- Coin Metrics’ four-year regression study finds that perpetual-futures funding rates explain roughly 10% of $USDC’s weekly supply-growth variation and only 4% of $USDT’s, highlighting fundamental structural differences between the two largest stablecoins.
- Aave v3 carry spreads add statistically meaningful explanatory power for $USDC supply growth but offer no significant incremental insight for $USDT, underscoring how issuer composition shapes analytical conclusions.
- No single model in the study accounts for more than 25% of stablecoin supply movement, and none of the identified historical associations establishes causation or predicts future Bitcoin returns.
Why $USDC and $USDT Regressions Tell Opposite Stories
A detailed quantitative study by Coin Metrics, using weekly observations spanning four years, has put the supply dynamics of the two dominant stablecoins — Circle’s $USDC and Tether’s $USDT — under rigorous statistical scrutiny. The research tests six lagged macroeconomic and market drivers against subsequent weekly supply growth for each issuer. The results reveal a striking divergence: higher perpetual-futures funding rates are positively associated with faster supply growth for both issuers, but the magnitude and significance of that relationship differs materially. For $USDC, funding rates alone account for approximately 10% of weekly supply-growth variation after controls. For $USDT, the same variable explains around 4%. That gap alone is analytically significant, but it is only the beginning of where the two issuers part ways.
A separate regression analysis, covering data from 2023 onward, introduces a carry spread variable defined as the gap between Aave v3’s annual percentage yield (APY) on lending and a three-month U.S. Treasury yield. This DeFi-versus-TradFi yield differential adds statistically significant explanatory information for $USDC supply growth. Critically, it contributes no meaningful incremental explanatory power for $USDT or for combined stablecoin supply beyond what funding rates already capture. The divergence is not a rounding error — it reflects genuine structural and behavioral differences in who holds each token, why they hold it, and what market activities drive issuance and redemption cycles for each issuer. In total, the lagged issuer models explain approximately 13% of $USDC’s weekly supply-growth variation and 16% of $USDT’s. The implication is direct: a positive relationship for one issuer and a negative one for another can coexist inside the same aggregate total, making combined stablecoin supply figures a blunt and potentially misleading analytical instrument.
Demand Composition and the Multiple Destinations of Newly Minted Stablecoins
Understanding what drives stablecoin issuance requires separating the demand motivations that sit beneath aggregate circulation figures. Collateral demand is especially relevant to any inference drawn from supply data. A trader acquiring dollar-denominated collateral to support a leveraged derivatives position and a long-term investor choosing to hold dollar balances both contribute to stablecoin demand, yet their intentions and subsequent market actions are entirely different. The same aggregate circulation reading can therefore sit alongside fundamentally different investor behaviors, making top-line supply numbers an incomplete signal of directional market intent.
Circle’s description of $USDC minting is mechanically straightforward: a customer deposits dollars and receives newly issued tokens in return, while redemption removes tokens from circulation. Tether’s documented use cases are broader and include facilitating payments and enabling holders to exit market positions. A holder who retains stablecoins after reducing risk exposure has made a materially different decision from one who is actively positioning to purchase Bitcoin, even though both require the same dollar-denominated token. Lending protocols introduce yet another destination. Aave explains that supplied assets enter liquidity pools, earn interest, and can serve as collateral — a use case that has nothing to do with spot cryptocurrency purchases. Existing tokens can also transfer between holders without triggering new issuance at all, which further separates growth in total circulation from raw trading activity.
Exchange Reserves, Timestamps, and the Limits of Snapshot Data
That same destination distinction applies directly to exchange reserve figures. An analysis published by CryptoSlate on July 20 used stablecoin balances held at Binance and Bybit to assess liquidity available for potential Bitcoin purchases. While venue balances represent a trading-proximate location and are therefore more actionable than total circulating supply, an exchange deposit still constitutes an intermediate step — not a completed purchase. The inferential leap from “stablecoins sitting on an exchange” to “imminent Bitcoin buying pressure” requires additional evidence about trader intent and execution that balance data alone cannot supply.
The timing of data observations adds a further layer of complexity. Aavescan’s Ethereum v3 dashboards for $USDC and $USDT demonstrate how quoted supply annual percentage rates (APRs) can vary significantly between live readings and daily snapshots. A midnight UTC snapshot taken on October 19, for instance, recorded $USDC at 11.5% — a point-in-time quote that Aavescan’s own methodology acknowledges may miss intraday rate changes. Interest-index growth, by contrast, measures accrual over an interval rather than a single moment. When set alongside the Federal Reserve’s FRED database Treasury observations — specifically the DGS3MO investment-basis series rather than the DTB3 discount-basis quote — the comparison requires careful handling. TreasuryDirect defines investment yield as annualized simple interest without compounding, which places it alongside Aave’s quoted APR without introducing asymmetric compounding assumptions. However, matching calendar dates does not synchronize observation methodologies, align holding periods, or reproduce the historical APY-based carry calculations used in the original Coin Metrics study. A yield comparison of this kind also leaves the underlying risk profiles entirely intact. Aave’s own documentation identifies smart-contract risk, oracle risk, collateral risk, and network risk as material considerations that a raw yield figure does not capture.
Why This Matters
The practical implication of Coin Metrics’ findings extends well beyond academic statistics. Stablecoin supply growth is frequently cited in crypto market commentary as a leading indicator of Bitcoin buying pressure, operating on the logic that dollars flowing into stablecoins are dollars preparing to rotate into risk assets. This study challenges that inference at a foundational level. When no single model explains more than 25% of weekly supply-growth movement, and when the historical associations identified do not establish causation or carry predictive power for Bitcoin returns, the evidentiary basis for that narrative is materially weaker than market commentary often implies. The opposing regression relationships between $USDC and $USDT further illustrate why aggregating the two into a single “stablecoin supply” metric obscures more than it reveals. Issuer composition — who is minting, why they are minting, and where those tokens are subsequently deployed — is the variable that determines whether supply growth represents speculative positioning, collateral demand, DeFi yield-seeking, payment activity, or simple dollar savings. Connecting issuance to destination and execution is what would show how much of any supply increase is actually becoming a Bitcoin bid. That connection, the study’s results confirm, remains largely unmeasured.
Frequently Asked Questions
Why do $USDC and $USDT show different regression results if they are both dollar-pegged stablecoins?
Despite sharing the same dollar peg, $USDC and $USDT have different issuer structures, user bases, geographic adoption patterns, and primary use cases. Coin Metrics’ regression analysis finds that DeFi lending yields — specifically the Aave v3 carry spread over Treasury rates — explain a meaningful portion of $USDC supply growth but contribute nothing statistically significant to $USDT supply growth. This reflects the fact that $USDC is more heavily used within DeFi lending protocols, while $USDT’s demand is more closely tied to trading activity, payments, and markets outside of Ethereum-native DeFi. Treating them as interchangeable overstates their similarity.
Does stablecoin supply growth predict Bitcoin price increases?
Based on Coin Metrics’ study, the answer is no — at least not in a statistically reliable way. The lagged regression models tested explain no more than 25% of weekly stablecoin supply-growth variation, and the study explicitly states that the historical associations identified do not establish causation or test future Bitcoin returns. Stablecoins have multiple end uses — collateral, lending, payments, savings, and trading — and aggregate supply figures cannot distinguish between them. Supply growth that reflects DeFi collateral demand or dollar savings, for example, is not the same as dry powder queued for Bitcoin purchases.
What is the significance of using midnight UTC Aave snapshots versus live rate readings?
Aavescan’s methodology notes that point-in-time snapshots, such as midnight UTC observations, can miss intraday rate changes that may be material in volatile market conditions. A snapshot reading like $USDC’s 11.5% APR on October 19 represents a single moment within a continuously adjusting lending market. Interest-index growth, which measures accrual over an interval, is a more complete picture of actual yield earned. Matching a snapshot Aave rate to a same-date Treasury observation also does not synchronize the two instruments’ observation methodologies, holding periods, or underlying risk profiles, meaning such comparisons should be interpreted with caution and cannot directly replicate the historical carry calculations in the Coin Metrics study.
