The Stringency and Complexity of Bank Regulation: What Should Simplification Preserve?

8th October 2026 

Key takeaways 

Across 21 advanced economies, preliminary evidence indicates that tightening measures combining high stringency and complexity are followed by a sustained decline in systemic risk, whereas equally stringent but simple rules yield gains that fade as institutions adapt. Complexity is valuable when it closes avenues for arbitrage. Reform should remove duplication and low-value burden while preserving specificity that makes stringent safeguards effective.

After more than a decade of building up the post-crisis rulebook, advanced economies are now reopening it. As several contributions to this Forum have discussed, the question is no longer only whether banks should hold more or less capital. It is increasingly about the architecture of regulation: whether the many overlapping constraints assembled since 2008 are necessary to contain systemic risk, or whether some of them impose operational cost without a commensurate benefit. The United States has moved toward capital relief. The United Kingdom is transposing Basel 3.1 on a broadly capital-neutral basis while reviewing its ring-fencing regime. The European Union is raising binding requirements for its largest banks through CRR III and CRD VI while, in parallel, pursuing an agenda to simplify the framework that already exists. Simplification has become a shared objective across jurisdictions. The open question is what, precisely, to simplify: when regulators cut rules in the name of simplicity, are they removing harmless red tape, or quietly giving up something that keeps the system safe? 

Measuring stringency and complexity

In work with Cameron Ellis and Chotibhak Jotikasthira, we argue that answering that question requires separating two attributes of a regulatory action that the literature, and much of the policy debate, has collapsed into one. The first is stringency: how binding the action is for the regulated institution — the breadth of institutions affected, the size of the constraint, and how much it actually bites. The second is complexity: the operational and compliance burden the action imposes — new tool types to track, conditional thresholds and carve-outs, coordination across subsidiaries and jurisdictions, or layering on top of rules that are already in place. The two dimensions are conceptually orthogonal. A loan-to-value cap is highly stringent for the borrower but simple for the bank to implement once the ratio is fixed. A Basel III Pillar-2 add-on with institution-specific calibrations and multiple conditional thresholds can be equally stringent and far more complex. Conventional wisdom treats complexity as pure deadweight — operating cost with no offsetting benefit. We set out to test that assumption directly, because a single index of how much regulation a country has cannot tell stringency and complexity apart, and so cannot say whether complexity is a cost or a benefit. We operationalize regulatory complexity along three observable dimensions: the number of distinct tools and reporting items; the use of conditional thresholds, exemptions, and institution-specific calibrations; and the extent to which a measure interacts with the existing rule stack. The index therefore captures regulatory design as experienced by regulated institutions, not merely the number of pages or legal provisions.

Three examples from our sample make the point concrete. In January 2007, Italy applied a 35% risk weight to mortgages on owner-occupied residential property within an otherwise unchanged framework. The action was stringent for the affected exposures, but operationally light: banks had one new risk-weight bucket to track and report. Sixteen months later, Spain's Circular 3/2008 moved capital requirements on real-estate exposures in the same direction, but through the full transposition of Basel II: in a single rule, it introduced the standardised and internal-ratings-based approaches to credit risk, a multi-tier real-estate risk-weight matrix, and a supervisory add-on regime with bank-specific calibrations, all operated in parallel with Spain's pre-existing dynamic-provisioning regime. Same regulatory direction; very different operational footprint. Germany illustrates the divergence from the opposite side: it added Basel III requirements like everyone else, yet ended our sample period more stringent and, on our measure, less complex than it began, because reserve-requirement simplifications in the 1990s and the transition to the Eurosystem framework more than offset the added complexity.

To capture these differences systematically, we construct two indices — one for stringency, one for complexity — for 21 advanced economies from 1995 to 2025. We take every dated regulatory action recorded in the IMF's iMaPP database and the ECB's MaPPED database and score each one along both dimensions using pairwise comparisons run by a large language model: for every pair of actions of the same direction, the model judges, tournament-style, which one added more to the stringency, and which to the complexity, of the regime. The result is a pair of event-dated, reversible indices that weight actions by their intensity rather than merely recording that a tool is switched on, and that hold the two dimensions apart. This lets us sort regulatory events by their direction and by whether they were high or low on each dimension, and then trace what happens afterwards to systemic risk, to equity values, and to the volume of credit.

When complexity makes stringency durable

Our preliminary evidence shows that tightening actions that raised both stringency and complexity were followed by a sustained fall in systemic risk — on our estimates, roughly a one-third reduction in the banking system's expected capital shortfall, holding across the full two years we can observe. Tightenings that raised stringency but left the rulebook operationally simple produced a similar initial improvement, but it faded within about a year — consistent with financial systems adapting around single, easily gamed rules. These stability gains do not seem to have come at a visible aggregate cost: neither broad equity-market values nor the volume of credit to the private sector contracted after tightenings, including the most stringent and most complex ones. Complexity seems to carry a cost only where it was not paired with stringency. In the combination of high complexity with low stringency, bank stocks underperformed the broad market by roughly a fifth. Complexity that disciplines nothing is deadweight; complexity that backs a binding constraint is not.

While these findings are suggestive, regulatory changes are not a controlled experiment, and we do not treat these estimates as strictly causal. We construct counterfactuals using synthetic-control methods and macroeconomic controls, and we read the results as informative patterns whose mechanism should be corroborated by more tightly identified micro-level evidence, including our own work discussed below.

The natural reading of the pattern is regulatory arbitrage. When a stringent rule is operationally simple, a bank can often take on equivalent risk in an adjacent activity or vehicle while remaining technically compliant, and the initial reduction in systemic risk decays as it does so. When the rule is operationally demanding — consolidated requirements, cross-subsidiary thresholds, reporting that reaches across affiliates — that substitution is harder to execute, and the gain persists. Two cases illustrate the mechanism. Stricter bank capital rules have pushed leveraged lending toward private credit funds outside the prudential perimeter — vehicles that nonetheless remain connected to banks through warehouse facilities and credit-line backstops. And the leverage ratio — a deliberately simple rule, often criticised precisely for treating all assets alike — pushed corporate bond market-making from bank-affiliated dealers toward open-end mutual funds. In our related work on bond markets,¹ we show that those funds then became the marginal providers of liquidity in investment-grade bonds, leaving bond liquidity more exposed to outflow-driven selling — a fragility that materialised at the onset of the COVID-19 pandemic. In both cases, tightening did not eliminate risk; it relocated it. A single blunt rule may therefore not be enough to keep the risk outside the banking system.

Simplification, supervision, and ownership

Complex rules are specific provisions that close loopholes, such as a capital requirement with liquidity backstops attached, or consolidated thresholds that prevent risk from being parked in an affiliate. Separating useful specificity from genuine bureaucratic burden is, in our view, the natural next step for this research agenda. Simplification should assign each layer an explicit purpose and owner, remove duplication across layers, and retain provisions whose specificity prevents avoidance or disciplines judgement.

Rules are, of course, only half of the story; the other half is whether supervisors act on the information available to them. Supervisory complexity is related but distinct: it arises from the methodologies, models, data, escalation procedures, and metrics used to interpret and enforce the rules. Some of these elements are burdens in their own right; others are guardrails that make judgement consistent, contestable, and accountable. The recent failures are instructive. Had the full liquidity coverage requirements applied to Silicon Valley Bank, it would likely have failed the test well before its collapse, giving supervisors earlier warning; and the slow deposit run at Credit Suisse was visible in market signals long before the end. Whether market signals can be made operational for supervision is, however, a genuinely hard question. In earlier work on credit booms,² we show that aggregate credit growth alone is a weak early-warning signal, since expanding credit is also a feature of a healthy, growing economy. What distinguishes booms that end in banking crises is who is lending: when new entrants and foreign banks — lenders without an established local franchise — account for a disproportionate share of credit growth, the probability of a subsequent crisis rises sharply. Yet even armed with that indicator, a supervisor would find it difficult to act on it in real time, since restricting credit from new entrants specifically would cut directly against competition objectives. Better indicators do not remove the need for supervisory judgement and for the legal mandate to exercise it.  

The debate about the post-crisis rulebook is often framed as a choice between more regulation and less. Our evidence suggests that the more useful questions are how stringent and how complex — and, within complexity, which parts are specificity that closes gaps and which are burden that closes nothing. The answer determines whether simplification makes the financial system cheaper to regulate, or merely easier to arbitrage. Importantly, the same language-model technology that made our measurement possible is likely to lower the cost of both compliance and enforcement, for regulated firms and supervisors alike, which may itself shift the cost-benefit calculus of complexity in future regulatory design.

Endnotes

1. Giannetti, Mariassunta and Jotikasthira, Chotibhak and Rapp, Andreas C. and Waibel, Martin, Intermediary Balance Sheet Constraints, Bond Mutual Funds' Strategies, and Bond Returns  (June 06, 2024). Swedish House of Finance Research Paper No. 23-10, SMU Cox School of Business Research Paper No. 23-19, Journal of Finance forthcoming, Available at SSRN: https://ssrn.com/abstract=4607843
2. Giannetti, M. and Jang, Y., 2025. Who lends before banking crises? evidence from the international syndicated loan market. Management Science, 71(3), pp.2289-2310.
   

To comment, please go to the discussion forum