By Lars Frisell
This article first appeared on The Business Post on March 17, 2019.
A decade after Ireland’s devastating financial crisis, the state of financial supervision is arguably much better. The flawed philosophy of “principles-based” supervision, whereby financial firms were largely left to their own devices as long as profits were high and credit losses low, has been replaced by “intense supervision”. Rigorous processes, independent controls and sound risk culture are now key concerns of the Regulator.
The resources devoted to supervision have increased dramatically. The Central Bank of Ireland’s staff count has almost doubled since before the crisis, to about 1900 employees today, and the vast majority of this increase is in supervisors. Most profoundly, it is now the ECB and not the CBI that is responsible for supervision of Ireland’s largest banks, adhering to a single rulebook that applies to all banks in the eurozone.
Yet despite this momentous ramp-up, and the associated costs for the industry and taxpayers, supervisors may soon find themselves on the back foot again. Developments in fintech – just a sound bite a few years ago – are shifting the nature of financial services in a way that defies regulatory frameworks and traditional, entity-based supervision.
Payments were once the preserve of banks and credit card companies. Now that fintechs and online retailers are entering the business, payments are gradually but inevitably shifting outside the regulatory perimeter. It is not difficult to imagine that some of the tech giants or social media platforms will eventually bypass traditional players and provide financial services entirely on their own. Which activities will be regulated, and by whom, is not a straightforward question.
Second, operational risk is quickly moving outside the supervised entities into firms to which they outsource automated functions such as data storage, reporting and transaction processing. Several global financial services providers have suffered serious data breaches and IT disruptions in recent years. Outsourcing does not itself reduce the likelihood of such incidents - sometimes the opposite - yet it removes these functions from the direct purview of supervisors.
Thirdly, and perhaps most challenging, is the pace with which artificial intelligence is removing humans from every part of the financial value chain – marketing, risk analysis, pricing, and decision making. The amount of data available to inform credit decisions has grown exponentially with cloud storage, which gives machines an undisputable advantage over people.
Neural networks can self-learn in a matter of days how to make loan decisions by training on gigantic data sets. Certainly, when something goes wrong, the bank or fund will be liable, not the computer
But how will supervisors be able to understand, much less control, the processes that will lead to the next financial crisis?
As business models continue to transform, the likelihood of supervisory blind spots occurring increases. Supervisors are, by nature, outward-looking. But they will inevitably need to be become more “outward-working”, engaging with industry participants to make sure they understand the nature and location of risks in the system. Several regulators have already started to work in this way, inviting firms to test their new technologies in “sandboxes” or “labs”. Going forward, supervisors will need to make use of the same advances in information technology that are transforming the financial industry – massive data storage, automation of laborious processes, and artificial intelligence.
The challenge for supervisors should not be underestimated: this transformation cannot be achieved with current staff resources. To be sure, lawyers, economists, and risk analysts will still be needed but authorities need to attract, train and retain cadres of data scientists and programmers. Most importantly, it will require a cultural paradigm shift – alien to many traditional supervisors – to stay on top of a dynamic, high-tech industry with muddled regulatory boundaries.
