FRTB Internal Models Approach: Achieving IMA Approval with Confidence

The Internal Models Approach (IMA) allows banks to use their own risk models to calculate market risk capital requirements. Unlike the Standardised Approach (SA-FRTB), which prescribes fixed risk weights, the IMA is based on Expected Shortfall (ES) at the 97.5% quantile. The goal: more risk-sensitive and potentially lower capital requirements — provided the bank meets strict requirements for model quality, backtesting and P&L attribution. The capital formula reads IMA capital = IMCC + NMRF + DRC.

IMA approval operates at desk level — each trading desk must be individually authorised. Prerequisites are passing backtesting examinations (traffic light approach) and the P&L Attribution Test (PLAT). Desks that fail these tests automatically fall back to the Standardised Approach. Additionally, each desk must demonstrate qualified risk management infrastructure, sufficient data history and independent model validation.

The P&L Attribution Test compares actual trading P&L results with the values predicted by the risk model. The deviation is assessed using statistical thresholds (Spearman correlation and Kolmogorov-Smirnov test). If a desk fails the PLAT, it loses IMA eligibility and must use the more capital-intensive Standardised Approach. The PLAT is therefore the central quality test for model accuracy under FRTB.

The Expected Shortfall model calculates the expected loss beyond the 97.5% quantile — capturing tail risks better than the previous Value-at-Risk. Under the FRTB, risk-class-specific liquidity horizons of 10 to 120 days apply. The IMCC (Internal Model Capital Charge) is calculated from the weighted combination of current and stressed ES values, multiplied by the supervisory factor of 1.5.

Risk factors that fail the Real Price Observation (RPO) test — those not sufficiently supported by observable market prices — are classified as non-modellable (NMRF). A separate stressed Expected Shortfall add-on must be calculated for these factors, which is added to the IMA capital. The NMRF treatment is often the most capital-intensive part of the IMA and requires a systematic data strategy.

Backtesting compares model predictions with actual losses over a period of 250 trading days. Results are assessed using the traffic light approach: green zone (0‑4 exceedances), amber zone (5‑9) and red zone (10+). In the red zone, the supervisor may withdraw IMA approval. Backtesting must be performed and documented at both desk and firm-wide level.

The EU has postponed the IMA application requirement to January 2028, while the FRTB trading book boundary and Standardised Approach already take effect from January 2027. Banks wishing to use the IMA should use the remaining time for model development, data infrastructure, test runs and the supervisory approval process. The EBA has already published final RTS on liquidity horizons, backtesting and risk factor modellability.

ADVISORI guides institutions from strategic assessment (IMA vs. SA: is the IMA worthwhile for which desks?) through technical model development (ES calculation, PLAT, backtesting, NMRF) to supervisory documentation and regulatory dialogue. Our consultants bring experience from IMA approval projects at European banks and support both initial approval and ongoing model monitoring and recalibration.

IMA implementation requires significant investment in model infrastructure, data management and personnel. Typical cost drivers include risk factor modelling, PLAT infrastructure build-out, NMRF data requirements and supervisory documentation. These are offset by potentially much lower capital requirements — for large trading books, the IMA capital saving compared to the SA can amount to several hundred million euros.

The EBA has published several final Regulatory Technical Standards (RTS) on the IMA: RTS on liquidity horizons for risk factor mapping, RTS on backtesting and P&L attribution requirements, and RTS on risk factor modellability (NMRF criteria). These standards define the technical details of IMA implementation and have been published in the EU Official Journal since November 2022.