FRTB, machine learning and swap transparency

The week in, March 31–April 6, 2017

FRTB reboot planned over P&L storm

MACHINE LEARNING gives quants market insight

SWAP TRANSPARENCY causes unexpected problems


COMMENTARY: FRTB and model risk

The clock is ticking towards the planned 2019 implementation of Basel’s Fundamental review of the trading book (FRTB) and major problems still remain unsolved. Chief among them is the continuing dispute over the use of internal models.

It’s an argument that has already been raging for years in other areas of risk management – notably operational risk, where a long-running effort under Basel II rules to prod major banks towards using the model-based advanced measurement approach proved largely unsuccessful. Op risk regulators are now heading towards rejecting the use of internal models for regulatory capital calculation altogether, in favour of a simpler (but also contentious) standardised approach. Now the same question is proving equally insoluble in the case of market risk: in the FRTB’s profit and loss attribution test banks are facing a painful choice between an internal model approach that poses huge implementation problems, and a standardised approach that promises massive increases in market risk capital requirements.

Here and in other areas, it’s often model risk management at the smallest banks that faces the biggest test. The largest banks will receive more attention, due to their systemic importance and scale, but they tend to have the resources to comply; the smaller banks will find themselves stretched by comparison.

And it is not a problem that will go away; the arguments on each side are familiar to everyone. Internal models are more sensitive than blunt-instrument standardised approaches, and designed by people with a deeper understanding of each institution. But the results are also less easy for outsiders to use in comparisons, and will always carry the suspicion that the models have been rigged by their insider designers, who may well have strong motives to do so. External model oversight is at best an imperfect and highly costly remedy.

Ultimately, of course, it’s an issue of trust; the financial sector’s regulators simply can’t rely on the institutions they supervise to be honest with themselves about their own internal risks. Hence the costly oversight/blunt instrument dilemma they are left with – which is, in the end, partly of the banks’ own making.



Volumes of cleared swaps and forwards have risen sharply since September 2016, when the first phase of the non-cleared margin regime required big dealers to exchange initial margin. LCH ForexClear has seen a sharp increase in monthly volumes of non-deliverable forwards (NDFs) from around $60 billion a month prior to this date to $175 billion, $265 billion and $370 billion in the subsequent three months. Research estimates 35% of global dealer-to-dealer NDF volume is now cleared, and as clients are offered NDF clearing and as further margin deadlines pass, this percentage should increase further.



We have to leave behind this theory that the cost of capital is an impediment to lending. It’s not obvious at all that if you relieve capital you create more room for lending – quite the contrary. There have been so many studies showing robustness in capital levels is a support to the flow of credit. But what we most need to defend is the idea of a risk-sensitive capital framework, and for this, an accurate ranking of risks is crucial” – Isabelle Vaillant, EBA

  • LinkedIn  
  • Save this article
  • Print this page