The Fundamental Review of the Trading Book (FRTB) is the Basel Committee's comprehensive overhaul of the market risk capital framework. Finalised in January 2016 and subsequently revised, FRTB replaces the Basel 2.5 regime that proved inadequate during the 2008 financial crisis. The crisis demonstrated that trading book capital requirements had been systematically too low — banks held far less capital than they needed to absorb the losses that materialised when markets dislocated. FRTB addresses this through a more risk-sensitive framework, a harder boundary between the trading book and the banking book, and much stronger requirements for banks that wish to use internal models.

Why Basel III Was Not Enough for Market Risk

Basel III addressed many of the post-crisis failings in bank capital but retained the Basel 2.5 market risk framework as an interim measure. Basel 2.5 had introduced the stressed VaR (SVaR) add-on and the Incremental Risk Charge (IRC), but the fundamental architecture — VaR as the primary risk measure, weak trading/banking book boundary, inconsistent internal model approvals — was preserved. Regulators recognised that a root-and-branch review was necessary.

The specific problems FRTB was designed to address include: the procyclicality of VaR, which produces low capital requirements when markets are calm and spiking requirements when markets stress; the inadequacy of 99% 10-day VaR for capturing tail risk; the permeability of the trading/banking book boundary, which allowed capital arbitrage; and the variability in internal model outputs across banks, which undermined the comparability of capital ratios.

The Trading Book / Banking Book Boundary

The FRTB boundary is one of its most significant reforms. Under the old framework, the boundary was largely determined by the bank's own trading intent — a position was in the trading book if the bank declared it was held for trading. This created obvious opportunities for capital arbitrage: positions could be shifted between books depending on which attracted lower capital charges.

FRTB imposes an objective, presumptive classification. Instruments must be in the trading book if they are held for the purpose of short-term resale, are managed on a fair value basis, or are taken for the purpose of hedging trading book positions. The Basel standard provides a detailed list of instruments that are presumed to be in the trading book, and a corresponding list of instruments presumed to be in the banking book. Transfers between books are heavily restricted and require supervisory approval. Switching a position from one book to the other to reduce capital is explicitly prohibited, and any capital reduction resulting from an approved switch must be reported as a capital add-on.

The Standardised Approach (SA)

All banks must calculate market risk capital under the FRTB Standardised Approach (SA), even if they also use internal models — the SA serves as a floor and a fallback. The SA itself is a significant upgrade from the old standardised approach, built around a Sensitivity-Based Approach (SBA).

The SBA requires firms to calculate sensitivities for each position to a set of prescribed risk factors across seven risk classes: general interest rate risk (GIRR), credit spread risk (CSR), equity risk, commodity risk, FX risk, and residual risk. For each risk class, the sensitivity to each risk factor (delta, vega, and curvature) is calculated, and the sensitivities are aggregated using prescribed correlation matrices to produce a capital charge. The correlation matrices incorporate assumptions about how risk factors move together — imperfect correlation reduces the capital charge relative to a simple sum of absolute sensitivities.

The SBA is more mechanically complex than the old standardised approach but more risk-sensitive: it produces a lower capital charge for well-hedged portfolios and a higher charge for concentrated, unhedged exposures. It also produces more comparable outputs across firms, because the calculation methodology is prescribed rather than model-dependent.

The Internal Models Approach (IMA)

Banks with sophisticated risk management capabilities can apply to use internal models for some or all of their trading desks. Under FRTB's IMA, the capital charge is based on Expected Shortfall (ES) rather than VaR. ES — sometimes called Conditional VaR or CVaR — measures the expected loss in the tail of the distribution beyond the confidence level, rather than just the loss at the confidence level. Using ES at 97.5% produces a more conservative capital charge than 99% VaR for most portfolios, and ES does not exhibit the procyclicality problems of VaR.

The ES is calculated using a stress period — a period of significant financial stress relevant to the bank's portfolio. This follows the Basel 2.5 logic of SVaR, but is more rigorously specified under FRTB.

Desk-Level Approval

Under FRTB, IMA approval is granted at the trading desk level, not at the firm level. Each desk must qualify separately. This is a major change from the old framework, where a firm-level IMA approval covered all trading desks. Under FRTB, a desk that fails to qualify for IMA falls back to the SA — and the capital requirement for an IMA-capable firm is the sum of the IMA charges for approved desks and the SA charges for non-approved desks.

P&L Attribution Test

The P&L Attribution (PLA) test is one of the most technically demanding features of FRTB. For each desk using the IMA, the bank must demonstrate that the risk model explains the actual daily P&L of the desk. The test compares two series: the hypothetical P&L (HTPL — the P&L calculated from the risk model, using the actual market moves of the previous day applied to the previous day's position) and the actual P&L (RTPL — the actual mark-to-market P&L of the desk). If the hypothetical and actual series are too different — measured by statistical tests on their mean, variance, and correlation — the desk fails the PLA test and must fall back to the SA.

The PLA test creates strong incentives for desks to align their risk models with their valuation models. Historically, these two sets of models could diverge: the risk model might use simplified assumptions for risk measurement purposes while the P&L was calculated from a more sophisticated valuation model. FRTB forces convergence — the risk model must capture the same risk factors as the valuation model, or it will fail the PLA test.

Back-Testing

Alongside the PLA test, desks must pass a back-testing requirement: the actual daily P&L must not exceed the 99% VaR (calculated alongside ES) more than a prescribed number of times in a 250-day rolling window. Exceeding the limit (known as "exceptions") results in an add-on to the capital charge and, if persistent, can result in the desk's IMA approval being revoked.

Non-Modellable Risk Factors (NMRFs)

A distinctive feature of FRTB is the treatment of Non-Modellable Risk Factors. A risk factor is modellable under FRTB only if the bank can demonstrate that there are sufficient real price observations for that factor — at least 24 observations per year, with no gaps of more than one month. Many risk factors in complex or illiquid markets fail this test: the credit spread of a bespoke structured product, the volatility of an exotic equity option at long tenors, or certain commodity forward curves in thin markets.

For NMRFs, the bank cannot use the ES model. Instead, it must calculate a Stress Scenario Risk Measure (SSRM) — a capital charge derived from a stress scenario calibrated to the specific NMRF. The NMRF charge is additive: it sits on top of the ES charge for modellable factors. For banks with significant exposure to illiquid or bespoke instruments, the NMRF charge can be a substantial fraction of total IMA capital.

The NMRF framework creates strong incentives to standardise products and use instruments with liquid, observable markets. It also incentivises banks to contribute to price transparency initiatives that generate the real price observations needed to classify risk factors as modellable.