Retail/Consumer Credit Model - EQUANT Learing

Consumer leading has been sleeping giant for financial sector. Its growth over years made retail banking and consumer lending enormous. But this growth was  largely  ignored till US sub-prime mortgage crisis of 2007 and later worldwide constrain in credit market in consumer lending because  of which banking regulator woke up to its importance.  Prior to banking crisis we would see very less research work in this area compared to corporate lending and derivatives pricing.  Reason of very less research  was due to the fact of  its success which focus on just one objective- to assess default risk of new and existing borrowers  or to put this in simple language whether to accept an applicant for credit or not.  But over few years there has been transition from Application based model to behavioural scoring which extend further credit or to try to cross sell another product to borrower and constantly understand underlying risk by accessing probability of default for each customer on lending book.

But over last decked there has been few major development- 1) There has been efforts to move away from default based scoring to profit scoring – i.e. to develop models which not only based on default risk but also their profitability.  This is in line with dynamic model which checks likelihood of existing customer to purchase additional financial products and also try to access their chance of moving to other lender if better rates are offered or settling their loan earlier. 2) Over last few years as channels of application has increased due to telephone and internet, lenders have been able to customize loan offer to borrower, due to which the problem of how to model such variable pricing decision which are related to similar problems in yield management as one of big issue in lending industry is how different default risks of borrowers are priced.  3) Initially Basel 2 and now Basel 3 wants models to focus on default risk of portfolios of loans and not just individual loans.  For easy understanding it means that one needs to consider how the methodologies used for credit risk of individual consumer loans can be extended to portfolios of such loans. Another reason for modelling portfolio is the need to price portfolios of consumer loans as part of securitization process, where existing methods have failed leading to temporary setback in asset-backed securities market.

So in future article we will concentrate our efforts on some deeper technical issue surrounding Credit Score and related concepts like weight of evidence, information value and decision making.  Also at some point we will measure performance of model build and how well its ability to discriminate between good and bad borrowers, the accuracy of its probability prediction of the chance of borrower defaulting with particular cut-off scores.  Understanding this different type of measurement help validate scorecard by rules driven by BASEL accord which specifies risk and capital requirement for bank and financial institutions.

Before we end this article I want to explore new dimension which has been very successful called profit scoring which uses dynamic scoring models – It has simple idea of risk and return and further ventures into simple stochastic process. Profit scoring has power to support credit limit adjustment and even have basic idea on Bayesian about individual default risk and repayment performance. With individual default risk we can also introduce important concept on survival analysis which estimate when default or other negative events occur rather than if they occur. Using Cox’s hazard model will gives you hazard score- this model helps you building credit life cycle which is extremely useful in building customer lifetime value models.