International Finance and Accounting Handbook

racy for the same periods was 89.6%, 89.6%, and 85.4% respectively. The prediction
accuracy on a small secondary sample (holdout) of ten failed firms was 100% one
year prior to failure, 70% two years prior and 40 percent three years prior. In the ab-
sence of the corresponding Type II accuracy, this result is difficult to interpret, how-
ever. The authors believe that the model is sufficiently robust as to be applicable to a
cross-section of firms and industries.

10.12 GREECE

(a) Gloubos and Grammatikos (1988). Companies in regulated economies are often
sustained in operation long after they have become economically bankrupt. This
causes taxonomic problems for the researcher because to treat such companies as
healthy is clearly wrong, while including them in the bankrupt group causes biases
because of the difficulties in identifying the date of the bankruptcy. The authors sug-
gest that estimated models in such economies as Greece may be expected to have a
higher degree of misclassification than similar models estimated in market-driven
economies. In this study the authors compare four techniques on a “new” sample of
healthy and bankrupt firms. The four techniques used are:

1.Linear Probability Model (LPM)

2.Probit Analysis (PROBIT)

10 • 28 BUSINESS FAILURE CLASSIFICATION MODELS

Standardized

Univariate F Coefficient Wilk’s Lambda

Variable Amount Rank Amount Rank Amount Rank Stepwise

EBIT/TA 26.4 3 0.79 3 0.23 5 3
EBIT/Interest 49.2 1 0.66 1 0.53 1 1
CA/CL 4.3 5 0.96 5 0.24 4 5
FD/SF 21.6 4 0.82 4 –0.25 3 4
MV/TL 36.9 2 0.72 2 0.44 2 2

Exhibit 10.8. Relative Contribution Tests and Ranks of Variables in the Distress Model.

Forward

Industry Relative

Ratios Raw Ratios

Actual No of

Classified Classified

Group Cases Failed Nonfailed Failed Nonfailed

Failed 51 48 3 46 5

(94.1%) (7.8%) (90.2) (9.8%)

Nonfailed 48 5 43 5 43

(10.4%) (89.6%) (10.4%) (89.6%)

Exhibit 10.9. Classification Accuracy of the Industry Relative and the Raw Ratio Models.