Agricultural loan analysis is an area that can benefit from knowledge-based technology such as expert systems. Evaluation of agricultural loan applications requires content specific experience and precludes the use of mechanical procedures alone. Performing an agricultural loan evaluation requires an in-depth understanding of the character of the applicant and the management team, capital structure and financial health of the entity, the ability to service the loan, what security is available to cover the loan and the economic, industry and market environment in which the applicant operates. Knowledge of the applicability of financial modelling techniques is also required. To appraise effectively an applicant's relative financial strengths and weaknesses requires knowledge about all of the relevant factors and their interrelationships, the efficient use of financial data, and the analytical techniques used to assess the financial position.
Computer-assisted loan evaluation has traditionally relied on statistical linear models derived using regression analysis or multivariate discriminate analysis. However, these models are unable to capture subjective and qualitative evaluations. Assessing the strengths of the management team or predicting how well they might cope with a financial crisis are subjective considerations. Informed opinions, even biases, about a firm's business plan, its market share and industry prospects, or the accuracy of the information obtained are also qualitative factors. Subjective and qualitative judgements of this nature are often the key to accurate loan evaluations. They put a critical perspective on the relevant quantitative data.
The purpose of this research is to explore the application of expert system technology to the agricultural loan evaluation process. Specifically, an expert system for evaluating loan applications in the agricultural sector has been developed. The characteristics of the loan evaluation process provide the rationale behind the development of an expert system. Further, the research formulates a framework for building and validating this expert system. The research evaluates loan officer perceptions of the expert system and discusses the implications of its use.
An expert system has been developed in conjunction with two institutions - PIBA and QIDC. Both institutions were selected because they have separate divisions for handling agricultural lending and a high portion of their lending is to the rural sector. The system has been developed from information provided by both banks and validated with reference to their loan officers. The agricultural loan evaluation expert system - ALEES - has 12 knowledge segments and 159 rules in total. The application of the expert system to the loan appraisal process has been successful for PIBA and QIDC - a recommendation from the system matches that of the loan officer in 97% of cases. The results are achieved by evaluating loan proposals from two perspectives, viz. subjective and quantitative. The overall response of loan officers to the system appears highly positive; it is apparent they consider the system as a useful tool for evaluating agricultural loan applications and to help in clarifying their thoughts about the loan application. This finding has particular relevance for the less experienced loan officers, who are the ones that found the expert system most useful. It would suggest the expert system could play a strong role in assisting less experienced loan officers understand the loan evaluation process, and in the provision of a training and education mechanism for loan officers.
The inclusion of subjective criteria in the knowledge base on which to base an assessment of the client provides additional information not available in similar expert systems. This result has potentially important implications for the refinement of agricultural credit assessment models and expert systems for evaluating rural sector loans. Further, loan officers introduced a bias into the loan evaluation process for clients who have financial and subjective assessments marginally outside the institutions accepted limits. Clients were more likely to have their loan application approved if they had a prior credit history with the institution.
An interesting issue is raised as a result of this research in that ALEES could be considered a generic expert system. The system in its current format is not specific to either of the institutions involved in the research. The implication of this contention is that the lending process is the same across all institutions and the expert system could be widely implemented by agricultural lenders. The demand for experienced loan officers could diminish as the users of the system need not be loan officers and further, they may not have training in loan evaluation procedures. However, a decline in decision quality may result. The scenario of a generic system for agricultural loan evaluation is highly unlikely because lending institutions do have specific strategies and requirements for lending. However, re-engineering of the loan evaluation process may occur with the implementation of expert systems and other technologies into an operational environment. The technology is available; lenders need only implement it.