Investigating the Usefulness of Accounting Information: Artificial Neural Networks (ANNs) versus Multivariate Analysis Linear Regression Models
Abdulati A. Abdou
Cairo University-Faculty of Commerce
Ahmed M. Badawi
University of Tennessee, Knoxville - Mechanical, Aerospace and Biomedical Engineering Department
Morgan State University
The 2nd Annual Accounting Conference, Future of Accounting and Auditing, Faculty of Commerce, Cairo University, June 2005
The main objective of this study is to investigate the usefulness of accounting information in explaining the stock price performance in the Egyptian stock market. This objective centers on introducing an artificial intelligence technique, namely, Artificial Neural Networks (ANNs), instead of traditional linear regression models often used in prior research. The predictive power of ANNs is tested against the predictive power of traditional linear regression models. Results suggest that the performance of ANNs outperforms that of the traditional linear regression models for some input categories not including the accounting ones. The predictive power of ANNs is almost as same as that of the traditional linear regression models for the input categories that include only accounting variables. The empirical evidence provides little support on the linear regression misspecifications as an explanation for the low documented usefulness (returns-earnings association) of accounting information in prior research.
Keywords: Usefulness, MBAR, ANNs, Association Studies
JEL Classification: G12, M41, M47, C30Accepted Paper Series
Date posted: October 4, 2005 ; Last revised: December 2, 2009
© 2013 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo5 in 0.328 seconds