Stress testing, cyber risk modelling and machine vision

BOE STRESS TEST of fund sector welcomed

IMAGE RECOGNITION is the latest tool for quants

CYBER RISK modelling isn’t making the grade

COMMENTARY: The macro view

The relatively recent concept of emergence – a belief that complex behaviour must be the result of a guiding mind – is why we refer to ‘queen’ ants, bees and termites; it seems impossible that the pinhead brains of thousands of mute insects could be collectively responsible for the architecture of a hive. But it’s now a core tool for understanding the behaviour of financial markets, in particular some of their otherwise inexplicable crashes.

So it’s good news that the Bank of England, leading the field, has carried out its first system-wide stress simulation of a crisis in the investment fund sector. It made some worrying discoveries about potential interaction between funds, banks, dealers and investors, which could lead to sharp jumps in corporate bond rates. There is some criticism of the assumptions on which the study was based, but all in all it’s a good first step.

Meanwhile, a Deutsche Bank analysis of pricing at liquidity aggregators used game theory to predict that aggregators with too many participants could end up harming overall liquidity, as the winning liquidity providers would be more likely to underprice trades – an effect known as the ‘winner’s curse’ – and may then decide to withdraw liquidity as a result.

US banks are even starting to use network analysis to assess the risks of their huge (and sometimes literally uncounted) inventories of internal risk and capital models.

All this is cheering stuff, but there’s still a need for caution – models of financial networks carry model risk of their own, whether because networks evolve rapidly and render them obsolete, or because omitting one of many layers of interaction can lead to serious misrepresentation of the true risks and connections involved. So risk managers at both banks and regulators should resist the temptation to rely entirely on their shiny new connection diagrams, as these could be almost as deceptive as the bank-level measures they replace.

STAT OF THE WEEK

Investment funds are expected to spend $7 billion a year by 2020 on so-called alternative data, such as satellite imagery and social media information, according to consultancy Opimas. 

QUOTE OF THE WEEK

“There are ways for firms to manage the [IFRS 9] standard. This is very controversial, because accounting standards should reflect reality rather than drive reality. However, within the standard there is a huge scope for judgement, interpretation, management – and manipulation” – Adrian Docherty, BNP Paribas