The Use of Structural Equation Modelling to Select Financial Indicators for a Bankruptcy-prediction Model

Abstract

The classical tool of bankruptcy prediction is the multivariate discriminant Altman model. The aim of this paper is to present a proposal for the use of structural equation modelling (SEM) to select financial indicators for an Altman-type bankruptcy prediction model. Financial factors, as diagnostic variables in bankruptcy-prediction models, are not in fact directly measurable variables, and they ought to be recognised as latent variables described by various measured financial indicators. So it is possible to use a structural equation modelling (SEM) approach for this purpose. A path diagram in terms of SEM for the Altman model is presented. Based on this diagram, three variants of SEM models for the general Altman model are estimated. The essential problem tackled in this paper is how to appropriately select non-bankrupt firms. Matching pair sample selection methods are applied. The non-bankrupt firms are from the same branch of industry and are similar in size. The major objective of our methodological proposal to use a general SEM model to study corporate bankruptcy is to overcome the difficulties in the modelling of bankruptcy risk through the use of previously-applied methods.