Predicting the Fate of Insolvent Companies in Administrative Receivership: Using Statistical and Neural Classification Methods

The paper focuses upon businesses which are in crisis and attempts to assess whether financial variables can be identified and used to discriminate between insolvent companies which can be rescued and those which cannot. In attempting to resolve this question, the research utilises a sample of companies placed into administrative receivership and applies four classification techniques: two statistical methods - Linear Discriminant Analysis and Logistic Regression; and two neural nets - the Backpropagation net and the Learning Vector Quantization net. The performance of the four different classification procedures was quite comparable and revealed estimated classification rates of between 70 - 80%. The analysis also identified six discriminating variables, but three of these - debtor turnover, gearing ratio and current ratio - were particularly important in predicting outcomes.