Case Study 14.1: The P&L Attribution Failure — Priya's FRTB Readiness Audit at Halcyon Securities

The Situation

Organization: Halcyon Securities (fictional EU-regulated investment firm) Priya's engagement: FRTB readiness assessment for Halcyon's four trading desks Timeline: Q1–Q2 2024 (pre-FRTB implementation) Regulatory backdrop: CRR3 bringing FRTB into EU law effective January 1, 2025; Halcyon's regulator (BaFin) expecting FRTB readiness plans by Q3 2024

Background: Halcyon's Trading Operations

Halcyon Securities operates primarily in European fixed income markets. Its four trading desks:

1. EGB Desk: European Government Bonds — German Bunds, French OATs, Italian BTPs
2. IG Credit Desk: Investment grade European corporate bonds
3. Rates Options Desk: Swaptions, caps/floors, options on Bunds
4. EM Rates Desk: Emerging market local currency bonds — Poland, Hungary, Czech Republic, Romania

Under the pre-FRTB regime, all four desks operated under the Internal Models Approach (IMA) using a historical simulation VaR model approved by BaFin in 2018.

Priya's mandate: assess FRTB readiness for all four desks; identify which desks could qualify for the IMA under FRTB and which would need to move to the Standardized Approach (SA).

The P&L Attribution Assessment

FRTB requires each IMA-eligible desk to pass a "Profit & Loss Attribution" (PLA) test. The PLA test compares:

• Risk-Theoretical P&L (RTPL): The P&L the risk model attributes to risk factor movements, calculated daily
• Hypothetical P&L (HPL): The actual P&L calculated holding positions constant (removing intraday trading effects)

Two metrics determine PLA test outcome: - Spearman correlation between RTPL and HPL: must exceed 0.80 - Kolmogorov-Smirnov (KS) statistic comparing the distributions: must not show significant distributional divergence

Desks with Spearman correlation ≥ 0.80 and KS test passing → qualify for IMA (Green zone) Desks with Spearman correlation < 0.80 or KS test failing → must use SA (Red zone) A middle zone allows continued IMA use with capital add-on during a remediation period

Priya's Findings by Desk

EGB Desk: Green Zone — Passed

The EGB Desk passed all PLA tests. European government bonds are among the most liquid instruments in the world — their risk factors (yield curves for Germany, France, Italy, Spain) are well-defined, heavily traded, and available in high-frequency data.

Spearman correlation: 0.94. KS test: passed.

One finding: the Italian BTP yield curve was modeled using a single generic EU sovereign yield curve in the risk model, rather than a country-specific BTP curve. This approximation introduced small but consistent deviations. Priya recommended adding a BTP-specific spread factor to the model — the correction would not affect PLA test outcomes materially, but would improve risk sensitivity for the Italian sovereign book.

IG Credit Desk: Amber Zone — Remediation Required

The IG Credit Desk showed a Spearman correlation of 0.76 — below the 0.80 threshold. Investigation revealed the root cause: credit spread volatility.

The risk model represented credit spreads using sector-level indices (e.g., "European financials IG credit spread index"). But the actual portfolio held bonds from specific issuers — Deutsche Bank, BNP Paribas, Société Générale — whose spreads moved idiosyncratically relative to the sector index. On days when there was credit news about a specific issuer, the risk model predicted flat spreads (because the sector index was unchanged) while the actual HPL showed a significant move.

Remediation required: add issuer-specific credit spread risk factors for the 20 largest positions (covering 68% of portfolio notional). This required subscribing to issuer-level CDS data (Bloomberg) and modifying the risk engine configuration.

Timeline: 4 months. Cost: approximately €140,000 in data subscriptions and implementation.

Rates Options Desk: Red Zone — SA Required

The Rates Options Desk failed the PLA test decisively. Spearman correlation: 0.41.

The root cause was the most technically complex: the desk's option book generated P&L from: - Delta (linear rate sensitivity) — well-captured by the model - Vega (sensitivity to implied volatility) — partially captured - Volatility smile dynamics — not captured at all

The options desk ran a large book of swaptions across different maturities and strikes. The implied volatility surface — the relationship between option price, maturity, and strike — is not flat. When market conditions shifted, the volatility smile moved — generating substantial P&L from the change in the smile's shape. The risk model assumed a flat implied volatility surface, capturing only ATM (at-the-money) vega, not the smile dynamics.

The model's RTPL showed vega P&L from ATM volatility movements. The HPL showed the same ATM vega P&L plus smile dynamics P&L — a material unexplained residual on days of significant vol surface moves.

Remediation path: The quantitative team estimated that adding smile risk factors (SABR model or similar) would require 12–18 months of model development and validation — too late for January 2025 FRTB implementation.

Decision: Rates Options Desk moves to SA for FRTB implementation. Remediation program continues in parallel with target IMA re-application in 2026.

Capital impact: The SA capital charge for the Rates Options Desk was 2.3× the previous IMA capital — an increase of €28 million in market risk capital requirements for this desk alone.

EM Rates Desk: Mixed — Partial NMRF Issue

The EM Rates Desk passed the PLA test (Spearman: 0.83) but faced a Non-Modellable Risk Factor challenge.

The desk traded bonds in Poland, Hungary, Czech Republic, and Romania. For the major markets (PLN, HUF, CZK rates), sufficient market price observations existed — these were classified as modellable risk factors.

But the desk also held Romanian RON-denominated bonds (RON rates) and Polish inflation-linked bonds (PLN real rates). Both of these risk factors had insufficient market price observations on at least 1 day per month over the prior year — triggering NMRF classification.

The NMRF capital add-on for RON rates and PLN real rates was calculated using stress scenarios. The stress scenario loss was €8.5 million per year — significant for a relatively small book.

Priya recommended: reduce the RON and PLN-linked positions to minimize NMRF capital; alternatively, seek additional market price data sources to improve the observation count (potentially reclassifying as modellable).

The Capital Impact Summary

The Rates Options Desk failure alone accounted for 91% of the capital increase. The board risk committee received Priya's analysis and approved the SA transition with the parallel IMA remediation program.

Regulatory Engagement

Priya facilitated Halcyon's BaFin pre-implementation discussions. BaFin's key expectations:

1. Written plan for SA transition: Halcyon submitted a formal FRTB implementation plan documenting: current desk classifications, PLA test results, capital impact analysis, timeline for SA transition, and the parallel remediation program for the Rates Options Desk's IMA re-application.

2. SA methodology validation: Under FRTB, even the SA is more complex than its predecessor — it uses sensitivity-based methods (delta, vega, curvature). BaFin expected validation documentation for the SA sensitivity calculations.

3. Ongoing PLA monitoring: For desks in the IMA (EGB and EM Rates, after remediation), BaFin expected quarterly PLA test results to be reported as part of the regular supervisory data package.

BaFin's response to the implementation plan: "The submission demonstrates adequate understanding of FRTB requirements and a credible remediation timeline. The decision to use the SA for the Options Desk pending model enhancement is appropriate."

Discussion Questions

1. The Rates Options Desk's PLA test failure resulted from the risk model's inability to capture volatility smile dynamics. This is a model risk issue — the model is conceptually incomplete. Under SR 11-7 equivalent principles, who should have identified this model limitation? When should it have been identified — during the original model approval in 2018, or is it acceptable that FRTB's more rigorous PLA test surfaced it?

2. The capital impact of the Options Desk moving from IMA to SA was €28 million additional capital — a 2.3× increase. From a business perspective, is this capital cost acceptable? What options does Halcyon have to reduce its SA capital charge (without immediately qualifying for IMA)?

3. The IG Credit Desk's PLA failure was caused by using sector-level credit spread indices rather than issuer-specific risk factors. This is a risk factor granularity issue. How does the granularity-vs-simplicity tradeoff apply to market risk models: what are the risks of adding many issuer-specific factors, and what are the risks of insufficient granularity?

4. Romania's RON rates were classified as Non-Modellable Risk Factors due to insufficient market price observations. Is this a data availability problem (not enough observable prices) or a market liquidity problem (the risk is genuinely harder to model)? Does the distinction matter for how the capital charge should be set?

5. Priya's engagement produced an assessment that significantly increased Halcyon's required capital. From a consultant engagement perspective: how should Priya have communicated the capital impact finding to Halcyon's senior management? What is the consultant's responsibility when the assessment result is unfavorable to the client?