Under the going-concern concept, an entity is assumed to be a going concern when it is able and willing to continue operations in the foreseeable future. Although substantial agreement exists as to the meaning and role of the going-concern concept, it is difficult to make going-concern assessments in the course of an audit. In particular, existing auditing guidelines contained in SAS No. 34 are inadequate and existing going-concern prediction models are flawed. In view of this, the objective of the dissertation is to construct a going-concern prediction model (hereafter called the Koh model) that is based upon improved statistical techniques and methodology.

A sample of 165 companies that filed for bankruptcy during the period 1980 to 1985 and a matched sample of 165 non-bankrupt companies are used to construct and test the Koh model. Following the lead taken by the proposed SAS on going-concern assessments, a non-going concern is operationalized as a bankrupt company. For each of the sample companies, six financial ratios as specified by the proposed theory of bankruptcy are obtained. Probit analysis with the weighted exogenous sample maximum likelihood procedure is used to estimate the coefficients of the Koh model. Using the Lachenbruch U method, the hold-out accuracy rates of the Koh model are computed. They are 85.45% for non-going concerns, 100.00% for going concerns, and 99.91% overall. With these accuracy rates, the Koh model compares favorably with other going-concern prediction models suggested in the literature and the auditors.

The effects of misclassification costs of Type I and Type II errors on the Koh model are also considered. It is found that the optimal cut-off probability for the Koh model is very insensitive to varying relative misclassification costs. Coupled with its high predictive ability and stability, the Koh model can be an effective prediction model, analytical tool, and defensive device for auditors. Further, the methodology developed and employed in the dissertation can contribute to the current state-of-the-art in constructing prediction models such as going-concern or bankruptcy prediction models, takeover/acquisition prediction models, and loan default prediction models.