What is Credit Risk
and Why is it Important?
The lender offers credit to a debtor, termed a borrower, based on trust in repayment. Credit risk emerges when there's a risk the borrower might not repay, leading to the lender not receiving the principal or interest, thereby incurring significant financial losses. Additionally, substantial costs are often incurred in efforts to recover the loan amount.
What is Credit Risk and why is it important?
The lender provides credit to a debtor who can be referred to as a borrower based on the trust that the borrower will repay the lender. Therefore, the likelihood that a borrower would not repay their loan to the lender is called credit risk. In this case the lender would not receive the owed principal. Moreover, they would not be paid the interest due to the fact that the borrower will not be able to pay the loan and will therefore suffer a substantial loss. In addition, it is likely that the lender will have to sustain substantial costs to recover the amount.
Executive Summary
The event of a borrower not being able to make the required payments to repay their debt is associated with each borrower’s characteristics. One way for lenders to counter losses due to borrower defaults is to require collaterals that would cover the outstanding debt and/or increase the price of the lending funds. This is commonly known as risk-based pricing.
The most important factor is to be able to estimate the credit risk of each borrower as precisely as possible. It's important to know that lenders' inability or failure to estimate borrowers' probability of default can have grave consequences for lenders and society. In general, lending to borrowers with a high probability of default is one of the main reasons for a serious financial crisis such as the global financial crisis in 2008.
Define Dependent Variable
The analysis on historical data is performed by defining the binary variable which has the values of 1 and 0. Variable 1 refers to defaulted borrowers and 0 refers to non-defaulted borrowers. Defaulted borrowers are borrowers who defaulted in payment. By 'default', it means if either or all the following scenarios have taken place:
- Payment is due for more than 90 days. In some countries, it is 120 or 180 days
- Borrower has filed for bankruptcy
- Loan is partially or fully written off
Steps of PD Modeling
- Data preparation
- Variable selection
- Model development
- Model validation
- Calibration
- Independent validation
- Model implementation
- Periodic monitoring
Statistical Techniques used for Model Development
- Logistic Regression is the most widely used technique for estimating PD
- Survival Analysis is generally used to compute lifetime PD
- Random Forest, Gradient Boosting and ANNs are also used techniques for estimating PD
Machine Learning Models Comparison
It can be seen that the Random Forest algorithm has the highest AUC score, however, ANNs has the highest accuracy score.
Expected Loss and its components: PD, LGD and EAD
It is normal for lenders to incur credit losses from every portfolio or exposure over a given period of time. In other words, lenders know that there is a certain amount of credit risk associated with every borrower. They expect that there could be a loss when lending to any borrower. Such expected loss comes as a result of different types of factors, borrower specific factors, the economic environment or a combination of the two.
In addition, lenders may also incur unexpected losses. These are most likely the results of adverse economic circumstances that affect the whole economic environment. It is highly unlikely, but still possible that lenders suffer exceptional losses, for example, due to a severe economic downturn. The ways to estimate expected loss, which is associated with credit risk are also called expected credit losses on exposure level. Expected loss is the amount a lender might lose by lending to a borrower. There may be many different approaches to estimate and forecast that amount. However, the established credit risk modeling framework defines expected loss as the product of three components namely: probability of default, loss given default and exposure at default.
Lenders build different statistical models to estimate each of these components and then multiply them to obtain the expected loss for a given exposure level. PD is defined as the likelihood that a borrower would be unable or unwilling to repay their debt in full or on time. In other words, it is an estimate of the likelihood of the borrower defaulting and usually refers to a particular time horizon. LGD is the share of an asset that is lost and unrecoverable when the borrower defaults. Therefore, it’s the proportion of the total exposure that cannot be recovered by the lender. EAD is the remaining proportion of the principal at a specific time horizon and is defined as the remaining proportion of the loan that is exposed when the borrower defaults. Therefore, it is the maximum amount a bank may lose.
The graph below shows the monthly and cumulative loss generated by the platform.
