A Discrete‐valued Risk Function for Modelling Financial Distress in Private Australian Companies

Abstract

This paper emphasises the inappropriateness of continuous measure predictors for both the logit and MDA models when dealing with the measurement errors that exist in much of the private company data used to model financial distress in that sector. Also, it is argued that the step function logit model that we get as a consequence of the necessity to categorise the predictors, may be more appropriate in explaining underlying nonlinear behaviour of firms at risk than the usual continuous response linear function. Within this context, the two models are compared using data from 140 private Australian companies. A logit model based on only three discrete‐valued ratios gave an overall accuracy rate comparable to that of an existing continuous‐valued multiple discriminant analysis (MDA) model based on six ratios. Of interest is the very different order of significance of the predictor ratios in the two models although neither model remains trustworthy for predictive purposes.

Risk management processes used in determining project feasibility in the property development process early stages by Australia/New Zealand property developers

“…Logistic regression (LR) in both binary and multinomial calculations allows for the use of categorical dependent variables and is often used in literature when dealing with firm decision models (Cybinski, 1995). Additionally, this statistical method is being used more frequently in recent times as it often produces superior results to traditional regression methods when describing the relationship between response variables and one or more explanatory variables (Hosmer et al, 2013).…”

“…ANNs have also been shown as a valuable technique for predicting variables when there are many groups (Lapedes and Farber, 1987) and also as a predictor of commercial property values (Connellan and James, 1998). ANN analysis using Multilayer Perceptron is used in this study to determine key differences among respondent attributes and specifically in relation to predictors of hurdle rate selection. Logistic regression (LR) in both binary and multinomial calculations allows for the use of categorical dependent variables and is often used in literature when dealing with firm decision models (Cybinski, 1995). Additionally, this statistical method is being used more frequently in recent times as it often produces superior results to traditional regression methods when describing the relationship between response variables and one or more explanatory variables (Hosmer et al.…”

An Exploratory Study of the Company Reorganisation Decision in Voluntary Administration

“…These studies have developed and refined multivariate financial distress prediction models since the seminal research in this area by Altman (1968). Prior studies of both Australian listed companies (Lincoln 1984, Izan 1984, Castagna and Matolcsy 1981, and unlisted companies (Cybinski 1995, Shailer 1990, McNamara et al 1988) have confirmed the usefulness of financial ratio data in the bankruptcy prediction task in the Australian setting. More recently, financial distress studies have focused on the problem of distinguishing between types of distressed firms rather than distinguishing distressed firms from healthy firms.…”

“…The categorised variable was coded with a value of one to five: five representing ratios with a negative equity position, and values of one to four being quartile groups for ratios with a positive equity value. The effect of applying this method of addressing problems with ratio calculation was tested by Cybinski (1995), who found the original ratio information value was preserved in models developed using categorised independent variables. 7 The dependent variables for statistical models required classification of subject companies based on (1) whether the company reorganised or was liquidated (for the decision event), and (2) whether a company was successful or unsuccessful in reorganisation (for the performance event).…”

“…Cybinski (1995) tested groupings consisting of three and five categories, and found that the five category grouping gave a slightly superior "goodness of fit" of the model to data. Cybinski's study did not aim to find the best categorisation regime, however, it did indicate that reliable models could be developed using discrete categorisation of financial ratio independent variables.…”