Regulators should be careful what they wish for on FRTB

New framework likely to reduce use of internal models, as planned; but is that a good thing?

A decade ago, regulators warned they were going to crack down on banks’ use of internal models for calculating risk in their trading books. In their own words, the intention of the Basel Committee on Banking Supervision back in 2012 was to “reduce the reliance” of the industry on the use of models, because the financial crisis had suggested some models underestimated risk.

Since then, the Basel Committee has devised a floor that prevents the capital requirements generated by internal models from falling too far below the outputs of regulator-set standardised approaches. They have also implemented tests that act as a guarded gateway into the internal model approaches (IMA) for market risk, which are supposed to allow only the proven best-in-class models to pass through.

Or at least, that was the theory.

In reality, banks in Europe that have begun to plan for implementation of the Fundamental Review of the Trading Book (FRTB) now expect to drastically cut the number of internal models they will use. One globally systemic bank plans to ditch the IMA altogether. Two others say they think every bank will switch at least some of its trading desks onto the standardised approach, with very few retaining models for more than half their trading desks.

IMA can lead to lower capital requirements, but banks have to balance that uncertain advantage against the certain increase in operational costs associated with internal model development, validation and approval.

And so, regulators have succeeded in reducing reliance on internal models, perhaps even further than they anticipated. But does that necessarily mean an improved way of calculating capital requirements? That all depends on the performance of the standardised approach.

Regulators have enhanced the standardised fallback for trading books by trying to align its risk sensitivity to that of internal models. This relies on making assumptions based on price movements seen during the 2008 financial crash.

However, the more time passes since that crisis, the greater the chance of discrepancies emerging between the estimates of risk projected by internal models and by the standardised approaches. The International Swaps and Derivatives Association argued that the setting of risk-weights for carbon emission certificates in the standardised approach overestimates the actual risk in a market that has matured significantly since 2008.

Success in the IMA model approval process has less to do with the quality of the models, and more to do with the trading strategy of the desk itself

The opposite is also possible: the standardised approach could eventually come to underestimate risks compared with banks’ internal models, if new events cause worse losses than 2008 for specific exposures.

Under the existing market risk capital regime, banks using internal models must measure stressed value-at-risk – estimating maximum losses at a given confidence level during a stress period. A survey conducted by Risk Quantum of regulatory reporting in the first quarter of 2022 found that nine out of 32 of the largest European and North American banks now base their stressed VAR observation periods on Covid-19, rather than 2008.

The new risk measure that will be used by IMA desks after the FRTB is implemented is the expected shortfall, which is an average of the worst-case losses for a portfolio beyond a VAR measure estimated at a 97.5% confidence level. As with stressed VAR, new risk events can enter the tail window if losses are so severe, opening up the possibility that the measure will become more conservative than the standardised approach.

Banks are also finding that success in the IMA model approval process has less to do with the quality of the models, and more to do with the trading strategy of the desk itself. Desks that don’t completely hedge risks such as delta – the sensitivity of a derivative price to changes in the underlying price – have more room for error than those that hedge thoroughly.

Paradoxically, this means the models for a proprietary trading desk that poses greater directional risk to the bank don’t need to be as accurate to receive regulatory approval as the models for a well-hedged desk that poses less directional risk.

The Basel Committee may get what it wanted on a superficial level: reduced use of internal models in the capital framework. But whether it will achieve the deeper aim of better market risk management remains to be seen.