Knowledge-Based Systems 30 (2012) 67–77

Contents lists available at SciVerse ScienceDirect

Knowledge-Based Systems journal homepage: www.elsevier.com/locate/knosys

Bankruptcy prediction models based on multinorm analysis: An alternative to accounting ratios
Javier de Andrés ⇑, Manuel Landajo, Pedro Lorca
University of Oviedo, Spain

a r t i c l e

i n f o

a b s t r a c t
In this paper we address the bankruptcy prediction problem and outline a procedure to improve the performance of standard classifiers. Our proposal replaces traditional indicators (accounting ratios) with the output of a so-called multinorm analysis. The deviations of each firm from a battery of industry norms (computed by nonparametric quantile regression) are used as input variables for the classifiers. The approach is applied to predict bankruptcy of firms, and tested on a representative data set of Spanish firms. Results indicate that the approach may provide significant improvements in predictive accuracy, both in linear and nonlinear classifiers. Ó 2011 Elsevier B.V. All rights reserved.

Article history: Received 9 February 2011 Received in revised form 2 October 2011 Accepted 3 November 2011 Available online 30 December 2011 Keywords: Bankruptcy prediction Classification techniques Nonparametric methods Quantile regression Accounting ratios

1. Introduction Under the current economic conditions, bankruptcy early warning systems have become tools of key importance in order to guarantee the stability of the economy, as a consequence of their potential to avoid losses to stockholders, creditors, managers and other interested parties. The passing of the Basel II accord makes the need for accurate systems to predict financial distress even stronger, as this agreement establishes that the reserve capital of banks will directly depend on the use of updated models to estimate the