Libor replacement, model risk and stress testing

FRTB proxy risk factors may trigger model failures

FCA’s Libor plans a ‘reality check’ for loans, bonds, RMBS

CCP stress testing gets real

COMMENTARY: Exact vs accurate

Once again concern is growing over the limitations of risk modelling – and the profession is caught in a vice. On one side, new regulations such as the profit and loss attribution test within the Basel Committee’s Fundamental Review of the Trading Book (FRTB) encourage the application of models to as many risks as possible – with the threat of a forced move to a potentially more costly and less sensitive standardised alternative to spur compliance. On the other side is the danger of developing a model that produces an expected loss figure or capital requirement that is exact but wrong.

FRTB also mandates large capital add-ons for non-modellable risk factors where data points are too sparse for a reliable model. One way round this, which many institutions are exploring, is to use proxy risk factors, which have a richer loss history and are presumed to correlate well with the non-modellable risk. If the correlation fails, however, the bank will face a costly move to the standardised model. And, more dangerously, if the correlation of proxy and risk holds most of the time and fails only in extreme circumstances, the bank will face huge and unexpected losses.

Non-modellable risks are among the most serious that banks now face. The risk of cyber attack, for example, is particularly intractable thanks to a rapidly changing threat environment and a paucity of data, causing growing concerns about the reliability of cyber risk insurance and the stability of the insurance providers who offer it. There are other reasons to doubt their ability to properly assess emerging risks like cyber risk: many insurers seem blind to entire clusters of emerging dangers.

Model risk can even be created by procedures aimed at reducing it; much academic research has focused on the problem of devising plausible and credible stress test scenarios, in particular for use by central counterparties (CCPs), but the problem here, critics say, is that standardisation of stress scenarios across a number of CCPs could be catastrophic if the set of scenarios is incomplete – and the neglected risk is able to hit every CCP at once.

STAT OF THE WEEK

Market participants estimate that around $24 trillion notional in cleared euro interest rate swaps may have to be transferred from LCH in London to an EU clearing house once the UK leaves the European Union…

QUOTE OF THE WEEK

…and the transfer is “going to be the most godawful bloody mess” – Simon Gleeson, Clifford Chance