Automation of Double-Entry Accounting: An Emerging Technology
A Research Report Wesley C. Sampson Ph.D., CPA, CDP Retired, University of Toledo
Automation of Double-Entry Accounting: An Emerging Technology
A Research Report
This new tool provides analytical information over audit, internal control, and accounting operations. Translation of account numbers in journal entries to matrix notation enables index classification of double-entry journals. The structure involves posting whole journal entries rather than ledgers and may be executed parallel with present programs. Real-time implementation is possible.
I. Double-entry vs. Single-entry Practice
From adding machine to punched card to relational database to XBRL, the coordinated effort to report accounting results on the web, single-entry systems have performed an increasing portion of Accounting tasks. During the earlier years of mechanization and automation, an emotional difference existed between the proponents of single-entry vs. double-entry. The clear advantage of automation has silenced the discussion, single-entry is now dominant. This research, however, points toward a revival of double-entry. The organization of double-entry accounting around journal entries appears to be a new concept.
Double-entry involves relationships among the individual accounts included in the entry. On the other hand, the individual ledger posting illustrates single- entry accounting. Often, the individual row of a relational database illustrates a single-entry posting. This new automation identifies all journal entries by the particular combination of accounts addressed, providing a complete classification of journal entries for analysis and reports.
The clerical task of double-entry caused many to seek single-entry solutions. Punched-card and one-write systems of the early and middle twentieth century sought to reduce the copying involved in the posting function. Reputedly, the dominant bookkeeping text of 1923, Twentieth Century
Accounting, taught making entries on the ledger first – only to find that students balked at copying the data into journal form (thereby hindering audit). A related problem exists on today‟s computer systems. Reputedly, some real-time systems suspend the log journal function at busy hours. The human tendency may be to avoid the transaction journal, the fundamental record of Accounting, at critical times. Electronic journal posting rather than ledger posting may make double-entry Accounting more acceptable.
The procedure discussed here is based upon balanced double-entries as input, although a programming exception has been granted for the many entities which use both double and single entry in their books.
II. . Post the Journal Entries
Posting whole journal entries rather than individual ledgers offers improved supervision, improved reporting of Accounting results, and faster execution for real-time accounting.
A new, hopefully more useful, definition of transaction flow results from posting journal entries. Each group of identical (except for monetary value) entries is totaled. For example, monthly adjustments should show as an annual total at year-end. Precise identification of transaction flows has been missing from previous accounting analysis. Many audit programs include the step of „read the journal to familiarize oneself with operations, material transactions and unusual transactions‟ This important but undocumented step has become increasingly difficult to perform as the volume of automated bookkeeping expands. AICPA Practice Alert 2003-2 and SAS-99 have addressed this need but various software firms have presented only partial solutions .
Incomplete knowledge of transactions also limits the evaluation of internal control. Supervisors and auditors rely upon rules and procedures, knowing that a portion of the transactions is not fully inspected. Analytical review of transaction flows with known monetary and transaction activity
totals enables personnel to evaluate internal control with a further degree of accuracy.
Indexes which manage data collection and reporting of complete journal-entry information appear to fit the description of simple tools to enable real-time accounting, as suggested by Vasarhelyi (1998, p.218). The combinatorial index presented here is dynamic, probably of sufficient speed for real-time accounting. Nevertheless, it remains a human tool. The structure was developed to enable knowledgeable personnel to compete with Expert System Structure. It was designed to improve supervision and audit.
Real-time activity involves persistent magnetic records, usually on magnetic disk. A whole journal entry requires one write operation whereas the component ledger entries require several write operations. Resultant saving of write operations is expected to improve computer execution of real-time accounting applications. Remaining indexes of the journal database (to produce trial balance and specific searches) are constructed at retrieval time. The relative efficiency and greater information available via this structure is independent of hardware improvements.
This computer-based combinatorial index originated from Accounting thought and is designed to organize „events‟ as proposed by Sorter (1969). This author describes the journal as combinatorial while the ledger is hierarchical. The ledger is easily organized line by line as in punched card systems or relational database organization. Journals entries, which occupy a variable number of lines, require a logic which has not been included in relational database. Although some audit packages have identified whole entries in limited situations, organization of the entire journal file by the (usually) primary keys related to the accounts posted is quite difficult. The return to double-entry logic adds operational information to existing financial/managerial usage of general ledger systems. Trial balance information is unchanged from traditional structure. Using the same journal input and producing identical trial balance output, the index method may either supplement or replace traditional ledger posting.
III. Intellectual History De Morgan (1846) is often credited with origination of the matrix approach to Accounting. Leech
(1986) organized the reports of writers from De Morgan onward. Seemingly all the matrices discussed were square, emphasizing either the debit-credit equality or effects upon financial statements. A more recent author, Koshimura (1988), extended matrix methods to preparation of the Statement of Funds. Many statement values were determined by a periodic method.
Mattesich (1964) took a different, potentially perpetual, approach. His matrix was intended to produce a flow statement, in his case the budget. The occupied cells of his matrix contained only simple journal entries supporting the budget plan. Sampson and Olan (1992) recast the square, two-dimension, matrix to a sparse matrix of multidimension form to provide for analysis of existing journals, which include compound entries. Sampson and Douthat (2000) developed index control over the journal entry summaries represented by occupied cells, making real-time processing practical.
The late Gerald Salton and others in Information Science(1975) developed an outwardly-similar-to-this-method vector-space concept of indexing in the early 1970s. Salton‟s text-processing approach has spawned a number of text-search programs which attempt to organize large areas of information. An important part of the vector-space concept was the abandonment of an expensive library concept, controlled vocabulary.
Business, however, uses a controlled vocabulary in many situations. The Chart of Accounts is the controlled vocabulary for Accounting. Sampson and Olan modified the Salton classification logic to include only one domain with its controlled vocabulary, in our case, the journal of the entity Accounting system.
IV The Journal in Accounting Theory
A cursory reading of Accounting history indicates that early journals were oriented toward individual trading partners. This writer speculates that a need similar to present-day customer relationship management was a driving force behind adoption of double-entry methods. The personal nature of transactions was preserved in narrative in early Accounting; but, by the late
1800s accounting practice had changed the position-sensitive usage of debit and credit. Chatfield (p 67) states that “By 1900 these last vestiges of the complete thought entry had disappeared from most ledgers, whose purpose had become simply the tabulation of account balances for easy statement preparation.” It was from this base „for easy statement preparation‟ that the first Accounting automation began. This writer speculates that early accountants attempted to define „events‟ but the difficulty of defining and copying complete (perhaps multidimensional) events on the paper medium contributed to emphasis of the ledger over the journal.
Paton and Littleton (1940) defined Accounting as transaction-driven. Theorists have long been aware that both the journal and the ledger have been vital for a complete double-entry system. Sampson‟s index, patented in 1995, requires only the Chart of Accounts with opening balances (a master list) and the complete journal. All ledger information is produced via indexes to read data from posted journal entry summaries -- or the actual journal entries when additional detail is needed. The definition of Accounting is satisfied using the index procedure. The same information of conventional systems is present in both the journal and the ledger; however, identical combinations of accounts addressed in repetitive entries are totaled into one summary. These summaries portray the operations of the accounting system.
Much computer technology of the method proposed has been available since the late 1960s. However, it was not until the popularity of PASCAL and its related languages that production-type programming was available for variable-length records. Patents have been issued for the matrix notation method of organizing data into variable-length records, its extension to the Accounting cycle, and to index management of the variable-length records. In the current implementation, comparatively recent public domain functions manage matrix notation of large master lists and the specific identification of large numbers of journal entry summaries.
V. What is the New Technology?
A specific journal entry summary of three accounts (modified by debit or credit) will occupy a cell within a three-dimension matrix and can be addressed in matrix notation as [entry] (i,j,k). More complex entries will use a longer notation. In this manner every summary can be identified quickly. The storage of only occupied cell addresses releases computer resources which would otherwise be directed to matrix management. Double-entry accounting fits sparse matrix logic.
This index organizes the journal by entry summaries, one record for each summary of identical entries (groups of account-sides). Total dollars are accumulated for each account-side addressed in each entry summary. The summaries are managed by a number of simple insertion indexes. There is an index of summaries related to any individual account-side. The number of accounts within each journal entry summary is also indexed. The count of transactions totaled in each specific summary record is also associated with an index. An index of storage locations for each entry leads the reader to the specific entries when needed.
The new technology places a series of journal entry totals between the original journal and the ledger. It also organizes ledger detail in the sequence of the summaries, rather than present date sequence.
X. Outline of New Information Available An output spreadsheet reports not only the usual debits, credits, and balances for each account, but also the tally of original transactions which post each the debit or credit sides and the number of summaries used. When reporting an account-side the full summaries are displayed in journal form. The 'other ends‟ of each pattern are visible. The count of repetitive entries is also displayed. The totals of repetitive entries are, of course, useful in judging materiality. Searches conditioned on the dollar amount summarized for a given account also aid the auditor.
Logical anomalies are more easily identified. Once an anomaly is sensed, the supervisor/auditor
may use Boolean searches to find specific combinations of accounts in any journal entry. Individual entries are identified within a summary, enabling a full detail search when needed.
VII. Advantages of Emphasis on Journal Entries “…(B)etter measurement insight can be derived by the individuals who have extensive hands-on experience with the process and are essentially the experts at the process.” (Sutton, et al, 1998)
(1)Reports Directly from the Journal
Organization of Accounting records around journal entries offers easy-to-prepare flow reports. The direct method of Cash Flow reporting, for instance, can be prepared „on the fly‟ via this index. Specialized cost reports appear to be a promising field of search. The trial balance is retained, allowing periodic reports to be continued in batch mode.
Computer access to a sorted and tabulated Accounting journal gives the auditor a new tool with which to work. Most audit plans contain a step of „read the journal and observe unusual transactions‟. The purpose of this step is economic measurement of the firm via identification of outlier transactions and major fund flows, as well as evaluation of workers in the accounting process. The index of journal entry summaries helps to find specific areas for inspection. The judgments made via the manual step of „read the journal‟ can be categorized and managed and documented through search via an index. This procedure provides analytic review of internal control. The ability to find unusual entries, the tally of entries related to specified ledger accounts, the separation of the preponderance of transactions from the outliers, and the identification of supersets and near-neighbors of specific transactions all add to the supervisor‟s ability. Real-time monitor usage of the index can provide managers with transaction reporting and warning as well as balance reporting.
(3) Related (Accounting-series) Applications
Another advantage may be the understanding of data combinations in computer-assisted management of other business disciplines (i.e., product mix and other mixes).
All accounting-series computer applications can be analyzed via a combinatorial index, even though single-entry is their logical organization.. In general, the index appears useful wherever transactions typically contain several items. Invoices, schedules, and professional cases are expected to contain patterns of activity: this aspect of their management fits combinatorial reasoning. The number of attributes of a loan application may fit combinatorial reasoning, but the often-single postings to demand deposit accounts do not. Master lists are found in much of business and other specialized endeavors. For accountants‟ purposes, the master list over journal entries is the Chart of Accounts, doubled to include all possible posting locations available to a single journal entry, and the opening balances of each account. Master lists of inventory, operational tasks, legal classification, and medical procedures are examples of posting situations where a combinatorial index is expected to add information value.
VIII. Alternative to Artificial Intelligence
The general combinatorial model underlying double-entry Accounting automation provides an alternate logic to multi-dimension designs. Several designs of artificial intelligence contain the concepts of multi-dimension matricies and near-neighbor. When used in prediction, artificial intelligence algorithms often calculate an uncertain result with concurrent estimates of correlation and/or confidence. The combinatorial index retrieves only historical transactions for its user, who must then make the decision under human uncertainty. The purpose of the index is to improve the user‟s data gathering. IX. Measurements to Date
Previous articles (Sampson and Olan, 1992, and Sampson, 1996) have described the concept of a journal-entry identifier record and its potential in audit. Unfortunately, the early program implementations were not scaleable. A scaleable prototype designed for real-time execution was written in 1997 - 1998. The prototype has been converted to Linux-based operation and a graphics user interface (GUI) is in process.
Much of the data targeted by a combinatorial index is confidential, with resulting external restrictions. Journals of large entities containing mixed double-entry and single-entry transactions were tested 1993-1995. No errors were found in classification of entries from single-line size to a 1503-line double-entry. A 142,000-case hospital record database was successfully tested in 1998 and a 179,000-line double-entry journal was successfuly organized and audited in 2003. The auditor changed the opinion regarding Internal Control as a result of this test.
Tests of cash-flow preparation have not been performed. The index supports extraction and recapitulation of all entries posting debits to all cash accounts, with the other ends providing the basis for reports of sources of cash; and similar extraction and recapitulation to support cash usage. No field test has been performed on this topic. XI. Limitations The public-domain functions which currently manage matrix notation and searches are untested for their upper limits. We believe that charts of one hundred thousand accounts can be managed and that ten million unique journal entry summaries can be properly classified. However, the collision-resolution routines have not been tested at these high levels of activity; and, the impact of function collisions upon real-time operation is unknown.
Human preferences may be a limiting factor.
XI. Future Research
(1)Subsidiaries and Consolidation
The identification of reciprocal entries between corporations of a controlled group should be quickly organized. Any lack of reciprocity should also be quickly identified. No test or research has been done on this topic.
Budget transfers are frequently a problem in fund Accounting. This method can identify, report, and recapitulate those entries which include accounts from more than one fund, providing the information for better management of budgets.
(3)Case Based Reasoning
The Nearest-Neighbor method of case retrieval appears to be enhanced under this procedure. However, the more fundamental change brought by this procedure is the effect of user judgment. The user can change selection criteria and thereby make comparative evaluations. The machine does not make decisions; it merely selects cases quickly according to the user‟s queries. “The Curse of Multidimensionality”, identified in data mining research, suggests that humans may prefer accounting-style presentation of cases.
A Linux-based desktop version of the software is being readied for field tests. The GUI-enabled software has been delayed.. Accounting faculty, particularly those specializing in Audit or Governmental, are invited to test the searches and reconcile the index against existing double-entry Accounting datasets. Hopefully, the same data and the same executions can be used for Accounting reconciliation and for system timing comparisons.
References AICPA Practice Alert 2003-2 AICPA Statement of Audit Standards 99 Chatfield, M. 1974 A History of Accounting Thought. The Dryden Press, Hinsdale IL. De Morgan, A. 1846. Elements of Arithmetic (5th edn), Taylor and Walton, London Hoskin, K. W. and R. H. Macve, 1993, Accounting as Discipline: The Overlooked Supplement, Knowledges, Historical and Critical Studies in Disciplinarity, Eds, Messer-Davidow, Shumway, and Sylvan, University Press of Virginia. Leech, S. A. 1986 The Theory and Development of a Matrix-Based Accounting System. Accounting and Business Research, Autumn. Koshimura, S. 1988. Accounting Matrix. (trans) Manzen. Laskaris, N. and S. Fotopoulos 1997. An Efficient Data Mining Technique with Applications in Multivariate Image Analysis. 13th International Conference on Digital Signal Processing Proceedings. 1103. Mattesich, R. 1964. Accounting and Analytical Methods. Richard D. Irwin Paton, W. A., and A. C. Littleton. 1940. An Introduction to Corporate Accounting Standards. AAA. Columbus OH. Salton, G, A. Wong, and C. S. Yang. 1975. A Vector Space Model for Automatic Indexing. Communications of the ACM 18 (11) (November): 613 - 620. Sampson, W. C.. 1995. Method and Electronic Apparatus for Performing Bookkeeping. U. S. Patent No. 5,390,113. Sampson, W. C. 1996. Transaction Index: A Tool for Auditors. Internal Auditing (Spring) 16 24. Sampson, W. C. and M. J. Olan. 1992. An Innovation in Auditing: Matrix Summaries of Journal Entries. ABACUS 28 (2) (September): 133 - 141. Sampson, W. C. and M. J. Olan. 1993. Method and Electronic Apparatus for the classification of Combinatorial Data for the Summarization and/or Tabulation Thereof. U.S. Patent No. 5,212,639.
Sampson, W. C. and D. Z. Douthat. 2000. Method for the Organizational Indexing, Storage, and Retrieval of Data According to Data Pattern Signatures. U. S. Patent No. 6,058,392. Sorter, G. H. 1969 An Events‟ Approach to Basic Accounting Theory. The Accounting Review 44 (1): 12-19. Sutton S. G., V. Arnold and D. Havelka 1998 Reply. Auditing, a Journal of Practice and Theory 17 supplement 103-107 Vasarhelyi, M. K. 1998 Toward An Intelligent Audit: Online Reporting, Online Audit, and other Assurance Services. Advances in Accounting Information Systems 7: 207-221.
Web References Jenson, Richard L., 1998 Reengineering as a Framework for Accounting System Study. http//:www.sbaer.uca.edu./Research/1998/WDSI/98wds015.txt Sampson, Wesley C., 1998 Suppose Pacioli Had a Computer. http//:www.data-pattern.com/pacindex.htm