The Bank of Canada published Staff Analytical Paper 2026-13 in April 2026, marking the most granular transaction-level study of a specific DeFi lending protocol ever produced by a G7 central bank. Researchers Jonathan Chiu and Furkan Danisman analyzed Aave V3 on Ethereum — the largest decentraliz...
"Every affected user will be fully reimbursed." — Omer Goldberg, CEO, Chaos Labs (on Aave's March 2026 oracle incident)
The Bank of Canada published Staff Analytical Paper 2026-13 in April 2026, marking the most granular transaction-level study of a specific DeFi lending protocol ever produced by a G7 central bank. Researchers Jonathan Chiu and Furkan Danisman analyzed Aave V3 on Ethereum — the largest decentralized lending protocol at $27.2 billion in total value locked — using on-chain data from January 2023 through May 2025. The core finding: DeFi lending is operationally viable with proper governance, but recursive leverage, revenue concentration, and wave-pattern liquidations create structural fragilities that current frameworks do not adequately address.
The paper arrives weeks after the European Central Bank published its own DeFi governance study showing that the top 100 token holders control over 80% of voting power across Aave, MakerDAO, Ampleforth, and Uniswap. Together, these two papers signal a shift in central bank posture — from theoretical risk assessments to empirical, protocol-specific audits. For DeFi protocols managing $94 billion in aggregate TVL, this transition from observation to measurement has direct regulatory implications.
Staff Analytical Paper 2026-13, titled "DeFi Lending: Returns, Leverage, and Liquidation Risk," is the Bank of Canada's second major DeFi study, following its 2023 theoretical paper "On the Fragility of DeFi Lending." The 2026 version is empirical rather than theoretical, using transaction-level data from Aave V3's Ethereum deployment.
Three primary findings emerged:
Zero bad debt on the protocol level. Aave V3 on Ethereum recorded no accumulated bad debt during the study period, including through the turbulent trading conditions of 2024. Over-collateralization requirements and automated liquidation mechanisms functioned as designed, triggering position closures before collateral values dropped below outstanding debt.
Recursive leverage is widespread. Despite the overcollateralization requirement, 20% of total borrowing volume and 8.2% of all transactions involved recursive leverage — the practice of borrowing against deposited collateral, re-depositing the borrowed asset, and borrowing again to amplify exposure.
Revenue concentrates in a narrow set of tokens. Protocol earnings derive disproportionately from a small number of asset pairs, raising questions about diversification and systemic dependency on specific markets.
The researchers classified DeFi's systemic impact on traditional finance as "currently limited" due to the sector's relative size compared to mainstream banking. However, they flagged intra-crypto contagion as a plausible transmission mechanism during stress events.
The 20% figure for recursive leverage volume is the paper's most policy-relevant data point. In traditional finance, leverage ratios are regulated, disclosed, and stress-tested. In DeFi, recursive leverage is permissionless, invisible to casual observers, and limited only by the protocol's loan-to-value (LTV) parameters.
Consider the mechanics: A user deposits 100 ETH as collateral, borrows 80 ETH worth of stablecoins (at 80% LTV), converts those stablecoins back to ETH, deposits the new ETH, and borrows again. Each cycle amplifies the user's exposure while remaining within per-transaction LTV limits. The aggregate leverage, however, can far exceed what the protocol's surface-level parameters suggest.
The Bank of Canada's data shows this is not an edge case. At 20% of borrowing volume, recursive leverage is a structural feature of how Aave V3 is used. The paper does not prescribe specific remedies but notes that this practice "amplifies risk" beyond what overcollateralization alone can mitigate.
For context, Aave has surpassed $1 trillion in cumulative lending volume as of February 2026. If recursive leverage consistently represents 20% of flow, approximately $200 billion in cumulative lending has involved amplified positions — a figure comparable to mid-sized national banking systems.
The paper's liquidation analysis found that four assets — WETH, wstETH, WBTC, and weETH — accounted for 90% of total liquidation value during the study period. Liquidations clustered in temporal waves during sharp price declines rather than distributing evenly over time.
Borrower losses during liquidation events ranged from 5–10% when accounting for liquidation fees alone. When including missed price recoveries — the appreciation borrowers forfeited by being liquidated at local lows — total losses reached 10–30% of the liquidated collateral value.
The finding of "limited broader market impact" from these liquidation waves requires qualification. During the study period (January 2023–May 2025), DeFi lending TVL ranged from approximately $15 billion to $45 billion. At current levels of $94 billion in total DeFi TVL, with Aave alone at $27.2 billion, the systemic relevance of liquidation cascades has increased.
The paper's data predates the March 10, 2026 incident in which an oracle misconfiguration on Aave triggered $27 million in erroneous wstETH liquidations — an event that provided a real-world stress test for the paper's theoretical framework.
Aave V3 generated $83.3 million in fees over the 30-day period ending in late March 2026 and recorded its highest-ever quarterly revenue of $22.56 million in Q4 2025. The protocol holds a 62.8% share of the decentralized lending market by TVL and a 28.9% share of total DeFi TVL.
The Bank of Canada's finding that revenue concentrates in a small number of tokens aligns with observable data. Borrowing demand skews heavily toward USDC, USDT, and ETH, with collateral concentrated in ETH derivatives (wstETH, weETH) and WBTC. This creates a dependency structure: protocol revenue and risk both hinge on the health and liquidity of a narrow asset set.
The revenue concentration finding has implications for Aave's transition to V4, which launched on March 30, 2026 with a hub-and-spoke architecture designed to pool liquidity across chains. If revenue remains concentrated in the same asset pairs regardless of multi-chain deployment, the architectural upgrade may distribute the same concentrated risk across more venues rather than genuinely diversifying it.
Three weeks before the Bank of Canada paper's publication, Aave experienced its most significant operational failure of 2026. On March 10, a misconfiguration in the CAPO (Collateral Asset Price Oracle) system caused the protocol to value wstETH at approximately 1.19 ETH when the actual market rate was 1.23 ETH — a 2.85% discrepancy.
The error triggered automatic liquidation of 34 high-leverage E-Mode positions totaling 10,938 wstETH ($27 million). Liquidation bots captured 499 ETH ($1.2 million) in bonuses by executing against positions that should not have been eligible for liquidation. The root cause: stale parameters in the CAPO smart contract where the reference exchange rate and timestamp were not synchronized, causing a maximum rate calculation below the live market rate.
Aave founder Stani Kulechov stated there was "no impact to the Aave Protocol" in terms of bad debt. Chaos Labs CEO Omer Goldberg confirmed all affected users would be "fully reimbursed," with 141 ETH recovered through BuilderNet refunds and 13 ETH in liquidation fees flowing to affected users, with DAO treasury funds covering the remaining shortfall up to 345 ETH.
The incident illustrates the Bank of Canada paper's central tension: Aave's automated liquidation system works as designed — it prevented bad debt — but it also liquidated users who should not have been liquidated. The system protected the protocol at the expense of individual borrowers, validating the paper's finding that the current design "shifts liquidation risk to borrowers."
In March 2026, the European Central Bank published a working paper by economists Alexandra Born, Zakaria Gati, Claudia Lambert, Mahvish Naeem, and Antonella Pellicani analyzing governance token concentration across Aave, MakerDAO, Ampleforth, and Uniswap using on-chain data.
The headline finding: the top 100 governance token holders control more than 80% of all holdings across these protocols. The ECB concluded that effective supervision requires "smarter identification of responsible parties rather than accepting surface-level claims of disintermediation."
Read together, the Bank of Canada and ECB papers construct a two-sided critique. The Canadian study demonstrates that DeFi lending works mechanically but concentrates risk in borrowers. The European study shows that the governance structures overseeing these protocols concentrate power in a small number of holders. The implication: a system that shifts risk to users is governed by a narrow group of token holders whose interests may not align with those users.
The ECB paper flagged these findings as relevant to MiCA enforcement, noting that governance concentration could provide regulatory "anchor points" for supervisory authorities seeking to identify accountable parties within ostensibly decentralized systems.
The Bank of Canada's study period captures a market that has grown substantially since. Current metrics:
| Metric | Value | |--------|-------| | Total DeFi TVL | $94 billion | | Aave TVL | $27.2 billion | | Aave market share (lending) | 62.8% | | Aave cumulative loans | $1 trillion+ | | Compound TVL | $2.8 billion | | MakerDAO (Sky) TVL | $7.1 billion | | EigenLayer TVL | $8.4 billion | | Aave 30-day fees | $83.3 million | | ETH deposited in DeFi | 25.3 million ETH |
DeFi TVL peaked at $120 billion in Q1 2026 before declining 21.7% to current levels amid broader market turbulence. The Fear & Greed Index stands at 8/100 (Extreme Fear), with ETH trading at approximately $2,133.
The institutional pipeline continues expanding independently of price action. Aave's Horizon product for institutional real-world asset markets reached $1 billion in TVL. Total RWA tokenization across the sector stands at $25.5 billion, a 21x increase from $1.2 billion in January 2023.
The Bank of Canada paper is categorized under JEL codes E50 and E58, placing it within the macroeconomics and central banking research framework rather than a technology or fintech classification. This taxonomic choice signals that the Bank views DeFi lending as falling within its monetary policy research mandate.
Three regulatory vectors emerge from the combined Bank of Canada and ECB research:
Leverage transparency. The 20% recursive leverage finding creates a basis for requiring protocols to disclose aggregate leverage metrics. No major DeFi protocol currently reports this data systematically.
Governance accountability. The ECB's 80% concentration finding provides a framework for identifying "responsible parties" under MiCA and potentially under the U.S. CLARITY Act, which remains in Senate deliberation.
Oracle risk management. The March 2026 Aave incident, combined with the Bank of Canada's liquidation wave analysis, suggests that oracle configuration is a systemic risk factor that existing governance structures have not adequately addressed.
The Bank of Canada paper does not make policy recommendations. It is classified as staff research, not as an official position of the institution. However, its empirical methodology — applying transaction-level analysis to a specific protocol — establishes a template that other central banks and regulators can replicate.
The Bank of Canada and ECB papers collectively represent a maturation in how central banks engage with DeFi. The era of generalized position papers on "crypto risks" is giving way to empirical, protocol-specific analysis using the same on-chain data that DeFi participants use themselves.
For Aave — and by extension, the DeFi lending sector managing tens of billions in user deposits — the implications are concrete. The system works mechanically: no bad debt, automated liquidations, transparent code. But "works mechanically" is not the same as "works for users." Recursive leverage amplifies risk beyond surface-level parameters. Liquidation waves impose 10–30% losses on borrowers. Oracle misconfigurations can trigger millions in erroneous liquidations. And the governance structures overseeing all of this concentrate control in a small number of token holders.
Central banks have now demonstrated they can read the chain. The question is what they do with what they find.