ECB Research Dismantles Haircut Orthodoxy: Brokers Evaluating Licensing Pathways Face Collateral Model Lock-In
The European Central Bank published Working Paper 3301 on 15 October 2026, titled "Stop believing in haircuts, at least the way you think about them." The provocative title masks a sober empirical finding with direct operational implications for anyone building collateral infrastructure today.
Using transaction-level supervisory data from January 2019 to June 2024 and exploiting the 2021 introduction of the binding leverage ratio requirement, the researchers show that three frictions jointly shape haircuts in the bilateral dealer-to-client segment: dealer balance sheet constraints, collateral convenience yields, and client outside options. This is not an incremental refinement of existing theory. It is a direct challenge to the foundational assumption that haircuts function primarily as risk-management tools calibrated to asset volatility.
The standard model assumes haircuts should be positive when a dealer lends cash, protecting against selling collateral at a loss if the client defaults, and negative when a dealer lends a security. In practice, haircuts in bilateral repo are mostly zero and often carry the opposite sign. The ECB data confirms what practitioners have observed for years: the textbook model does not describe how bilateral markets actually function.
Negative haircuts mean that market participants endowed with high-quality collateral can turn to the repo market to obtain more funding than the market value of the pledged security. These negative haircuts are particularly prevalent in euro-denominated trades, where they represent 9% of outstanding volumes compared to 4% in dollar-denominated trades. This is not an anomaly, it is a structural feature of modern repo markets, driven by collateral scarcity and dealer economics.
The mechanics matter. The dealer's interest in collateral includes the rehypothecation value: as legal owner of the collateral, the dealer can re-use it during the term of the trade, whether by pledging it in another repo, delivering it against a short, or posting it as regulatory margin. Haircuts are not set in isolation, they emerge from a negotiation where both parties have complex, sometimes opposing incentives shaped by regulatory constraints and market structure.
The ECB model delivers sharp segment-specific predictions: leverage constraints raise haircuts in collateral-driven repos only, convenience yields flip sign across the two segments through a rehypothecation channel, and client outside options do the same through the shadow value of the conserved resource. Within this framework, zero and wrong-sign haircuts are equilibrium contracting outcomes under regulatory and intermediation frictions, not departures from rational risk management.
The timing is significant. The 3% leverage ratio requirement became binding on 28 June 2021, and the ECB researchers use this regulatory shift as a natural experiment to identify how balance sheet constraints alter haircut setting. Their finding: the effect depends entirely on the type of trade. In collateral-driven repos, binding leverage constraints raise haircuts. In funding-driven trades, the leverage constraint has no effect, but higher convenience yields raise haircuts through the rehypothecation channel, and broader client outside options lower them.
This has direct relevance to the regulatory trajectory facing digital asset brokers. Regulators have long contemplated using requirements for haircuts as a mechanism to curb leverage in repos. In 2014, the Financial Stability Board recommended introducing a framework for establishing numerical haircut floors, aimed at the non-bank financial sector. This framework targets non-centrally cleared securities financing transactions involving collateral other than government securities, where cash is provided to non-bank financial intermediation entities. The framework mandates minimum haircut levels determined by the type and maturity of the underlying collateral.
The FSB framework assumes haircuts are a lever that regulators can pull to constrain leverage. The ECB research suggests the relationship is more complex: haircuts are already responding to leverage constraints through dealer behaviour, and imposing floors may simply shift activity to other margins or segments. Requirements on dealer banks can be effective where SFT markets are largely intermediated by bank-affiliated dealers, while market-wide regulations could be more effective where intermediation by nonbanks is more prominent. Minimum haircuts or initial margin requirements are intended to act as a floor, and so market participants should still set haircuts according to their internal risk management policies.
For a broker founder evaluating licensing pathways, the operational question is concrete: the infrastructure you select today will encode a collateral model. Most white-label solutions and institutional custody platforms implement haircut schedules based on static risk assumptions, asset class, credit rating, maturity. These are inputs to a formula, not outputs of a negotiation. The ECB research suggests this approach may be directionally correct for centrally cleared transactions but fundamentally misaligned with how bilateral markets work.
Since the transitional period ended across the EU on 1 July 2026, any firm providing crypto-asset services must hold CASP authorisation. A MiCA licence is an authorisation granted by the national competent authority of an EU member state under Article 63 of Regulation (EU) 2023/1114, letting the holder serve clients in every member state from a single home base. The licensing clock is running. But the infrastructure decisions made during the transition period will persist long after authorisation is granted.
The risk is not that regulators will impose haircut requirements, they almost certainly will, eventually. The risk is that you build on infrastructure designed for a static, volatility-based haircut model, then find yourself unable to adapt when regulatory frameworks evolve to reflect the ECB's findings. The FSB has already delayed implementation of its minimum haircut standards multiple times, in part because the original framework did not account for how markets actually function.
A significant portion of bilateral contracts have zero haircut, 35% of all transactions in one study of UK markets. If your infrastructure cannot represent a zero haircut, or cannot adjust haircuts based on counterparty relationships and market conditions, you are building on a foundation that does not match market reality.
The ECB paper does not prescribe what brokers should do. It describes what markets do. The implication is that collateral models will need to accommodate dealer constraints, convenience yields, and counterparty bargaining power, not just asset volatility. Infrastructure that cannot support this flexibility will require replacement, not reconfiguration.
The licensing transition is a constraint on operations, but it is also a design window. The collateral architecture you select now will shape your risk management capabilities for years. The ECB's research suggests that the assumptions embedded in many off-the-shelf solutions may already be outdated. Whether that matters depends on how quickly regulators and counterparties update their own models, and whether you have the flexibility to respond when they do.
References
[2] ECB Banking Supervision, "Leverage ratio Pillar 2 requirement,"
[3] FSB, "Leverage in Nonbank Financial Intermediation: Final Report," July 2025
[4] Regulation (EU) 2023/1114 on Markets in Crypto-Assets (MiCA)
