Academic journal article International Journal of Business Studies

Forecasting Credit Ratings Using an Ann and Statistical Techniques

Academic journal article International Journal of Business Studies

Forecasting Credit Ratings Using an Ann and Statistical Techniques

Article excerpt

In a liberal environment the conceptual importance of credit rating has increased significantly. The objective of this study is to explore and find out the effect of the financial performance data of a firm relative to the credit rating of a debt issue of that firm. The study also proposes to capture the relationship, if any, between financial performance data and credit rating given by experts in an appropriate model.

Financial data relevant to debt issue ratings are obtained from the publications of a premier credit rating agency in India. Data analysis involved the building of a model using conventional multiple linear discriminant analysis and Artificial Neural Network Systems. Artificial Neural Networks (ANN) model was found to be superior to the discriminant analysis model. The ANN model can be used to increase speed and efficiency of the rating process in practical applications. In addition, if experts provide better-input data, it can be relied upon to provide an automatic rating to a significant extent.

Keywords: Credit Rating, Rating Methodology, Discriminant Analysis, Artificial Neural Network, Experts System.

I. INTRODUCTION

In the present day liberalized environment of many developing countries, the issue of Credit Rating has become a crucial aspect not only for issuers and subscribers of a debt, but also for investors (ranging from individual investors to foreign institutional investors). Credit rating, which is the indicator provided by an autonomous professional agency, reflects the willingness and the ability of the issuers of debt to honor the terms of the client's obligations in terms of repayment of interest and principal payments.

To evaluate a bond's potential, rating agencies rely upon a committee analysis of various aspects of the issuing company such as the issuer's ability to repay, willingness to repay and protective provisions for an issue. It is not known what model, if any, is used for analyzing various issues. It is almost impossible to capture all relevant information used by expert(s) to arrive at a specific rating by quantitative analysis. Subjective evaluation of qualitative information forms significant, if not only part, of the analysis.

This is the main reason why conventional analysis techniques yield very poor results when used for the prediction of ratings. It is possible to develop rule-based expert systems to predict ratings but usually expert information is confidential.

Usually past performance of a firm is reflected in its current financial information and is not necessarily an indicator of future performance. Nevertheless, financial figures may indicate creditworthiness of an on-going firm because of their significant influence on future performance. In addition, they usually reflect a host of subjective data like management effectiveness, competitive position of firm, customer relations, employee morale etc. Subjective and qualitative information used by experts in the rating a debt obligation of a company may also be reflected by financial figures.

The objective of the study is to verify to whether a relationship exists between financial information of a company and the ratings on debt obligations awarded by experts. Assuming the existence of such a relationship, the study proposes to use Artificial Neural Networks, which have a proven ability to capture hidden relationships between the dependent variable, and a set of independent variables and compare the classification and prediction results with results of conventional discriminant analysis.

The practical utility of the study is in the fact that if a significant relationship is found with the help of an Artificial Neural Network, the same can be used as a benchmark from which expert/s can continue his/her analysis. This will considerably reduce the time and efforts' of experts because experts can delegate routine analysis to the Network model and concentrate on qualitative and subjective factors. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.