| Abstract [eng] |
Credit rating methods were first introduced in 1940 and over the years they have become an increasingly important tool for successful financial institutions. In the 1960’s when credit cards were created, credit rating methods became particularly important to banks. Nowadays, creditworthiness assessment models are widely used as tools for making consumer credit financing or non-financing decisions, or even predicting bankruptcy of companies. The main purpose of creditworthiness assessment is to build a model that establishes a customer's creditworthiness rating using historical data. Credit rating models are significantly improving with the growing amount of information collected. An effective credit rating system can help protect financial sector companies from bankruptcy, maximize profits and minimize the risk of customer insolvency. A wide number of different credit assessment models are currently being developed and they produce different results across different customer segments. The subject of this master's thesis is the market of consumer credit for natural persons. The Scientific Problem: Major problem faced by banks and other financial institutions in Lithuania - how to determine which customer will be “good” and which will be “bad” before providing financial services, how long does it take to determine that the customer is “bad”, which variables best describe a “bad” and “good” customer, and which method to use to get the best possible result. The creditworthiness evaluation of companies is a widely discussed topic in Lithuania, while the analysis of the creditworthiness evaluation of natural persons, although not less important, is not so widely covered. Companies providing financial services in Lithuania often have to buy models for assessing the creditworthiness of natural persons from third parties, because they themselves have very little understanding about creating such models. Therefore, the main problem is to determine which statistical model is best suited for companies providing financial services in the Lithuanian market. Aim of the work: to evaluate the creditworthiness assessment methods used in practice, to understand the importance of risk assessment in business and in the financial sector in particular, to determine which method is best for predicting customer insolvency for a particular financial services company operating in Lithuania. The work consists of literature analysis, methodology and empirical research results. The literature analysis defines the most often used statistical models for natural persons’ creditworthiness evaluation, concept of risk itself and types of risks faced in financial sector. In the empirical analysis section analysis id performed on clients from one of the companies providing financial services in Lithuania. Variable significance tests are made, and models compared to each other. Finally, the paper presents the results of empirical research and their evaluation. The results of the empirical analysis reveal that different models are using different variables to forecast whether or not client will be insolvent and in particular case Random forest model showed the best results while Linear discriminant analysis provided the worst test results. It is recommended that financial institutions would take into account that significant number of variables is important when trying to make a creditworthiness evaluation model. The results showed that models, which are hard to interpret provided the best results. It is recommended to use such models with caution, since it is very hard to understand if it is in line with business plan. There is no universal creditworthiness evaluation model so financial institutions are recommended to develop models themselves so they would represent their client profile. |
