You are not currently logged in.
Access JSTOR through your library or other institution:
If You Use a Screen ReaderThis content is available through Read Online (Free) program, which relies on page scans. Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
Recognizing Financial Distress Patterns Using a Neural Network Tool
Pamela K. Coats and L. Franklin Fant
Vol. 22, No. 3 (Autumn, 1993), pp. 142-155
Stable URL: http://www.jstor.org/stable/3665934
Page Count: 14
You can always find the topics here!Topics: Artificial neural networks, Training, Analytical forecasting, Bankruptcy, Modeling, Discriminant analysis, Standard deviation, Financial ratios, Statistics, Statistical median
Were these topics helpful?See something inaccurate? Let us know!
Select the topics that are inaccurate.
Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
Preview not available
This study builds neural networks (NNs) which estimate the future financial health of firms. A neural network is a relatively new mathematical approach for recognizing discriminating patterns in data. We use NNs here to identify financial data patterns which consistently distinguish generally healthy firms from distressed ones. The purpose is to detect early warning signals of distressful conditions in currently viable firms. Being able to form highly reliable early forecasts of the future health of firms is critical to bank lending officers, investors, market analysis, portfolio managers, insurers, and many others in the field of finance. The traditional approach and present standard for predicting financial distress uses multiple discriminant analyses (MDA), MDA is a statistical means of weighting the relative value of information provided by a combination of financial ratios. But MDA has been sharply criticized because the validity of its results hinges on restrictive assumptions. These restrictions are, in many cases, incompatible with the complex nature, boundaries and interrelationships of financial ratios. In such cases, the power of MDA is compromised and the results may be erroneous. Other studies have suggested alternatives to MDA, including logit, probit, recursive partitioning, expert systems, and nonparametric models. However, none of these approaches has replaced MDA as the standard for comparison. We show that NNs are a promising alternative to MDA. The paper presents Cascade-Correlation, the particular NN used in our comparison study of NN and MDA models for predicting financial distress. We discuss how the NN computer-implemented training process works, how it autonomously gleans relationships from the data, and how it builds a unique neural network structure. The completely flexible NN method for capturing and communicating knowledge about the data allows a neural network to uncover complex, imbedded patterns that other techniques cannot detect or describe. Our study examines 282 firms which were in operation during the period 1970-1989. Ninety-four of the firms were formally identified and reported by their auditors as (at some point over the period) being financially distressed. The remainder (188 firms) were reported by their auditors to be healthy and viable. Specifically, we looked at five ratios widely considered to be prime determinants of financial health: working capital/total assets, retained earnings/total assets, earnings before interest and taxes/total assets, market value of equity/book value of total debt, and sales/total assets. Half of the firms of each type were used to develop NN and MDA models and the rest served as a test sample. Our decision to use auditors' reports rather than the traditional bankruptcy filing as our indicator of financial distress was based on a desire to focus on the "practical" relevance of a correct prediction. Bankruptcy of a firm may occur after a prolonged period of financial distress. If so, there is little practical use for a predictive model since the distressed nature of the firm is obvious to virtually all of the firm's stakeholders, i.e., shareholders, employees, vendors, etc. On the other hand, the published auditors' reports we used are expected to precede bankruptcy, perhaps quite substantially. In addition, bankruptcy is only one outcome of financial distress. Others include reorganization, liquidation, and acquisition by a viable firm. Regardless of the outcome, losses and risks preceding the final resolution are likely to be incurred by stakeholders. Bad audit reports can cause bond ratings to be lowered, lines of credit to dry up, and other business relationships to be disrupted. Thus, the use of auditors' reports as the prediction criterion covers a broader range of events than does the bankruptcy filing and should have more relevance to decision-makers. The test results suggest that the NN approach is more effective than MDA for the early detection of financial distress developing in firms. The NN models consistently correctly predicted auditors' findings of distress at least 80% of the time over an effective lead time of up to four years. A statistical comparison of results showed that the NNs were always better than the MDA models for identifying firms which eventually received bad audit reports. Thus, NNs appear to be reliable forecasting tools, whose use may be particularly warranted when the costs associated with misclassifying a distressed firm as healthy are high.
Financial Management © 1993 Financial Management Association International