The Prediction of Small Business Instability--Loan Noncompliance: a Discriminant Analysis Approach.
Date of Award
Doctor of Philosophy (PhD)
The purpose of this study was to determine if a small business' tendency towards loan noncompliance could be ascertained from the business' financial information. By developing linear discriminant models a firm's tendency towards loan noncompliance was accurately determined. For the objectives of this study loan noncompliance was defined as the borrower not complying with the terms of the original loan agreement. Examples of loan noncompliance are: (1) alteration of the loan agreement to the disadvantage of the lending institution, (2) late payment, and missing an interest and/or principal payment. Data for this research was obtained from Robert Morris Associates (RMA) member banks. The information received from each bank was their RMA Data Submission Forms. On each form the bank indicated whether the firm was, or was not, in compliance with their original loan agreement at the end of the firm's fiscal year. A total of 347 firms were received resulting in 51 matched pairs. The discriminant models were developed using a stepwise procedure and the Lachenbruch-Mickey leaving-one-out (LM) validation technique. Additional validation was provided by employing a hold-out sample of complying firms. The model that was most effective at determining a firm's tendency towards non-compliance consisted of: (1) earnings before taxes to total liabilities, (2) cash to current liabilities, and (3) current liabilities to cash flow. The accuracy of the model was 76.5 percent employing the LM validation technique and 62.1 percent based on the holdout sample of complying firms. This shows that a small business' tendency towards loan noncompliance can be effectively determined based on financial ratios. The inclusion of industry and economic data did not enhance the financial ratios ability to indicate the tendency towards noncompliance. Improved specification of the model to include nonfinancial information is likely to be difficult. If qualitative information could be obtained (e.g., number of employees, education of owner, integrity of management, etc.) a better model could possibly be developed. The question of coefficient stability and variable relationship stability could be evaluted by employing inter-temporal testing.
Moores, Charles Thomas, "The Prediction of Small Business Instability--Loan Noncompliance: a Discriminant Analysis Approach." (1982). LSU Historical Dissertations and Theses. 3815.