Synopsis: In today’s blog, we will elaborate on some of the basics of credit risk models that lenders utilize to evaluate a borrower’s creditworthiness. 

Over the recent years, the lending landscape, particularly in the context of credit risk management, has undergone a sea change. Several lending organizations, including banks, NBFCs, and fintechs, have developed sophisticated risk management systems– with some even employing artificial intelligence (AI) and machine learning (ML) models to ascertain a borrower’s credit risk. 

Such credit model validation enables financial companies to make key lending decisions, i.e., whether to sanction a loan. Credit modeling has materially improved internal risk management practices and profitability by measuring, aggregating, and managing risk across a variety of loan instruments. 

So, what are credit risk models, and what really goes into developing them? Let’s find out. But, first, let us elaborate on the concept of credit risk.

What is Credit Risk?

Credit risk is the probability that a borrower will default on their debt obligations. In other words, it is the likelihood of a borrower failing to make interest and principal payments. 

Commonly, traditional lenders charge high-interest rates to borrowers with high credit risk. To illustrate, if an MSME has poor credit (CIBIL) scores and a weak financial position, it will have to fork out higher interest payments, assuming they are extended a loan at all. 

So, when a borrower approaches a bank or a financial company for a loan, the lender must evaluate their creditworthiness through risk models before making a decision. This process is based on the 5Cs of credit: character, capital, conditions, capacity, and collateral.

What are Credit Risk Models?

A loan extension to high-risk borrowers can drastically affect lenders’ cash flows on repayment defaults. Lenders may even incur additional collection costs by hiring a debt collection agency to recover the loan principal amount. Hence, it becomes imperative to utilize advanced credit risk models to mitigate any future losses. 

Credit risk models are data models, which compute the probability of a borrower’s default and its impact on the lender’s financials. This process is known as credit risk modeling. Traditionally, it was done by pouring over a borrower’s balance sheet and financial position, along with insights gained by the credit analyst from interviews with them.

As technology has matured, new ways of estimating credit risks, including credit risk modeling with machine learning, have emerged. These include the use of cutting-edge data analytics (using Python and R) and big data techniques to model credit risk. 

Moreover, with the economic expansion, several diverse kinds of credit risks have cropped up, thus influencing credit risk models even more. Below, we spotlight the major credit risk categories that are accounted for in credit risk models.

Types of Credit Risk

Credit risk models evaluate the following credit risks:

1. Credit Default Risk

First and foremost, credit risk models are concerned with the borrower’s default risk, which is their abject failure in making interest and debt payments. As per the non-performing assets (NPA) framework, it becomes a major cause of concern when the payments are delayed 90 days past the due date. 

This risk level may change owing to a dynamic economic outlook, like a recession or mounting inflation. It may worsen further if a borrower’s situation changes radically due to a harsh competitive landscape or a temporary stalling of payments.

2. Concentration Risk

In addition to borrowers’ capacity, lenders are also concerned about their suppliers and buyers. For instance, if an MSME primarily sells to one major distributor, its default risk may jump suddenly in case its distributor becomes insolvent.

So, concentration risk is basically the risk of high exposure to a single counterparty or sector. Likewise, lenders must diversify their own credit portfolios, without being unduly exposed to specific borrowers or sectors.

3. Country Risk

Finally, country risk is the risk of a payment default over a payment freeze imposed by a foreign country (as happened in the case of sanctions imposed on Russian companies over the Ukrainian war). Such risks may also increase if the foreign country is politically unstable or has poor financial positioning.

What are the Parameters of Credit Risk Models?

To minimize credit risk exposure while maintaining healthy loan books, lenders must employ sophisticated credit risk models that are both scientific and ensure great accuracy. The main aim is to compute expected loss, which is the product of the probability of default, exposure at default, and loss given default. We elaborate on them below.

1. Probability of Default (PD)

As the name suggests, the probability of default is associated with determining the likelihood of a borrower defaulting on their obligations over a particular time horizon. Generally, it is calculated based on historical credit performance and default data, like credit scores and debt-to-income ratio.

Poor credit scores and high default probabilities result in loans that are sanctioned at very high-interest rates. However, these rates can be managed by pledging security or basing loans on cash flows.

2. Exposure at Default (EAD)

Exposure at Default (EAD) is the amount of loan outstanding at the time of a borrower’s default. It is an indicator of the risk appetite of the lender. For example, if a loan of Rs. 1 crore is sought, and the borrower defaults after 3 years when the amount outstanding is R. 75 lakhs, then EAD is Rs. 75 lakhs.

EAD is a highly dynamic concept, as this metric changes every time the borrower pays a loan installment. Also, it is a conservative measure as it does not account for any recoveries.

3. Loss Given Default (LGD)

Loss given default (LGD) refers to the loss incurred by the lender when a borrower defaults on their loan. It is the best-case scenario, as lenders ascertain their losses after estimating the probability of some recovery.

For instance, if a lender sanctions a loan worth Rs. 2 crores; this entire amount may not be LGD. Financial companies will also allow for collateral, down payments, or any installments already made, so the loss will be smaller than the amount loaned.

While there are several ways of calculating LGD, one method commonly used is where,

LGD = EAD * (1 – Recovery Rate)

Here, the recovery rate is an estimate of the percentage of the loan that may be recovered by selling the collateral or through other recovery methods. 

The Key Takeaway

With the onslaught of new credit products, such as embedded financing, cash-flow-based loans, buy now pay later loans, etc., along with technological advances, lenders have been upping their game on the ways by which they model credit risk. The access to aggregated and alternate data has popularized the use of ML-driven and ANN models.

However, credit risk models form only one aspect of managing credit risk; stress testing and scenario analysis techniques supplement them for a better estimation of a potential loss on defaults.