DeFi buy-borrow-die tied to higher loan defaults

Researchers studied the Venus lending protocol on BNB Smart Chain and found changes in borrower behavior after the 2021 Infrastructure Investment and Jobs Act. The paper, authored by Lisa De Simone of the University of Texas at Austin, Peiyi Jin of the National University of Singapore and Daniel Rabetti of NUS, reports a 24.5% decline in trading among wallets the authors infer are linked to US taxpayers. The decline is associated with increased loan defaults and a 39.6% rise in the value of defaulted loans.

The team analyzed Venus activity from Nov. 12, 2020, through July 31, 2022. They converted about 13 million transactions into 1.36 million daily borrower observations and focused on the protocol’s 15 largest tokens. The authors used the Nov. 15, 2021 enactment of the infrastructure law as an external event to isolate tax-driven behavior because the law expanded digital-asset information-reporting that traders expected would increase visibility to the IRS.

To identify likely US users, the researchers inferred residence from patterns such as trading concentrated in US business hours, inactivity on US-only holidays, and holdings of US-regulated stablecoins. The paper defines a protocol default as a loan that remains above Venus’s 60% loan-to-value limit for at least seven days without later borrowing or additional collateral. Using that definition, the sample accumulated about $133.34 million in daily outstanding defaulted debt over the study period; the figure counts daily exposure and can include the same troubled loan on multiple days.

The paper describes a tax-driven borrowing strategy that preserves unrealized gains while creating spendable stablecoins. An example: an owner who bought ETH at $1,000 and saw it rise to $4,000 could borrow $1,000 in stablecoins against the full holding instead of selling 25% of the ETH and realizing a taxable gain. The loan starts at a 25% loan-to-value ratio; if ETH falls to $2,000 the ratio rises to 50%, and continued declines or accumulated interest can trigger liquidation once the protocol’s limit is breached.

Using an instrumental-variable approach based on the law’s enactment, the authors estimate US-linked borrowers reduced portfolio trading by 24.5% after the law. Borrowers holding stablecoin debt showed an additional 23% decline in trading. The authors link the reduced trading to worse loan outcomes: a 1% rise in tax-induced illiquidity is associated with an 11.2% increase in defaulted accounts and a 39.6% increase in defaulted loan value. The study estimates the average inferred US-linked borrower deferred about $3,357.42 in annual capital-gains tax, about 17% of their trading portfolio in the sample; that estimate assumes the residency inference and reconstructed portfolios used in the paper.

The paper outlines how liquidations are meant to work and how they can fail in practice. When liquidation mechanisms operate, outside actors repay part of the loan and seize collateral at a discount, restoring the pool. When liquidations fail because of rapid price moves, thin markets or blockchain congestion, losses can fall to protocol reserves, token holders or suppliers, spreading the cost of an individual borrower’s tax-driven strategy across the pool.

The authors note several limitations. The results come from a single protocol on one blockchain, rely on inferred residency and use a specific definition of default. The study also identifies volatile collateral and operational issues in Venus’s liquidation process as contributing factors. The paper presents wallet-level activity, collateral and liquidation data to document how tax-driven holding behavior coincided with measurable changes in trading and loan performance on Venus.