Restatements Due to Improper Revenue Recognition: A Neural Networks Perspective

Abstract

ABSTRACT: The Securities and Exchange Commission (SEC) issued Staff Accounting Bulletin No. 101 (SEC 1999) in an attempt to curb improper revenue recognition practices. Nonetheless, revenue restatements and the subsequent earnings restatements have continued unabated. Our goal is to contribute to the emerging technologies literature by applying the neural networks methodology to the study of revenue restatements. We also compare the results of the neural network classification with classifications obtained from multiple discriminant analysis (MDA) and logistic regression (Logit) models. Six financial and governance variables were used to train the neural network on a sample of 180 firms, and the model was validated using a holdout sample of 51 additional firms. The results show that the neural network model has superior predictive power for predicting revenue restatement firms when compared to the MDA and Logit models. However, the Logit and MDA models predict nonrevenue restatement firms better. Moreover, when misclassification costs are included, the neural network (NN) model performs the best with the lowest relative misclassification costs.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 12 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.