are devised based on non-numerical data and the delphi technique (numerous experts
in various areas) is also recommended. Each characteristic is rated over time into the
five categories. The sum of development patterns from varying sources of informa-
tion builds the basis for a final classification. Clustering techniques are also used by
Fischer to clarify information types.
(e) von Stein and Ziegler (1984). This is an ambitious attempt to identify bank-
ruptcy risk from three separate, yet inter-related perspectives. They are: (1) balance
sheet analysis using financial ratios; (2) analysis of the bank accounts of firms, and
(3) analysis of the behavioral characteristics of company management. The study
thus addresses criticism leveled at relying exclusively on one of the three approaches
in assessing failure risk.
The balance sheet analysis considers medium-sized firms in Germany. The failure
dates for the “bads” covered the years from 1971 to 1978. The date for all the
“goods” was fixed (1977). There were 119 failed companies; the failure date was de-
fined as the date of the first value adjustment or write-off, or only in a few cases, the
date of the bankruptcy or composition petition. The “goods” consisted of 327 com-
panies. The companies in the “bad” sample were from the following industries: man-
ufacturing and processing (54.5%), building (17.7%), trade (22.7%), others (5.1%).
The companies in the “goods” sample were comparably distributed across industries.
Thirteen financial ratios were identified as the most discriminating of the 140 ra-
tios initially considered. These ratios are:
- Capital borrowed/total capital
- (Short-term borrowed capital ×360)/total output
- (Accounts payable for purchases and deliveries ×360)/material costs
- (Bill of exchange liabilities + accounts payable for purchases and deliveries ×
360)/total output - (Current assets – short-term borrowed capital)/total output
- Equity/(total assets – liquid assets – real estates and buildings)
- Equity/(tangible property – real estates and buildings)
- Short-term borrowed capital/current assets
- (Working expenditure – depreciation on tangible property)/(liquid assets + ac-
counts receivable for sales and services – short-term borrowed capital) - Operational result/total capital
- (Operational result + depreciation on tangible property)/net turnover
- (Operational result + depreciation on tangible property)/short-term borrowed
capital - (Operational result + depreciation on tangible property)/capital borrowed
Three nonparametric methods (Nearest-Neighbor Classifications: Fix/Hodges,
Loftsgaarden/Quiesenberry and Parzen) and two parametric methods (linear and
quadratic multiple discriminant analysis) were tested. The method of Fix and Hodges
was found to be the most discriminating. The results of the tests on the development
sample are given in Exhibit 10.2.
In the second phase of the analysis, 45 bad and 37 good cases were examined
using the following account characteristic variables:
10 • 10 BUSINESS FAILURE CLASSIFICATION MODELS