nanced by debt were the major causes of Irish failures. Several of the firms contin-
ued to pay dividends right up to the year prior to failure. On the other hand, only
one company actually made payments to unsecured creditors after insolvency, in-
dicating that asset value had deteriorated beyond repair and only then was failure
declared.
10.17 KOREA
(a) Altman, Kim, and Eom (1995). As a growing and potentially overheated econ-
omy, Korea may be following in the footsteps of its neighbor, Japan, which had a pe-
riod of rapid economic growth only to be followed by increased business failures. For
this reason, the authors suggest, that a failure prediction model for Korea is timely,
even given the 1995 robustness of the South Korean economy. In particular, because
of the increased deregulation and greater autonomy in decision-making by financial
institutions, the availability of predictive models is relevant.
The distress classification model described in this study consists of two versions:
the K1 model is applicable for both public and private firms, whereas the K2 model,
which uses the market value of equity in one of its ratios, may be used only for pub-
licly traded firms.
Linear discriminant analysis was the technique used in building the model. The
sample of failed firms consisted of 34 publicly traded industrial and trading compa-
nies with assets ranging from $13 million to $296 million. Failure and failure dates
were defined based on technical insolvency or liquidation whichever came first.
Technical insolvency is defined as the condition when the credit of a company is no
longer accepted. Most of the failures in the sample occurred in 1991–1992. It is sig-
nificant to note that 30 of the 34 distressed firms had their shares publicly traded only
since 1988, and 23 of the 30 were listed during the explosion of new IPO listings in
1988 and 1989. For this reason, the results of the model may be of interest to in-
vestors and regulators of new issues in the Korean stock market.
Because the nondistressed group of firms tended to be significantly larger in size
on average, the pairing of the healthy firm with the failed firm was based mainly on
industry sector grouping. For 34 distressed firms a larger sample of 61 nonfailed en-
tities was chosen, with the actual one to one pairing done by random selection from
the universe of 61 firms during model building.
The time series analysis of the individual ratio averages revealed that some early
warning financial indicators such as book value of equity to total liabilities do not be-
have in the same way as they do for U.S. firms. This ratio, contrary to expectations,
actually improves for failed firms until just before bankruptcy. However, the same
ratio based on market value behaves as expected. For this reason, the authors have
proceeded with two different models: one employing the book equity leverage vari-
able and the other with a market equity variable.
The criteria for selecting the final variable set were:
- High univariate significance test (see Exhibit 10.14).
- Expected sign for all the model coefficients.
- Original (in-sample) and holdout (out-of-sample) test results.
- Reasonable accuracy levels over time.
The K1 model had the following variables:10.17 KOREA 10 • 37