Article: DeFi-ying the Fed? Monetary policy transmission to stablecoin deposit rates
Date: 2026-09-14
Publisher: ECB Working Papers / Andrea Barbon, Jean Barthélemy, Benoit Nguyen
Score: ₿₿+
Read time: 25-30 min
Summary:
- Overview: The paper asks whether Federal Reserve policy reaches the yields stablecoin holders earn by lending USDC, USDT and DAI on Aave v2 and v3 (Ethereum) from January 2021 to January 2026. DeFi deposit rates averaged about 100 bps above the Fed funds target. However, they sat roughly 160 bps below it from mid-2022 to end-2023 and at times spiked above 1,000 bps. The authors build a model with two opposing forces. Liquidity providers with TradFi access arbitrage rates up after a hike, while falling crypto prices cut leverage demand and push rates down. A Bayesian proxy-VAR on high-frequency FOMC surprises, plus wallet-level data, tests both channels.
- Opposite Short-Run Response: A 100 bp shock to the 2-year Treasury yield lifts USDC deposit rates by roughly 100 bps on impact, because supply falls about 50% against a 40% drop in borrowing in the first week. Later deleveraging erodes that gain. For USDT, borrowing contracts faster than supply, rates initially fall, and convergence takes beyond 42 weeks. The weekly first-stage F-statistic is 7, below the strong-instrument threshold, and the monthly backup runs on a sample the authors themselves call too small.
- Weak Unconditional Pass-Through: Around a median 50 bp hike, DeFi rates rise about 15 bps after one week, a 0.3-to-1 pass-through. While the Fed added 525 bps over 2022-2023, Aave rates fell from 2.5% to 0.5% before recovering; in late 2024 a 100 bp cut coincided with a 600 bp rise. Error-correction half-lives are 5 weeks for DAI, 4 for USDC and 2.2 for USDT, and Bitcoin returns enter every spread regression significantly.
- Sticky Liquidity Supply: Of 89 daily rate jumps above 20 percentage points, 59% of the 49 increases were demand-driven, and 60% of the 40 decreases came from new supply. During the SVB depeg, USDT deposit rates exceeded 60% annualised as borrowing surged and supply lagged. Gas fees explain 33-51% of decay time for 5-25% spread spikes but nothing above 30%. USDC-only wallets (7.29% of LPs, 15% of volume) cut liquidity 1-2 points more than others after hikes.
- Utilisation Curve Consequences: Aave's USDC and USDT curves (0% base, 4% slope to a 92% kink, 10% beyond, 10% reserve factor) tie deposit yield to crypto leverage demand, not to the policy rate. Four risk proxies (liquidation, depeg, hack, rate volatility) explain only about a third of spread variation. Exchange earn products that route balances into lending protocols, as Coinbase does through Morpho, inherit a BTC-correlated yield that can fall during tightening. Tokenized T-bills could close the gap, though KYC requirements limit their reach.
Article: Not All LPs Are Equal: The Active-Passive Gap in Automated Market Maker Liquidity Provision
Date: 2026-09-29
Publisher: arXiv / Agathe Sadeghi, Dingyue Liu, Ciamac Moallemi, Xin Wan, Brian Zhu
Score: ₿₿+
Read time: 15-20 min
Summary:
- Overview: The paper splits Uniswap liquidity provider profitability into active and passive components. Standard pool-level analysis treats all LPs as identical, and this split tests that assumption. Using Allium event logs and second-level Binance mid-prices from February to November 2025, the authors study v2, v3 and v4 pools on Ethereum, Arbitrum and Base. The sample covers WETH, USDC, USDT, WBTC and cbBTC pairs with more than $10 million in volume. Profitability is a fee-inclusive 15-second markout that excludes gas. The authors use two estimators. One matches mint-burn positions closed within 20 seconds using LIFO. The other prices an infinitesimal, always-in-range LP from the pool price path. In concentrated-liquidity pools, passive LPs consistently do worse than the pool aggregate.
- Hidden Passive Losses: The method finds no short-lived positions in v2, so overall and passive markouts match (20.82 bps on Base USDC-WETH 30 bps). In v3 a gap appears: Ethereum USDC-WETH 5 bps falls from -1.15 bps overall to -1.51 bps passive, and USDT-WETH 5 bps falls from -0.91 to -1.31. The gaps are fractions of a basis point and many passive confidence intervals straddle zero, but the authors never test the gap itself for significance.
- Block Time Shapes Gap: The gap is largest and most spread out on Ethereum, where blocks take about 12 seconds and active matching happens almost entirely inside one block. On Arbitrum (250 ms blocks) and Base (2-second blocks, 200 ms Flashblocks) it stays near zero, and active LIFO volume in Arbitrum v3 is close to nil. The authors treat this as descriptive only, since gas, MEV infrastructure and LP mix are not separated. Sandwich trades exceed 50% of volume in Ethereum's 1 bps WETH-stablecoin pools.
- Fee Tiers Cushion Passives: Passive LPs earn more at 30 bps than at 5 bps. On Ethereum v3, passive USDT-WETH is +1.00 bps at 30 bps against -1.31 at 5 bps, and USDT-WBTC is +2.52 against -0.25. Both estimates flatter real passive positions. The full-range benchmark understates costs for narrower ranges, which suffer more adverse selection per unit of liquidity. Address-level matching misses vaults and multi-address strategies, so the measured active share is a lower bound.
- Liquidity Incentive Benchmarks: The paper does not analyze liquidity mining, emissions or protocol-owned liquidity. Its tokenomics contribution is showing that TVL, volume and pool-level markout overstate what passive capital earns in concentrated pools. LP rewards calibrated to aggregate pool returns therefore misjudge the passive depth they are meant to buy. The infinitesimal benchmark needs only the pool price path plus a reference mid-price and runs block-by-block. It gives a yardstick for sizing LP rewards and choosing launch-pool fee tiers that assumes nothing about LP behaviour.
Article: A prelude to the theory of Real-World Asset (RWA) Tokenization
Date: 2026-09-14
Publisher: arXiv / Wenpin Tang
Score: ₿₿
Read time: 20-25 min
Summary:
- Overview: The paper sorts tokenization by the economic role of the underlying asset rather than by asset class. Standard RWAs are the object being tokenized, stablecoins use RWAs as backing, and Digital Asset Treasuries (DATs) turn crypto holdings into listed corporate securities. It then builds two stylised models. The first extends DeMarzo-Duffie security design with continuous secondary trading, so order flow gradually reveals the issuer's private information. The second treats a DAT as a stochastic control problem in which the market-to-NAV premium, equity issuance and crypto accumulation are set jointly. Context figures include tokenized Treasuries at about $16 billion of a $38.1 billion non-stablecoin RWA market in August 2026.
- Learning Through Trading: The issuer chooses what share of cash flows and of the terminal payoff to tokenize. Kyle-style informed trading shrinks the market's uncertainty about asset quality over time, which lowers the adverse-selection cost priced in at issuance. With no secondary-market discounting, the optimum comes from five quadratic equations and a comparison of nine candidate designs. The comparative statics confirm the expected: better disclosure or oracles, deeper liquidity and stronger outside investment opportunities all favour tokenizing more.
- Premium Drives Issuance: In the DAT model, optimal common issuance is a* = (m-1)/γ, where m is the market-to-NAV ratio and γ is a balance-sheet friction. Issuance is zero at or below NAV. A premium exists only when expected crypto growth beats the risk-free rate, and it decays to 1 at the horizon. It holds only over a finite window τ*. Convertibles become more attractive as the premium shrinks, and preferred stock and debt are roughly independent of m.
- Strategy Back-Of-Envelope: Strategy's ratio fell from 1.70 in December 2024 to 0.65 by June 2026 as BTC holdings rose from 447,470 to 846,000. The calibration rests on three quarters of 2025 and an assumed 27% BTC growth rate; γ swings from 0.020 to 8.498 depending on the horizon chosen, so the implied 2.63-year financing window is illustrative only. Q4 2025 issuance below NAV (annualised rate 0.345) contradicts the base model.
- Treasury Design Parallels: The paper does not model native token emissions, staking or governance. Its usable output is the DAT issuance rule: selling a claim to buy backing assets adds value per unit only while the claim trades above NAV, and the incentive falls in proportion to the premium. That logic carries over to protocol treasuries that sell their own token for reserves. Model 1 gives a parametric frame for deciding how much of an RWA's cash flow to put on-chain.
Article: Hidden by complexity? Measuring stablecoin, crypto and decentralised finance ecosystems
Date: 2026-09-15
Publisher: BIS Working Papers / Timothy Aerts, Ronald Heijmans, Jan Paulick, Violeta Vuletic
Score: ₿₿
Read time: 15-20 min
Summary:
- Overview: The paper, by BIS and DNB economists, argues that public blockchain data is fully transparent yet hard to read in economic terms, and that headline crypto metrics depend heavily on methodological choices. It draws on full-node data for Bitcoin, Ethereum and Tron from the Mercurius platform (100 billion records) and documents three measurement problems: change outputs inflate Bitcoin transfer volumes; Ethereum is crowded with unclassifiable and imitative smart contracts; and the same stablecoin serves different functions on different chains. The authors propose bounded estimates, bytecode-based classification and chain-level disaggregation. They conclude that on-chain indicators are noisy approximations rather than direct measures of economic activity.
- Bitcoin Volume Bounds: Across 1.3 billion transactions and 3.6 billion outputs from 2009 to 2026, excluding self-transfers cuts reported Bitcoin transfer value by up to a factor of six. The gap widens from March 2016 as address reuse spreads. Simple market capitalisation reached up to four times realised capitalisation during price surges and fell below it in 2022. Both volume bounds rest on deliberately crude change heuristics, so the factor of six measures the spread between two rules of thumb, not a calibrated overstatement.
- Dormant Supply Adjustments: About 3.5 million BTC have not moved in over 10 years and 1.8 million in over 15 years. The authors treat the 15-year cohort as lost and remove it from market capitalisation. The heuristic leaks: roughly 240,000 BTC moved after more than 10 years dormant, and almost 3,000 after 15. Dust (UTXOs of 10,000 satoshi or less) totals only about 1,500 BTC, yet it spans over 250 million outputs, up from about 60 million in early 2020.
- Token Proliferation And Imitation: Of 67.5 million active Ethereum contracts, 54.2 million match no technical standard; 11.8 million are proxies and 1.4 million issue fungible tokens. The symbol uni-v2 appears on 465,576 contracts, about a third of all ERC-20 tokens. Roughly 7,000 contracts reuse the USDT symbol. Their spurious issuance reaches about 100 billion units and their transfers exceed 15 billion, rising sharply from 2020. Uniswap V2 swaps concentrate in the WETH-USDC and WETH-USDT pools.
- Chain-Specific Stablecoin Roles: The paper does not analyse supply schedules, emissions or governance. Its tokenomics contribution is measurement. Smart contracts held over 20% of Ethereum USDT at peak (10-15% since late 2024) against roughly 1% on Tron, so one token works as DeFi collateral on one chain and as a payment or store-of-value instrument on the other. Issuance surges after the US election and the GENIUS Act did not lift smart contract holdings. Counting holdings from transfer events avoids TVL double-counting.
Article: Measuring DeFi Risk
Date: 2026-09-07
Publisher: arXiv / Jeremy Bertomeu, Xiumin Martin, Ibrahima Sall
Score: ₿₿
Read time: 10-15 min
Summary:
- Overview: The paper builds a risk metric for DeFi lending pools that needs only aggregate deposits and borrowings per coin. It splits each pool into a synthetic lender holding stablecoins and a synthetic leveraged borrower holding unpegged coins. It then asks how far all unpegged prices must fall before that borrower is liquidated (RI) or becomes insolvent (RII). Applied daily to Compound and Aave from April 2021 to May 2022 (27,804 token-level observations), the measures show fragility above prudent levels in June and July 2021. The authors treat deposit tokens such as cUSDC as new stablecoins backed by a crypto basket.
- Two Synthetic Ratios: RI divides stablecoin borrowings by net unpegged holdings weighted by each coin's liquidation threshold. RII divides borrowings grossed up by the liquidation bonus by unweighted net holdings. A value of one means current prices already trigger the event; one minus the value gives the uniform price drop required. Both assume every unpegged coin falls in proportion and treat the pool as one borrower, so they measure system distance to default, not positions at risk.
- Fragility Timeline: At the mid-2021 peak, RI exceeded one, and a further fall of roughly 35% in coin prices would have impaired depositor collateral. Over the final six months of the sample, RI implied that a crash of at least 30% was needed to compromise borrowers. The authors rate risk at that point as moderate against daily price variation but exposed to a crash. The sample ends on 18 May 2022, before the June 2022 Celsius stress that a footnote mentions.
- Thin Validation Evidence: A 1,000-point rise in the Nasdaq Crypto Index (61.7% Bitcoin, 35.09% Ethereum) lowers RI by 0.12 to 0.15 and RII by 0.06 to 0.09, with R-squared no higher than 0.176 across 852 daily observations. Liquidations scaled by lagged borrowings rise across RI and RII quintiles. The validation is weak: OLS on price levels with no stationarity treatment, plus a liquidation check shown only as a quintile chart with no formal test.
- Collateral Parameter Levers: The paper does not model token supply, emissions or staking. Its design content sits in the risk parameters: RII scales with one plus the liquidation bonus, so a richer bonus that attracts liquidators also thins the cushion protecting lenders, and a lower liquidation threshold raises RI. The authors also flag that covering shortfalls by issuing a governance token such as COMP mirrors the LUNA backstop behind the 12 billion USD Terra-UST collapse.
