¿Proporciona la auditoría evidencias para detectar y evaluar tensiones financieras latentes? Un diagnóstico comparativo mediante técnicas econométricas e inteligencia artificial

Abstract

Forecasting financial failure is a critical issue for both financial theory and practice, as bankruptcies cause severe effects, not only for shareholders and creditors, but also for third parties throughout the economic system. We have developed a logit model and an artificial neural network to help forecast financial distress based on the information content of audit reports and auditors contracts. These models are built on empirical evidence indicating that it is possible to infer the existence of unrevealed financial pressures, given a small number of quantifiable signs, e.g. changing of auditors, nonfulfillment of formal requirements, and the accumulation of qualified reports. Even with its parsimony, logit model reaches an 85% hit rate, and neural network is able to correctly classify up to 90% of the companies in training, testing and cross-validation phases. We discuss whether the expression of going-concerns in emphasis paragraphs may reduce the ability of investors to evaluate corporate financial risk and/or forecast bankruptcy events.