Credit Score Cards — Marking the Defaulters
One of the most important parts of building Credit Score Cards is creating a default definition that is — at what action of the customer are we going to declare him/her as Defaulted or Gone Bad. To break down even further, defaulted here implies that the customer is unable to repay the loan provided by the Financial Institution.
Why is Default Definition Important?
Carve it in stone that default definition is the cornerstone around which the entire story of Credit Score Card revolves. It is what photos are to Instagram and Tesla to Elon Musk — yes, that important. To jot down in points, default event is used to –
1) Set the performance of applicants over which a Credit Score Card is modeled
2) Access the risk of a portfolio/market
3) Build the profile of risky customers
This is definitely not a thorough list but gives an idea about the usage of a default event. Having provided the context, let’s breakthrough into the exciting stuff.
How is Default Event Constructed?
There are two main components that go into the construction of Default Definition.
a) Default Event
b) Performance Window
a) Default Event –
For my ease of explanation and an eventual assurity of your grasp, I’ll take an example here. Suppose a customer applies for a credit card and gets approved. The bill of the credit card is generated every month and the customer has to pay back it in full or the minimum due (Depends on the customer) within a month’s period to stay current on the card. However, either due to low ability or willingness to pay, he/she misses the repayment within that period. That’s where the status of the customer changes from “Current” to “Delinquent”.
The exact status of a customer goes in the following stages –
Current → 30 Days Past Due → 60 Days Past Due → 90 Days Past Due → 120 Days Past Due and so on.
Now, here’s where defining the default event kicks in. To keep things simple, I’ll not consider the role of local regulatory norms (The stupidest crime that can be committed).
The Financial Institution can now decide to choose a status for a customer upon which it will change his/her status from “Delinquent” to “Defaulted”. That means the customer has gone bad and the chances of recovery (repaying the loan) are very slim. This decision of choosing a status among 30 DPB, 60 DPB, 90 DPB, and so on is based on the past behavior of applicants and determining that “Point of No-Return”.
In this example, let’s assume that the default event has been marked as 90 Days Past Due.
b) Performance Window -
Now that we have defined the default event, the concept of Performance Window comes into the picture that is over what period are we going to monitor the behavior of a customer, post-acquisition. If in this example, we take a performance window of 12 months that means whether a customer has gone 90 Days Past Due in a span of 12 Months. If yes, we will tag them as a defaulter.
The performance window can change through the need of the business and product.
One key thing to point out is that a very major part of constructing Default Definition is dependent on the local regulatory norms which are outside the scope of this piece and for now, my knowledge.
Once we have the default definition handy, the team can go in and tag defaulters over a sample window of applicants and generate what we call in modeling — The Dependent Variable. This is the field around which our model will now be trained and tested to build a Credit Score Card. Therefore, it’s highly imperative that we get this step correct. Easy peasy, right?
“A powerful model is the one where the default predictions differentiate the population in such a way that there is a maximum separation between actual defaulters and non-defaulters”.
Please feel free to reach out to me at p17anshikap@iimidr.ac.in or Linkedin — https://www.linkedin.com/in/anshika-parihar-bb69a7145/
