Using Artificial Neural Networks To Examine Semiotic Theories Of Accounting Accruals

Abstract

Although the primary purpose of accounting is to communicate information, few studies have investigated the communicative nature of accounting. This study uses semiotics, a theory of signs and signals, to examine the purpose and usefulness of accounting accruals with a popular tool in forecasting--artificial neural networks. Two primary theories are proposed by this study. The Theory of the Functions of Accounting Accruals categories accounting accruals by their basic functions and states that two general types of accounting accruals exist: syntactic accruals and semantic accruals. Syntactic accounting accruals reflect incomplete transactions under a system of cash receipts and cash disbursements. Semantic accounting accruals present messages in a different format than their counterparts in a system of cash receipts and cash disbursements. The Theory of the Pragmatic Information of Accounting Accruals states that accounting accruals contain pragmatic information (value) because of their functions. The pragmatic information (value) of accounting accruals is examined by comparing the ability of accrual accounting data to forecast future cash flows compared to that of cash-flow accounting data. Forecasts are made using backpropagation neural networks. The results indicate that both syntactic and semantic accounting accruals contain pragmatic information (have value when forecasting future cash flows). The study provides evidence of the pragmatic information (value) of annual accounting accruals. However, no evidence was found in support of the pragmatic information (value) of quarterly accounting accruals. This result implies that annual accrual accounting data is a better predictor of future cash flows than cash-flow data while quarterly accrual accounting data is not.