Each financial crisis brings more financial supervision, more models and larger buffers – but still fragility persists. The paradox of perfect supervision is that the very attempt to safeguard stability can increase systemic risk by increasing complexity and synchronising behaviour. This column argues that resilience, not ever-tighter risk-informed control, should be the organising principle of supervision, rooted in the basic idea that has long governed finance itself: diversification.
Director, Systemic Risk Centre London School Of Economics And Political Science
Supervision is the process by which public authorities oversee financial institutions to ensure that they operate prudently and comply with regulations. It involves monitoring balance sheets and governance, curtailing undesirable risk-taking, keeping risk within acceptable limits, and using quantitative models to inform capital and liquidity buffers.
Officially, in both Europe and the US, the aim of supervision is to ensure the safety and soundness of institutions, and the stability of the system as a whole. It is a demanding role, performed under public scrutiny and political pressure, with limited information and substantial uncertainty.
Supervision would likely work as intended if risks were known, measurable and stable, and if the system did not react to the act of supervision. None of this holds true in practice.
Supervision focuses on risks that can be quantified, perhaps by using daily market, credit and liquidity measures. These measures largely capture routine fluctuations, not the events that truly matter: the probability of institutional failure, the likelihood of systemic crises or the build-up of instability in opaque corners of the system. Quantifiable risk tends to the least important risk, and when we rely too much on such quantifiable risk in supervision, it can can lead to complacency and the very outcomes supervision seeks to avoid.
When supervisors do their jobs well, it is almost axiomatic that dangerous risks emerge where they are not looking. By drawing attention to what is easily measured, supervision can create a false sense of control. The unmeasurable becomes unseen and hence ignored. When attempting to quantify the unmeasurable, we mistake precision for understanding – a classic McNamara fallacy.
At the heart of modern supervision lies a fallacy of composition: if every institution is made prudent, the system will be safe. It is not so.
When a large shock occurs – perhaps a geopolitical crisis, a pandemic, or market panic – prudent institutions must reduce risks because regulations tell them to do so. That means selling into falling markets. These collective attempts at prudence become crisis amplifiers.
The problem is structural. Most banks now use Basel III standard models, identifying the same risks in the same assets. When stress arrives, they run for the same exit because this is what the regulatory framework requires.
Actions that appear prudent in isolation become dangerous in aggregate. In the words of Milton Friedman (1953), “[t]he conditions for the optimum of the individual are not the conditions for the optimum of the group.”
The push for ever more detailed rules not only standardises behaviour but also changes who carries the burden of judgement, as risk management is outsourced to the authorities. Banks disclose vast amounts of information to their supervisors, whereas the supervisors can only examine a small fraction of that information in real time.
That creates the illusion that supervisors could and should have identified the most serious risks ex ante. In reality, the almost infinite flow of information makes it impossible to know which signals matter at the time. Yet once a crisis has revealed where the danger lay, that same information can always be read backwards as if it had been obvious – the benefit of hindsight. The disclosure itself becomes grounds for blame, and responsibility is shared.
The prevailing supervisory logic is to identify and contain risks. Whereas this can make institutions safer by first approximation, it can also be counterproductive because it treats risk as exogenous.
Much financial risk is endogenous in the language of Danielsson and Shin (2003) and Danielsson et al. (2009). It is the interaction of market participants that creates endogenous risk in the system. By contrast, exogenous risk arrives from outside the financial system: we are affected by it, but our actions do not cause or change it. Think of the asteroid that wiped out the dinosaurs 65 million years ago or COVID-19 in 2020.
The models that generally inform supervision assume that risk is exogenous. But in reality, supervision that tightly defines and targets specific risks causes institutions to cluster around the same models, metrics and responses. The system becomes synchronised: vulnerabilities emerge precisely because institutions behave alike, not because of anything intrinsic to the assets themselves. The result is endogenous risk.
Supervision thus reshapes the very risks it seeks to control, often for the worse. Each new rule alters incentives and encourages adaptation. The private sector adjusts rapidly, now aided by AI, exploiting gaps in supervisory coverage. Supervisors, by contrast, must negotiate consensus across jurisdictions – a slow, political process. The need for common standards pushes supervision towards the lowest common denominator: those risks easiest to define and measure internationally, but not necessarily the most dangerous.
This ratcheting effect increases costs without improving stability since is makes the system more uniform, more constrained and, paradoxically, more fragile.
The aim of supervision should be a system that remains stable even when individual parts fail. That requires shifting from risk-informed buffers to resilience. The key to resilience is variety in institutions, business models and responses to stress, as discussed in Goodhart and Wagner (2012), Vos and Beetsma (2016), and Danielsson (2024).
Diversification is one of the most successful principles in finance. It works for portfolios, and should work for the financial system.
Policymakers already have the tools to promote diversity. Licensing and regulatory frameworks can encourage institutions with distinct business models, tailored to different markets and risk profiles, rather than pressing all firms into a single mould. Proportional regulation can allow smaller or specialised institutions to operate without being forced into homogeneity.
Resilience also depends on judgement. Quantitative tools are essential, but they should inform rather than replace supervisory discretion. Experienced supervisors who can challenge models, interpret behaviour and recognise cultural shifts contribute more to stability than any uniform metric.
Supervision defines its mandate as ensuring safety, soundness, and stability. These are essential, but they are means rather than ends. The deeper purpose of financial regulation is to ensure that finance supports prosperity while robustly resisting failure.
Supervision cannot eliminate crises. The system is too complex, too subject to endogenous amplification, and too uncertain for that. The goal should be to prevent shocks from cascading.
The paradox of perfect supervision is that the more it seeks uniform control, the less stable the system becomes. Strength lies not in ever-greater measurement and control, nor in increasing liquidity and capital buffer, but rather in institutional variety, independent judgement, and acceptance that uncertainty cannot be regulated away.
A resilient system is one that withstands what we cannot foresee.
Danielsson, J (2024), “Why so many crises happen when we know why they happen and how to prevent them”, VoxEU.org, 30 May.
Danielsson, J and H S Shin (2003), “Endogenous risk”, in Modern Risk Management: A History, Risk Books.
Danielsson, J, H S Shin and J-P Zigrand (2009), “Modelling financial turmoil through endogenous risk”, VoxEU.org, 11 March.
Friedman, M (1953), “The methodology of positive economics”, in Essays in Positive Economics, University of Chicago Press.
Goodhart, C and W Wagner (2012), “Regulators should encourage more diversity in the financial system”, VoxEU.org, 12 April.
Vos, S and R Beetsma (2016), “Stabilisers or amplifiers: Pension funds as a source of systemic risk”, VoxEU.org, 23 February.
Source: cepr.org
While artificial intelligence offers considerable benefits to society, including accelerating scientific progress, improving economic growth,…
The global minimum tax represents the most ambitious international effort in decades to curb profit…
While globalisation has increased productivity and incomes around the world, extensively linked value chains have…
Textbook monetary economics views inflation as fundamentally driven by aggregate shocks, such as money supply…
Artificial intelligence differs from other technological advancements in finance, such as the initial adoption of…
Industrial raw materials such as nickel, cobalt, and rare earths are critical inputs in countless…