physical Units Sampling, and Dollar Units Sampling,” The Accounting Review (April 1993): 323345.
Ko, C, Nachtsheim, C. J., Duke, G. L., and Bailey, A. D., “On the Robustness of ModelBased Sampling in Auditing,” Auditing: A Journal of Practice and Theory (Spring 1988): 119136.
Pillsbury, C, Capozzoli, L., and Ciampa, A., “A Synthesis of Research Studies Regarding the Upward Mobility of Women in Public Accounting,” Accounting Horizons (March 1989): 6370.
Plante, R., Neter, J., and Leitch, R. A., “Comparative Performance of Multinomial, Cell, and Stringer Bounds,” Auditing: A Journal of Practice & Theory (Fall 1985): 4056.
Stettler, H. F., “Statistical Interpretation of Test Checks,” Journal of Accountancy (January 1954): 4957.
Stringer, K. S., Document source: Stringer’s personal files (1993).
Stringer, K. S., “Practical Aspects of Statistical Sampling in Auditing,” (1963). Presented at the 1963 Annual Meeting of the American Statistical Association.
Stringer, K. S., “A Statistical Technique for Analytical Review,” Studies on Statistical Methodology in Auditing, supplement to Journal of Accounting Research, (1975): 119.
Stringer, K. S., and Stewart, T. R., Statistical Techniques for Analytical Review in Auditing, 1st Edition, New York: John Wiley & Sons, Inc. (1986).
Trueblood, R. M., and Cyert, R. M., Sampling Techniques in Accounting, PrenticeHall, Inc. (1957).
Vance, L. L., and Neter, J., Statistical Sampling for Auditors and Accountants, New York: John Wiley and Sons, Inc. (1956).
Viator, R. E., and Scandura, T. A., “A Study of MentorProtege Relationships in Large Public Accounting Firms,” Accounting Horizons (September 1991): 2030.

APPENDIX
The following is a reproduction of Stringer’s questionnaire followed by the results which he compiled:

1959 MEETING OF INCHARGE ACCOUNTANTS
QUESTIONNAIRE CONCERNING EXTENT OF AUDIT TESTS

This questionnaire is designed to contribute to our study of the problem concerning the extent of audit tests by obtaining a crosssection of the views and practices of our incharge accountants. It need not be signed. Please answer all questions specifically and as promptly as possible.

The questions pertain to an examination of the following condensed hypothetical financial statements:
Balance Sheet
Current Assets: Current Liabilities:
Cash $ 1,900,000 Notes payable $ 1,000,000
Receivables 3,000,000 Accounts payable 2,000,000
Inventories 5,000,000 Accrued liabilities 1,000,000
Prepaid Expenses 100,000 Total 4,000,000
Total 10,000,000 Capital Stock 10,000,000
Propertynet 15,000,000 Retained Earning 11,000,000
Total $25,000,000 Total $25,000,000
Statement of Income
Net Sales $30,000,000
Costs and Expenses:
Cost of sales $20,000,000
Selling and general expenses 4,000,000
Income taxes 3,000,000
Total 27,000,000
Net Income $3,000,000

It is to be assumed that all items investigated in connection with the analytic review have been cleared satisfactorily.

1. The receivables include 2,000 accounts whose balances range from small amounts to a maximum of $5,000, with no particular concentration of balances at any level. There are no accounts that appear to be especially old or otherwise unusual, and the internal control is good with respect to cash receipts, sales and receivables. State the total number of accounts you would select for confirmation and indicate briefly how you would select them.

2. What changes, if any, would you make in your confirmation program if the receivables consisted of 200 accounts with none exceeding $50,000 and with other conditions remaining the same as in Question 1?

3. What changes, if any, would you make in your answers to Question 1 (2,000 accounts, none exceeding $5,000) if internal control over cash receipts, sales, and receivables were bad but previous audits had not revealed material errors in the receivables?

4. The physical inventory includes 5,000 line items, the extensions of which range from small amounts to a maximum of $5,000 with no particular concentration of amounts at any level. Perpetual inventory records show quantities and unit prices but they are not under general ledger control. The gross profit rate did not vary significantly from that of the preceding year. Inventory prices have not been rechecked by the client. How many items would you select for your test of inventory prices?

5. If your test of inventory prices revealed 5 errors resulting in a net overstatement of $2,500 (overstatements of $3,810, $1,010, $410, and $310, and an understatement of $3,040 in the extended amounts of the 5 line items) would you:

a. Consider the test satisfactory?
b. Extend the test? If so, how much, assuming no additional errors were found?
c. Take some other action? If so, describe briefly?

6. Your answer to Question 5 was undoubtedly based in part on your judgment as to the aggregate dollar amount of errors in the total inventory that would be material in relation to the above financial statements. What is the approximate maximum dollar amount of such errors that you would consider immaterial, assuming that there was no indication of fraud and that the amount given in your answer is before the income tax effect?

7. Your inquiries concerning the preparation and approval of vouchers indicates that internal control is good. The client issues approximately 1,000 vouchers each month. How many vouchers would you examine to afford satisfactory evidence of internal control in this respect?

8. If your test disclosed 3 instances of failure to comply with procedures which were necessary for control, but the vouchers appeared to be valid in all other respects, would you:

a. Consider the test satisfactory and proceed on the basis of good internal control?
b. Consider it unsatisfactory and proceed on the basis of bad control?
c. Extend the test? If so, how many more vouchers would you consider it necessary to examine to satisfy yourself concerning control if no additional procedural or other errors were found?
d. Take some other action? If so, describe briefly.