International Journal of Accounting Information Systems 1 (2000) 88–90
Discussion of perceived semantic expressiveness of accounting systems and task accuracy effects Scott L. Summers* School of Accountancy & IS, Brigham Young University, Provo, UT 84602-3098, USA
1. Introduction From the inception of accounting as a business discipline, the Debit-Credit-Account (DCA) accounting model was the fundamental technology of the discipline. It was designed to “. . . furnish an orderly, continuous record of the assets and equities of the business enterprise.” (Paton, 1937, p. 56) This model of accounting has changed little over its life. Paton (1937, p. 56) wrote, “Further, it is in general safe to say that there is no possible scheme of abbreviated bookkeeping which is as satisfactory as the doubleentry system from either the clerical or the informational standpoint.” While Paton’s thoughts may still be true for abbreviated bookkeeping systems, abbreviated systems are no longer satisfactory. Times have changed. Changes in information technology have created the opportunity for new accounting models to come forth. The selling points of these new models include the ability to capture more data about business activities, store this data in a disaggregated form, and summarize the data according to the particular demands of a user at any time. One such accounting model is the Resource-Event-Agent (REA) accounting model (McCarthy, 1982). This model is able to meet the demands of today’s information hungry users as well as produce the traditional outputs of the DCA model. McCarthy (1982, p. 555) states that an important feature of REA is its emphasis on the semantic expressiveness of the data model. In other words, the elements in the data model correspond to the elements in the modeled corporate reality. Fundamental differences exist between the REA and DCA data models. First, the DCA model was designed as an abbreviated, efficient accounting system focused on producing summarized financial informa-
tion, while the REA model was designed as an expanded, data-rich, flexible accounting system focused on producing information at any level of summarization. Second, the DCA model was not designed to be semantically expressive of the underlying reality, while the REA model was. Given these differences, one might question the contribution of research that compares the semantic expressiveness of the two models. One should expect a model designed with semantic expressiveness as a “feature” to be more semantically expressive than a model which does not focus on semantic expressiveness; nevertheless, good science is not limited to monumental discoveries, but is forwarded through exacting study of elements within an existing paradigm. This research (Dunn & Grabski, 2000) contributes in the latter way; moreover, this research adds to the body of research on semantic expressiveness of accounting models by testing whether semantic expressiveness affects task accuracy. 2. The experiment 2.1. Experimenter bias The experiment was conducted using 73 students who were enrolled in an undergraduate accounting information systems (AIS) class. One of the topics being introduced to students in this class was the REA accounting model. These students had completed an average of six undergraduate accounting courses which were most likely taught with reference to the DCA accounting model; thus it is likely that the students had significantly more experience with the DCA model than the REA model. This additional training may bias the students toward favoring the DCA model as well as performing DCA-oriented
* Corresponding Author. Tel.: 1-801-378-9790. E-mail:
[email protected] 1467-0895/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved. PII: S1 467-0895(00)00 0 0 5 - 1
S.L. Summers / International Journal of Accounting Information Systems 1 (2000) 88–90
tasks more accurately with the DCA model than with the REA model. Offsetting this bias toward favoring the DCA model is the bias induced by the presence of, and subsequent evaluation by, the person who taught the REA model. Pleasing the experimenter is a potential bias. The first bias regarding the numerous classes of training in the DCA model is not of significant importance in this article since it works against the hypotheses; however the second bias with regard to the experimenter’s presence and subsequent evaluation may confound the results with regard to Hypothesis One. 2.2. Problem frame The students were asked to complete four tasks: two of which were easier to complete with the REA model, and two of which were easier to complete with the DCA model. While it is clear that either model was capable of generating the answer to the DCA-facilitated tasks, it is not clear that the DCA model is capable of generating the answers to the REA facilitated tasks. The resolution of this issue hinges on whether the reader accepts the source documents as part of both models. For example, one task required the student to determine the company’s gross sales by inventory stock number. Some quick manipulations of the tables within the REA model provide the appropriate answer; however, students assigned to complete this task with the DCA model are not able to use any of the model features of the DCA model since the DCA model is not capable of answering this question (one cannot answer the question with Debit, Credit, or Account information). Rather, the student must use the raw material provided with each model: the source documents. While the authors argue that the source documents are part of each model, I am concerned that including the source documents as part of both models calls into question the validity of the comparison between the DCA model and the REA model. In my view, comparisons between two models need to be based upon elements in which the models are distinct. This particular comparison forces one model (the DCA model) to be evaluated on an element that is shared in common by both models. In other words, the students who answered this question with the REA model had the choice of basing their solution on elements unique to the REA model (tables and relationships) or on elements shared by the DCA model (source documents). Students answering this question with the DCA model could not answer the question based on elements unique to the DCA model. In my view, the source documents (whether paper or electronic) are the raw materials upon which any
89
accounting model is built. It is the capture of data from these source documents and the subsequent manipulation, storage, and perhaps presentation of this data/information that constitutes the accounting model. Without splitting hairs, I freely confess that any accounting system includes source documents, whether paper or electronic; however, a comparison of accounting models should not include source documents. The authors use these terms somewhat interchangeably. The end result of this difficulty is the potential for students to become disenchanted with the DCA model because it cannot answer the questions without reverting to the source documents. Moreover, required use of the source documents for the DCA model may affect a student’s task accuracy. Most accounting courses do not teach students how to interact with source documents. In my interactions with students, they are usually more familiar with a t–account or journal entry than a purchase order. 2.3. Terminology The authors use a covariate that they call “cognitive fit.” This variable is named based on the work of Iris Vessey (1991). While Vessey has produced a significant amount of work that identifies “cognitive fit” as congruence between problem-solving task and problem presentation, I believe that the term connotes an examination of cognitive style or ability. In that the subject’s cognitive workings are not investigated, I think that the literature would best be served by using a term that better represents the elements being studied. Perhaps the term “problem representation-task fit,” described as the congruence of the task with the information model architecture, would better fit the essence of the measured effect. This naming convention follows the same lines that other authors have followed. For example, Goodhue and Thompson (1995)use the term “technology-task fit” in a similar application. 3. Summary The authors are to be commended for their efforts in this area of research. Despite the criticisms, which I provided above, I found the paper to be both intuitive and insightful. The authors demonstrate an impressive command of the literature on semantic expressiveness and its abstraction of business reality into accounting models. I do not believe that the criticisms that I propose are fatal. I accept the results and implications as reported for Hypothesis One. I find these results appealing. I accept the results and implications of Hypothesis Two with some caution. I hope that the results with regard to task accuracy can be replicated in an extension to this work. I encourage the authors and the readership
90
S.L. Summers / International Journal of Accounting Information Systems 1 (2000) 88–90
in general to continue to explore the effects of semantic expressiveness on task accuracy as well as other measures of performance. Of particular appeal would be a study comparing the semantic expressiveness of the REA model as it overlays entity-relationship technology with the semantic expressiveness of the REA model as it overlays object-oriented technology. References Dunn CL, Grabski SV. Perceived semantic expressiveness
of accounting systems and task accuracy effects. Int J Account Inform Syst 2000;1:79–87. Goodhue D, Thomson RL. Task-technology fit and individual performance. MISQ 1995;19(2):213–36. McCarthy WE. The REA accounting model: A generalized framework for accounting systems in a shared data environment. Account Rev 1982;57:554–78. Paton WA. Accounting. New York: The Macmillan Company, 1937. Vessey, I. Cognitive fit: A theory-based analysis of the graph versus table literature. Decision Sciences 1991;22:219–40.