Accounting choice for measuring investment properties. Data mining techniques contribution to determine decision patterns

Abstract

International Accounting Standard 40 (IAS 40 - Investment properties) offers an ideal setting for research on accounting choice as it represents a paradigmatic case choosing between the fair value and the historical cost as the measurement criteria. In this paper, we take the opportunity of this standard to provide additional evidence in a multinational and multi-context on the determinants that explain the accounting choice. Furthermore, in this paper, we introduce and compare the use of artificial neural networks and decision trees in order to assess the predictive capability of these methodologies, compared to other techniques commonly used to solve classification problems in this area such as the logistic regression. The classification results indicate that both neural networks and decision trees can be an interesting alternative to classical statistical methods such as the logistic regression. In particular, both methods outperformed the logistic regression in terms of predictive ability, although no significant differences were found between both.