How is the Loss Component calculated in the SMA framework, and what qualifies as an operational risk loss for inclusion?
I understand the SMA formula at a high level for FRM Part II, but I'm unclear on the Loss Component details. How exactly are the 10-year average annual losses computed? Are there thresholds for which losses get included? And what happens to banks that have had a single catastrophic loss event -- does that one event distort the capital requirement for a decade?
The Loss Component (LC) is the internal-loss-data-driven element of the SMA that adjusts capital based on a bank's actual operational risk experience. It is calculated as a multiple of the average annual operational risk losses over a minimum 10-year observation period, using only losses above a specified threshold.\n\nLoss Component Formula:\n\nLC = 7 x Average Annual Op Risk Losses (10 years) + 7 x Average Annual Op Risk Losses above EUR 10M (10 years) + 5 x Average Annual Op Risk Losses above EUR 100M (10 years)\n\nThis three-bucket structure means larger losses receive disproportionately higher weight, reflecting that tail severity matters more than frequency for operational risk.\n\nWorked Example -- Ashford International Bank:\n\n10-year operational loss data (EUR millions):\n\n| Year | Total Losses | Losses > EUR 10M | Losses > EUR 100M |\n|---|---|---|---|\n| 2016 | 42 | 28 | 0 |\n| 2017 | 55 | 35 | 0 |\n| 2018 | 38 | 15 | 0 |\n| 2019 | 310 | 290 | 180 |\n| 2020 | 48 | 22 | 0 |\n| 2021 | 61 | 40 | 0 |\n| 2022 | 44 | 18 | 0 |\n| 2023 | 52 | 30 | 0 |\n| 2024 | 39 | 12 | 0 |\n| 2025 | 47 | 20 | 0 |\n| 10Y Total | 736 | 510 | 180 |\n| Annual Average | 73.6 | 51.0 | 18.0 |\n\nLC = 7 x 73.6 + 7 x 51.0 + 5 x 18.0 = 515.2 + 357.0 + 90.0 = EUR 962.2M\n\n`mermaid\ngraph TD\n A[\"All Op Risk Losses
(above EUR 20K threshold)\"] --> B[\"Bucket 1: Total losses
Weight = 7x\"]\n A --> C[\"Bucket 2: Losses > EUR 10M
Weight = 7x\"]\n A --> D[\"Bucket 3: Losses > EUR 100M
Weight = 5x\"]\n B --> E[\"LC = 7 x Avg_all
+ 7 x Avg_>10M
+ 5 x Avg_>100M\"]\n C --> E\n D --> E\n E --> F[\"Catastrophic single events
dominate for 10 years\"]\n`\n\nThe Catastrophic Loss Problem:\n\nIn Ashford's case, the 2019 event (a EUR 310M rogue trading loss) dominates the Loss Component for a full decade. Without that single year:\n- Annual average total losses would be ~47M (not 73.6M)\n- LC would be approximately 7 x 47 + 7 x 24 + 5 x 0 = 497M\n\nThe 2019 event adds approximately EUR 465M to the Loss Component, translating to roughly EUR 465M x ILM in additional capital, held for 10 years.\n\nQualifying Losses:\n- Minimum threshold: EUR 20,000 (losses below this are excluded)\n- Gross losses before any recoveries (though recoveries net against the same year)\n- Include: litigation settlements, regulatory fines, fraud losses, system failures, execution errors\n- Exclude: credit losses (unless arising from operational events), market losses, strategic losses\n- Timing: losses are dated to the accounting recognition date, not the event date\n\nData Quality Requirements:\n- Banks must maintain a comprehensive, validated loss database\n- External auditor review of loss data is required\n- Losses must be mapped to Basel event type categories\n- Near-miss events should be captured but do not enter the LC calculation\n\nPractice SMA capital calculations in our FRM question bank.
Master Part II with our FRM Course
64 lessons · 120+ hours· Expert instruction
Related Questions
How is the swap rate curve constructed, and why does bootstrapping from deposit rates to swap rates matter for valuation?
Why did the industry shift to OIS discounting for collateralized derivatives, and how does it differ from LIBOR discounting?
How does a knock-in barrier option actually activate, and what determines its value before the barrier is breached?
How does linear interpolation work on a bootstrapped yield curve, and what artifacts does it introduce?
How does the cheapest-to-deliver switch option work in Treasury bond futures, and when does the CTD bond change?
Join the Discussion
Ask questions and get expert answers.