Insolvency risk: characterisation and prediction

Abstract

The present document sets out to analyse the concept of insolvency risk in a firm and how it can be objectively measured. Our main objective is to predict whether a firm will face an insolvency situation, based on its most recent historical data stored in its accounts. In order to achieve it, the prediction of insolvency risk is studied reviewing some of the most relevant literature and explaining the accounting and financial implications which lie behind it, understanding the concept of insolvency from this perspective. In mathematical terms, this is an example of the so-called Problem of Classification (or Discriminant Analysis), which is usually approached using Statistics. More specifically, the chosen way to mathematically measure insolvency risk is through some of the most popular statistical prediction methods which deal with this problem. Some of these methods consist of the classical Altman’s Z Score, essentially equivalent to the Linear Discriminant, or more contemporary methods like Classification and Regression Trees or Neural Networks. These methods are applied on two samples. The first one is a sample of 40 Spanish firms selected under some certain criteria, gathering its data from SABI database (Sistema de Análisis de Balances Ibéricos). The second one is the sample that Professor E. I. Altman used in his famous 1968 article, where he introduced its aforementioned Z Score. A balanced approach between financial theory and statistical theory is used in order to effectively convey the message that we cannot totally rely on the statistical methods without taking into account the non-mathematical implications, for this is a complex issue involving many other areas such as finance, accounting or economics.