Regulators’ FRTB estimates based on faulty premise—industry study

US regulators may be underestimating the capital impact of proposed trading book rules due to take effect in 18 months, according to a new industry study.

Two industry groups—the International Swaps and Derivatives Association and the Securities Industry and Financial Markets Association—collected market risk data from the eight US global systemically important banks (G-Sibs) as part of their response to the Basel III endgame proposals issued by US prudential regulators in July 2023.  

The study found market risk capital would increase by 73%—close to the regulators’ estimate of 75%—if banks that currently use the internal models approach (IMA) continue to do so once the Basel Committee on Banking Supervision’s Fundamental Review of the Trading book (FRTB) is fully implemented in the US.

We know the number of banks which will actually go for internal model approval, certainly at the initial date, is very low on a global basis

Mark Gheerbrant, Isda

“That estimate assumes the banks are able to maintain the current internal models approvals,” says Panayiotis Dionysopoulos, head of capital at Isda. “Depending on how appealing regulators make internal models, that estimate could start going up.”

The regulators’ assumption about internal model usage looks increasingly dubious, however. A growing number of banks, including some G-Sibs, have already indicated that they will not seek approval to use the IMA under FRTB when the rules first enter into force. 

“We know the number of banks which will actually go for internal model approval, certainly at the initial date, is very low on a global basis,” says Mark Gheerbrant, Isda’s head of risk and capital.

Isda and Sifma therefore also looked at the capital impact of FRTB if all market risk is calculated using only the regulator-set standardized approach. In such a scenario, they found that market risk capital requirements for the eight US G-Sibs would rise by 112%—significantly more than regulators’ current estimate. 

Changing the rules

In a letter to US regulators, the trade groups argue that an over-reliance on the standardized approach could pose a risk to the financial system. 

“However good you make the standardized approach, it’s still one model that everyone’s going to be using, so any flaws will apply to every bank,” says Gheerbrant, “and the danger of such huge reliance on that, we think, is significant.”

To address this risk, Isda and Sifma suggest a series of modifications to the proposed rules. The first of these relates to capital add-ons for non-modellable risk factors (NMRFs)—risks for which banks are deemed to have insufficient data to produce reliable model outputs. The Isda-Sifma paper says NMRFs are one of the most significant disincentives for the adoption of the IMA.

The industry bodies suggest distinguishing NMRFs where there is sufficient data to run an expected shortfall model from those where there is not. The first category could still be modeled using expected shortfall—the measure at the heart of the IMA—but with an assumption that positions would take longer to liquidate than an equivalent modellable risk factor, while the second category would receive a capital add-on. 

Other suggested changes include capping IMA outputs at the level of the standardized approach—recognizing that the regulator-set methodology is meant to be very conservative—and relaxing the rules around a key regulatory test of model effectiveness. Trading desks that fail the profit-and-loss attribution (PLA) test face higher capital requirements and could be forced onto the standardized approach under the current Basel rules. 

Banks have complained about the counterintuitive nature of the test, which is harder to pass with well-hedged portfolios or during less volatile market conditions. The Isda-Sifma letter argues the PLA test should be used only for supervisory monitoring to avoid volatile capital requirements that could also disincentivize IMA adoption. 

“[The PLA test] is another big novelty where the ask from the industry working group is to make this a qualitative supervisory tool, to allow time both for the banks and supervisors to collect data before this is an automatic requirement that will introduce uncertainty and instability in the way the banks capitalize across the different trading desks,” says Dionysopoulos.

The paper also suggests replacing the proposed July 2025 go-live date for the US version of FRTB with a deadline of 18 months after the rule is finalized, which would allow more time for approval applications for the IMA.

Regulatory fragmentation

The Isda-Sifma paper also suggests changes in the calibration of the US version of FRTB, to allow more recognition of diversification in both the IMA and standardized approaches—and between the two approaches across different desks. In particular, the industry bodies say introducing an inter-risk-class correlation parameter of 0.5 would reduce standardized capital charges by 14%.

If adopted in full, the trade groups estimate their suggested changes to FRTB would limit the capital increase to around 24% if the eight G-Sibs maintain their current mix of IMA and standardized approaches, or 44% assuming only standardized approaches are used.

The changes suggested by the trade groups deviate significantly from the Basel version of FRTB and US regulators may be wary of adopting them. However, US bankers argue the US implementation of Basel III is more severe than the European version of the rules, and that softening the calibration is therefore justified. 

The US Basel endgame proposal scraps the use of internal models for credit risk and includes the stress capital buffer (SCB) as part of the overall capital stack. The Federal Reserve’s annual comprehensive capital analysis and review stress test, which is used to set the SCB, includes a global market shock scenario. Banks believe the SCB partly duplicates FRTB internal models because expected shortfall is also calculated using a stress event.

“It’s about the calibration of a particular standard. You have different parameters, and you find that the different jurisdictions make different decisions on how to go about it, and that makes it much more complicated to compare,” says Dionysopoulos. “Even with the three main jurisdictions—the US, the European Union and the UK—you get a very different profile of the overall estimates, even if you only focus on the data the regulators release themselves.”