Improved Credit Scoring with Multilevel Statistical Modelling
Loading...
Date
2011-01-18
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This dissertation introduces a new type of credit scoring model which assesses credit
worthiness of applicants for a loan by forecasting their probability of default. The multilevel
scorecard is an improved alternative to the conventional scoring techniques which are
regularly applied in retail banking such as discriminant analysis, decision trees and logistic
regression scorecards. In addition, this thesis proposes a new way of data clustering for a
multilevel structure which is more intuitive and relevant for efficient credit worthiness
assessment.
The motivation for the topic and the core idea of this research project are closely
related to the main advantages of improved credit scoring and its application to the decisionmaking
process in retail banking. The main advantages are accuracy gain and cost-saving.
Improving credit scoring techniques helps to increase operating efficiency by increasing
predictive quality and reducing misclassification errors. From the cost-saving prospective, it
also leads to profit growth and gives a higher return on capital.
In credit scoring the main goal is to define factors which influence riskiness of
individuals who apply for a bank loan. Accordingly, the thesis develops two types of multilevel
structures which allow including random-effects at the higher-level of the hierarchy. The first
structure nests applicants for a loan within second-level groups, microenvironments. Each
microenvironment determines the living area of a borrower with a particular combination of
socio-economic and demographic conditions. Microenvironment-specific effects impact the
riskiness of borrowers additionally to the observed personal characteristics. The second type of
multilevel structure extends the first. It cross-classifies individuals with different
classifications according to similarities in particular characteristics of their occupational
activities, living area condition and infrastructure of shopping facilities in their residence
areas.
Description
Table of contents
Keywords
Credit scoring, Logistic regression scorecards, Retail banking