The study involves a multivariate discriminant analysis (MDA) of financial ratios of 172 small businesses that during 1971-1976 borrowed fom 29 banking institutions representing 174 Florida banks. Of the 172 borrowers, 92 made timely repayment and 80 defaulted their loans. The study also contains an extensive review and evaluation of previous contributions to ratio analysis. / The study analyzes 54 industry and size-relative financial ratio variables ("relatives") for their usefulness as indicators of loan outcome. The determination of a value of a relative involves (1) computing the value of a financial ratio from financial data submitted by a borrower to a bank prior to a loan and (2) dividing that value by the mean value for a similar ratio computed from data appearing in the Robert Morris Associates (RMA) Annual Statement Studies for the period approximately concurrent with the borrower's financial statements. / The sample of 172 borrowers was split into two subsamples, each consisting of 40 defaulters and 46 non-defaulters. Analysis of one of the subsamples showed that dispersions of ratio values for defaulters and nondefaulters in the sample differed greatly. Consistent with that difference, a discriminant model using a quadratic decision rule proved more effective than a linear discriminant function for separating defaulters from nondefaulters. / The analysis produces a quadratic model containing 39 variables that correctly classifies 96.5% of the defaulters and nondefaulters in the subsample analyzed. For validation, the model was tested for efficacy in classifying the second subsample. The model successfully classifies 63% of the borrowers in the validation subsample. This result was tested for significance using normal approximation of the binomial distribution and was found significant at the .01 level. / The study concludes that information in small business financial statements is useful for short-term bank loan decisions, despite apparent inaccuracies in some of the unaudited data. The study also finds that MDA is effective for analyzing financial ratios as indicators of loan outcome. / Data for this study consists of a sample drawn from a population of loan recipients, rather than a population of loan applicants. Accordingly, if the model developed were to be used by a lender it should be regarded as an appellate model whose use is limited to reviewing decisions to lend that have been made using independent alternative models. / The study concludes that progress on three fronts is needed if empirically-derived, statistically sophisticated discriminant models are to play an important role in lending decisions. First, larger quantities of data must be collected for analysis and the reliability of data must be improved. Reviews or audits by CPA's of a larger percentage of the financial statements of small business borrowers are apparent means for enhancing data reliability. Second, analytical methods require additional refinement. In particular, methods are needed for selecting the variables to be retained for use in a nonlinear discriminant model. Presently available techniques appear to be less satisfactory for selecting variables for retention in nonlinear discriminant models as compared with linear discriminant functions. / Third, lenders must be educated as to the usefulness and the limitations of statistical lending decision models. Only if lenders are convinced of the potential usefulness of the models and the financial data on which they are derived will those lenders require from borrowers and provide to researchers the quality and quantity of data needed for continued progress in financial ratio analysis. Only if they are convinced of potentially beneficial results will lenders adequately test statistical models in real decision situations. / Source: Dissertation Abstracts International, Volume: 41-02, Section: A, page: 0714. / Thesis (D.B.A.)--The Florida State University, 1979.
| Identifer | oai:union.ndltd.org:fsu.edu/oai:fsu.digital.flvc.org:fsu_74085 |
| Contributors | SELPH, CARL JACKSON., The Florida State University |
| Source Sets | Florida State University |
| Detected Language | English |
| Type | Text |
| Format | 207 p. |
| Rights | On campus use only. |
| Relation | Dissertation Abstracts International |
Page generated in 0.0018 seconds