Survival Models in Retail Banking
Application of Survival Modeling in Personal Loan1
Credit Scoring has been in focus for some time and gradually focus is shifting to maximize profitability and not just minimizing risk. Hence time of default (customer not able to meets its obligation of payment) and prepayment (paying of Personal Loan before its term). So, survival modeling techniques could be of great use for
- Estimating time of default along with probability of default
- Predicting time of prepayment for Personal Loan and Mortgage
Survival Analysis for PD and LGD Modeling2,5
In Basel Framework, Probability of Default (PD), Exposure at Default (EAD) and Loss Given Default (LGD) are 3 scoring models being used. In PD Model, a time window (Performance Window) is fixed for defining target variable (whether a customer has defaulted or not). Survival Model may be helpful in incorporating time of default and a longer analysis period.
Attrition Modeling using Survival Modeling4
For estimating future customer loyalty and profitability at a account level, it may be useful or rather required to predict probability of attrition (closing an account relationship) at various future periods. In this example for TD Bank survival model is used to predict customer likelihood of attrition at different period in next 12 months.