The authors present three case studies where the PNB-Score was able to correctly
predict the outcome in advance. They also note that the test of the model on over 600
companies showed that the results predicted by the model were found to be relatively
consistent with the actual performance of the companies. The model is very sensitive
to the liabilities of the company, because failure is most often caused when the com-
panies’ cash flows are relatively low compared to its fixed debt commitments. The
study does not present any information on Type II accuracy. It is also not clear
whether the 600 companies tested are all problem companies or they included some
healthy ones as well. To the best of our knowledge, a revised PNB model is still ac-
tively used.
10.19 SINGAPORE
(a) Ta and Seah (1988). Singapore was and still is a dynamic and growing economy
that has attracted a large amount of foreign investment. A business failure prediction
model is justified both for preserving Singapore’s image as a major financial and
manufacturing center and as a way to assist rational investment in Singapore compa-
nies by investors and creditors. This study examines 24 financial ratios using linear
discriminant function analysis.
The failed firm sample consists of 22 firms with failure dates in the period
1975–1983. The failure characteristics of the firms in the sample are as follows: 9%
went into receivership, 18% went into creditors’ voluntary liquidation, while the rest
were involuntary “winding up” by the order of the court. The matched sample con-
sists of 21 nonfailed entities. Only industrial and commercial firms are considered in
the samples. The mean asset size of the firms in the sample is S$89.5 million. The
data sources for the sample are:
- Singapore Registry of Companies and Businesses
- Singapore Stock Exchange
- National University of Singapore’s Financial Database
The discriminant analysis process produced a four-variable model consisting of:1.Total debt/equity
2.Profit before tax/sales
3.Profit before tax/equity
4.Interest payment/profit before interest and taxesThe results of the model on the original sample and a validation (holdout) sample
are reported in Exhibit 10.17. The results for the original sample are based on data
from one year prior to failure. The validation test results are for one and two years
prior.
Although the sample size is relatively small, the results of the model are fairly
good, and its performance is assured as good quality data is available on a larger
number of Singapore companies. The model does suffer from several variables meas-
uring similar firm attributes (e.g. profits).
10 • 40 BUSINESS FAILURE CLASSIFICATION MODELS