A working paper on tax planning and DeFi credit risk argues that a common crypto borrowing strategy can leave lending pools carrying risks that originate in personal tax decisions. The paper focuses on users who avoid selling appreciated crypto, post it as collateral instead, and borrow dollar-pegged stablecoins against it.
The study was written by Lisa De Simone of the University of Texas at Austin, Peiyi Jin of the National University of Singapore, and Daniel Rabetti of the National University of Singapore. It examines Venus, a DeFi lending protocol on BNB Smart Chain.
The paper uses a simple example. If an investor bought ether for $1,000 and the position later rose to $4,000, cashing out $1,000 by selling one-quarter of the holdings would realize a $750 capital gain under the U.S. tax treatment described in the article. A DeFi loan offers another route: deposit all of the ether into a lending protocol, then borrow $1,000 worth of a dollar-linked stablecoin. The loan itself is not taxable income, and the investor keeps exposure to any further upside in ether.
That trade comes with a different set of risks. In the example, the initial debt is backed by $4,000 of collateral, putting the starting collateral ratio at 25%. If ether drops to $2,000, the loan-to-value ratio doubles to 50%. As interest accrues, the ratio rises further. Once it breaches the protocol limit, the code allows liquidators to repay part of the debt and seize part of the ether at a discount.
In other words, the borrower may delay a taxable sale, but the lending pool absorbs the risk tied to collateral volatility, debt growth, and the borrower’s willingness to act before liquidation. Lenders, meanwhile, usually see only a wallet address.
Venus data covers about 13 million transactions
The sample runs from Nov. 12, 2020 to July 31, 2022 and covers 15 major tokens on Venus. The researchers analyze about 13 million transactions and construct 1.36 million borrower-day observations, meaning one record per active wallet per day. About 3% of traders fall into the paper’s default category.
That default definition is not the same as delinquency in a mortgage market. The paper defines default as a loan remaining above Venus’s 60% loan-to-value threshold for at least seven days, with no additional deposits and no new borrowing during that period. Using that measure, cumulative default debt exposure totals $133.34 million.
The paper also notes that this number is cumulative daily exposure. A risky loan can be counted on multiple observation dates, so the figure does not represent principal losses from a single event.
From “buy, borrow, die” to public-blockchain lending
The paper places the strategy in a broader wealth-management context. Its traditional version is often referred to as “buy, borrow, die”: buy an asset, let it appreciate, borrow against it for spending needs, and avoid selling the asset outright.
That can defer capital gains taxes for years. As described in the article, U.S. estate rules may also allow heirs to reset cost basis when they inherit assets, wiping out much of the unrealized gain accumulated during the original holder’s lifetime.
Historically, this was a strategy associated with wealthy clients and private banks. Banks could review a client’s full financial situation, set lending limits, negotiate terms, and work through risks before forced collateral sales became necessary.
DeFi turns that process into code. Smart contracts do not review a borrower’s broader finances. They read wallet balances and apply the same collateral rules to everyone. Access becomes much broader, but the system relies on overcollateralization from the start.
Under the Venus parameters described in the paper, $10,000 in eligible collateral can support up to $6,000 in debt. Borrowing the full amount leaves very little room for a decline in collateral value. Borrowing only $2,000 leaves a much wider buffer. Oracles continuously feed market prices into the protocol, and once collateral values fall through a threshold, liquidators can repay part of the debt and receive collateral at a discount.
That mechanism is designed to protect the pool before collateral falls below the debt amount. But rapid selloffs, weak market liquidity, and blockchain congestion can all make liquidation less effective or delay execution.
Tax incentives can shape whether borrowers manage risk
The paper argues that tax motivation complicates the borrower side of risk. Reducing risk usually requires some action: trade, repay debt, or sell part of an appreciated position. For borrowers trying to postpone a taxable event, that action may be delayed.
The tendency can be stronger when unrealized gains are large or when a position is close to qualifying for a lower long-term capital gains rate. Stablecoins make the strategy more attractive because they convert volatile collateral into dollar-like spending power without forcing a sale. Borrowers can keep ETH or similar assets pledged, borrow USDT or USDC, and use the proceeds elsewhere.
At origination, the protocol sees a healthy collateral ratio. It cannot see whether the borrower’s original cost basis in ETH is very low, whether a sale would crystallize a large gain, or whether waiting a few more months would improve tax treatment. The paper’s argument is that those off-chain considerations can influence behavior precisely when a loan becomes more fragile.
The 2021 U.S. law serves as an external shock
To separate tax-driven behavior from crypto-market moves, the researchers use the U.S. Infrastructure Investment and Jobs Act, effective Nov. 15, 2021, as an identification event. According to the paper, Section 80603 expanded information-reporting obligations for digital asset brokers. Traders then had reason to expect that a larger share of their on-chain activity could eventually be reported to the Internal Revenue Service, or IRS.
The key point is expectation. The paper’s contribution does not depend on when the later reporting system became operational. It depends on the shift in beliefs triggered when the law was passed. Potential U.S. taxpayers had reason to care about the change; international users did not face the same policy shock.
The article notes that custodial brokers began using Form 1099-DA from Jan. 1, 2025 to report gross proceeds from relevant sales and exchanges. Later IRS rules also required cost-basis reporting for some transactions starting Jan. 1, 2026. Those rules apply to institutions that actually custody user assets. Non-custodial DeFi services are not currently covered, as described in the article.
The researchers compare data from before and after the law’s passage, well before formal implementation, to capture how users reacted to the prospect of future visibility rather than to a live reporting form.
Because blockchains do not disclose nationality or tax residency, the paper infers likely U.S.-linked wallets through behavioral signals: trading concentrated in U.S. working hours, unusual behavior around U.S.-specific holidays, and holdings of U.S.-regulated dollar stablecoins. The paper acknowledges that any single proxy can misclassify wallets, so it reports multiple specifications and stricter screens that combine several conditions.
Main result: inferred U.S. borrowers traded less after the law
In the paper’s main model, inferred U.S.-linked borrowers were 24.5% less likely to trade assets after the law took effect than international users. Borrowers with stablecoin liabilities showed an additional 23% drop in trading activity. The paper says that pattern fits its mechanism: borrowers already obtained cash through stablecoins while leaving appreciated collateral locked in place.
Liquidity in this study has a narrow wallet-level meaning. It refers to the daily probability that a borrower executes any asset trade, not to order-book depth, bid-ask spreads, or the market impact of large sales. The measure is designed to capture how actively a borrower manages a portfolio and whether appreciated assets remain effectively frozen inside collateralized positions.
The pattern is stronger for borrowers with larger unrealized gains and higher loan-to-value ratios. The paper also reports seasonal behavior consistent with tax incentives. Trading activity declines in December, especially in the final week, when investors tend to defer gains into the next tax year. Activity then rebounds after positions pass the one-year holding mark associated with lower U.S. long-term capital gains rates. The authors treat those patterns as support for a tax-based explanation rather than a one-off response around the law’s announcement.
The researchers estimate that U.S. borrowers in the sample deferred an average of $3,357.42 in capital gains taxes per year, equal to about 17% of portfolio size during the same period. The paper notes that this is an estimate based on the assumption that the wallets belong to U.S. taxpayers and is derived by reconstructing holdings from on-chain data and applying relevant tax rates. It is meant to indicate rough magnitude at the sample level.
Lower trading activity is linked to weaker loan performance
The final part of the paper looks at how declining activity affects loan quality. Borrowers who trade less often tend to leave risky positions open for longer. They may miss repayment windows or fail to add collateral before their ratios cross critical thresholds. A tax preference formed outside Venus can then show up inside the protocol as unpaid debt.
The authors also address a reverse-causality problem. Default itself might cause users to abandon a wallet and stop trading, rather than the other way around. To deal with that, the paper uses an instrumental-variables approach and isolates the decline in trading activity attributed to the policy shock.
Its estimates show that a 1% increase in tax-driven illiquidity is associated with an 11.2% increase in the number of defaulted accounts and a 39.6% increase in defaulted loan volume. A one-standard-deviation increase in illiquidity corresponds to about $350 more defaulted debt per borrower, or 2.7 times the model baseline.
The paper warns against treating those percentages as universal multipliers for all DeFi loans. They apply only to the subset of borrowers affected by this policy event and sensitive to tax considerations.
Smart contracts can read prices, not motives
The broader point is a blind spot in automated lending. Smart contracts can observe collateral prices, debt balances, accrued interest, and liquidation thresholds. They cannot observe a borrower’s cost basis or tax motive.
Two wallets may post the same amount of ETH and borrow under identical terms. Inside Venus, they look equally risky. In practice, one borrower may be willing to sell and cut risk, while the other will do almost anything to avoid a taxable sale. The paper frames this as a borrower-selection problem.
Overcollateralization can absorb ordinary market moves. It does not fully solve the case where borrowers are unusually reluctant to dispose of appreciated assets, continue carrying debt as their safety buffer shrinks, and trade less as conditions worsen. That concentrates risk in a group whose motives the protocol cannot identify.
If liquidation works smoothly, outside participants repay debt and sell collateral, and lenders may avoid losses. If liquidation fails, losses can be shared through the protocol’s own mechanisms by reserves, tokenholders, and pool liquidity providers. In that sense, a private tax preference can become a financial consequence borne across the pool.
The paper also sets clear limits on its findings
The authors say tax is only one source of risk. The study covers a single protocol during the 2020-2022 bull and bear cycle. U.S. identity is inferred from behavior. Default is defined in a specific way that captures long-lasting high-LTV positions rather than every troubled loan.
The paper also confirms that sharp collateral-price moves and flaws in liquidation mechanisms can create loan problems on Venus. Results may differ on protocols with deeper liquidity or different collateral parameters. Outcomes may also differ in markets where liquidation bots are more sophisticated than they were on Venus during the sample period.
Even so, the paper offers a view that is difficult to obtain in traditional lending datasets. On-chain records let researchers observe, at the single-wallet level, the full sequence of collateral, debt, user actions, liquidation, and eventual abandonment. Its conclusion is that moving a strategy once associated with private banking onto public blockchains does not erase human motives. Tax considerations and a reluctance to sell appreciated assets can still pass through the code and into pool-level risk.
