Defensive interval measures, first introduced by Sorter and Benston in 1960, have been largely ignored in the theoretical and applied literature. In this dissertation, the conceptual superiority of these ratios is explored and empirical investigations are undertaken to determine if these measures actually impart information different from the more traditional liquidity position indicators. Correlation tests of the cross-sectional degree of association between liquidity variables were performed. Significant associations between the traditional and defensive ratios were generally found, although the actual parameter estimates were usually quite small. In a number of other cases, statistical independence was established. These results were corroborated by time-series analyses.

A literature review of bankruptcy studies indicates the important role that liquidity variables play in discriminating between failed and nonfailed firms. In view of the alleged superiority of the defensive intervals, it was postulated that consideration of these refined liquidity measures might improve discriminatory ability. The primary purpose of this dissertation was therefore to investigate the contribution that defensive intervals make to business failure prediction.

Multiple discriminant analysis (MDA) was the basic technique employed to evaluate this contribution. Using ratio sets found to be good predictors in prior research as a starting point, discriminant models were constructed that incorporated various combinations of defensive interval measures. A number of refinements over the typical application of MDA were considered in this model development: a priori odds of group membership were identified; a range of relative costs of misclassification errors was considered; tests of the equality of group dispersion matrices were performed in order to select the appropriate form of statistical analysis; the paired sample design was rejected; and a Bayesian inference approach was adopted to evaluate the models.

Various quadratic MDA models were developed and evaluated, Evidence indicates that incorporating defensive interval measures in the analysis does indeed improve discriminatory ability. Most striking was the improvement noted in the correct classification of failed firms.

The analysis was extended to a comparison of model predictions and going concern evaluations as reported in auditor opinions on financial statement presentations. Evaluation of a subsample of the failed firm population indicated that the selected quadratic models provided advance signals of going concern problems much more frequently than the auditor opinions.

An independent sample was drawn containing companies that had been identified by their auditors as having going concern problems. For those firms that actually filed for bankruptcy, the discriminant models consistently outperformed the auditor opinions in terms of correct classification of going concern status. This advantage extended up to three years prior to the actual filing date. For those firms that did not file for bankruptcy, the models generally indicated going concern problems earlier than the auditor opinions.

Discriminant models which incorporate defensive interval measures can provide some important input to the auditor's going concern review, As demonstrated in this dissertation, these models often provide early signals of imperiled continuing operations and thus may offer the auditor a valuable alternative perspective to consider in going concern evaluations.