Credit Underwriting: Minimize credit risk losses using Data Science and Analytics

Growth of customers and portfolios is important for any organization and no different for financial institutions such as bank.

Banks and financial institutions take prudent measures to minimize the risk specifically at the time of acquisition.

Credit underwriting is a process to verify (the document and information provided by the applicants) and assess risk involved in granting credit facility to the applicants.

Data Science and Analytics have been used for many decades to measure the risk of applicants. Specifically, Credit Score is used for assessing credit worthiness of the applicants.  Some credit bureaus provide credit score at a customer, which is used for screening applicants. Financial institution may use credit bureau along with their internal risk scorecard to measure the risk of the applicants.

 

What are key decisions involved in Credit Underwriting?

There are two main decisions involved in granting a loan or credit facility the applicants or customers. First one is whether to grant a loan to a customer and second is how much loan/line of credit can be given to the customer if first decision is “Yes”?

 

Credit Risk Scorecard: What is probability of a customer going bad?

During underwriting process, the bank measures credit risk involved in granting a credit facility to a customer using Predictive Model and this predictive model is called Application Scorecard.

An application scorecard predicts whether a customer, given a credit facility, goes “bad” (not able to meet payment obligation on the credit product) within defined time period or not.  What variables are considered and how credit scorecard is developed?

When a predictive model is developed, it gives probability values between 0 and 1. Based on definition, the probability is the probability of a customer defaulting. The probability values are converted into score value using a transformation. Hence, you will notice that FiCO score has values between 300 and 850 and CIBIL score in India ranges between 300 and 900.  How probability values are converted to score values?

 

How application scorecard cut off value is decided on?

From underwriting perspective, calculation of probability of default or application score value is only part of the solution. The other part is the decision on cut off value.

Credit Scorecard and Underwriting Decision

The banks use different strategy to manage cut off value decision.  In a simple scenario, a score value will be selected as a cut off and based on whether the application score for a customer is higher than the cut off value the application is accepted or reject.  In other scenario, the bank may define range of score values, if applicant score is higher than upper value of the score range, then the applicant will be approved for credit facility. If the score value is within this range, application will be forwarded for manual override decision where a credit officer review and takes decision on whether to approve the credit facility to the customer. And if the application score value is lower than the minimum value of the range, the applicant will be declined for credit facility.

Once an applicant is approved for credit facility next decision, how much credit should be given to the applicant? How credit limit is assigned to an approved applicant?

Most of the credit facility products – e.g. Personal Loan, Home Loan, Credit Card, Home Equality Line of Credit and Vehicle loan – have limit of credit facility approved by the lender (bank or financial institution). Lime of Credit and Credit Limit has different meaning associated with.

In this example, our focus is Credit Limit assignment in case of Credit Card product. The credit limit in the case of credit card is set by the credit card issuer to approve any purchase transaction which does not take the total due balances beyond the credit limit. There are few more details related to Credit Limit and Transaction approvals, these are beyond scope of this blog.

Some of the factors involved in determining credit limit for a customer are

  • Financial Condition of the customers (one of the measure is Credit Risk Score): Less risky customers will be given higher credit limit. Also, sometime credit card issuer consider employment and employer type to grant higher credit limit.
  • Competition: In a competitive and growth scenario, the credit card issuers typically grant higher credit limit

Typically, the decision to approve a credit facility and credit limit assignment is sequential decisions. Some of the banks which have strong analytics capabilities have moved to a “comprehensive decision framework”. Key drivers to move to comprehensive decision framework are

  • Decision based on Customer Life Time Value instead of only based on probability of default
  • Recommending or issuing a product instead of bank accepting or rejecting an applicant for a product. The bank system evaluate scenarios based on the applicant data and review all the available products to match the right product considering defined risk measures. This scenario avoids bias toward pre-selection of a product.
  • Based on Risk Reward Mix the bank can devise appropriate pricing decision framework which is based on value or risk of involved in issuing a credit facility to a customer.

1 thought on “Credit Underwriting: Minimize credit risk losses using Data Science and Analytics”

Leave a Comment