Expected loss models for credit risk quantification: the Peruvian experience for supervisor and supervised

What it is about

Manuel Luy Molinié, deputy superintendent of Economic Studies at Peru's SBS, presented at the Second Annual Risk Conference (September 27, 2024) the Peruvian experience in moving from provisions based on incurred losses to an expected loss model, from the perspective of both the supervisor and the supervised institutions.

The 34 slides compare both provisioning schemes, explain the loss distribution and the risk parameters (probability of default, loss severity and exposure), describe the characteristics of the Peruvian financial system, the calculation of the parameters with statistical models, and the challenges and the communication strategy with the industry.

Objectives

• Show the limitations of a provisioning model based on incurred losses.
• Explain how the parameters of an expected loss model are estimated with the available information.
• Share the implementation challenges and the communication strategy with the industry.

Key takeaways

1. An incurred-loss scheme does not recognize losses until the event occurs and does not capture the heterogeneity across loan types, institutions and borrowers.
2. Expected loss is calculated as probability of default times loss severity times exposure at default and is tied to provisions; unexpected loss is tied to capital.
3. Knowing each portfolio's loss distribution makes it possible to hold provisions and capital that preserve stability and solvency.
4. The SBS built the parameters with credit bureau data, public information and institutions' data; loss severity based on actual recoveries (workout LGD) was the biggest challenge.
5. Implementation requires time and a lot of data, so it relied on technical meetings, information requests, impact studies and an adjustment timetable.