PREDICTING DEBT REPAYMENT PATTERNS IN SOUTH AFRICA
Date
2011-03-22
Authors
Brodkin, Michael
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Debt recovery agencies and attorneys specializing in high volume debt recovery have in
the past either used judgmental systems or standard credit bureau scores to prioritise
recovery efforts and to make decisions on how to allocate scarce resources. Both of these
methods are insufficient and inferior to statistical credit scoring methodologies that are
tailored specifically for the debt recovery function. The purpose of this research is to
apply credit scoring methodologies to the South African debt recovery environment and
to discover which variables are the most predictive in identifying the re-payment patterns
of written-off, post legal, credit card debt in South Africa.
Bivariate analysis was used to distil six variables which were then subjected to logistic
regression to produce a scorecard which was then tested on a holdout sample of debtors
that were not used in the development of the model.
The research concludes that just six variables produce a scorecard that predicts repayment
significantly better than chance. The research also concludes that standard credit
bureau scores are very poor predictors of re-payment of written-off, post legal, credit card
debt in South Africa.
The relevance of the findings are firstly, that just six variables are necessary to predict
debt repayment patterns in South Africa and secondly, collection agencies and collections
attorneys should not rely on standard credit bureau scores to assess the best re-payment
prospects or to optimise the allocation of scarce resources. Instead, collections agencies
and collection attorneys should develop scoring models that are specific to the debt
recovery function as well as the specific type of debt being worked on.
Description
MBA - WBS
Keywords
Debt repayment patterns, Debt recovery