This analysis builds on data from Kim Snyder’s team at KlariVis, whose 122-bank, 17-month transaction-level dataset forms its empirical foundation.
The question for community bankers is not whether digital assets pose a risk to the deposit base. Most of us have worked that out. The question is what path we are already on. The data suggests it is clearer — and further along — than the public debate acknowledges.
Earlier this year, KlariVis published transaction data from 92 community banks showing customer fund flows to Coinbase. The analysis has since been extended to include 122 institutions across more than 30 states, all major exchanges and 17 months of data through April 2026. Customers are routinely moving money from community bank accounts to crypto platforms — at almost every bank we measured, in patterns that have held across a full crypto market cycle. That’s the floor.
Where We Are
Of the 122 community banks in our updated dataset, 113 (93%) show meaningful crypto-platform transaction activity — up from 90% in the original Coinbase-only analysis. Of the 84 original banks still in the dataset, 96% show meaningful activity. The pathway between community bank deposits and crypto platforms is built and active at virtually every bank we measured.
For the 52 banks where we can reliably determine the direction of money flow, the picture is clear. Customers moved $154.8 million to crypto platforms over 17 months and brought back $76.5 million — a 2-to-1 outflow ratio sustained across a full crypto market cycle. By transaction count, 91% of all crypto-platform activity was outflow. Thirty-eight of the 52 banks (73%) showed net outflows.
The persistence matters most. When Bitcoin pulled back below $65,000 in early 2026, the outflow-to-inflow ratio moderated from a high of 5.5-to-1 during the post-election rally in early 2025 down to 1.44-to-1 in Q1 2026 — but it did not reverse. Money continued to leave through the downturn, just more slowly. When Bitcoin climbed again in April, the ratio snapped back to 2.34-to-1 at the same banks.
The 52 banks have aggregate deposits of roughly $89 billion. About $78 million in net outflows against that base is less than one-tenth of one percent over 17 months. Today’s flows are speculative trading activity, not deposit substitution — but they are also not bidirectional, as a healthy two-way trading channel would be.
Broken down by deposit product, the gradient is clean: money market accounts show a 96% outflow rate; checking and DDA sit in the middle; savings is the only product where most transactions are inflows — customers bringing money back from exchanges. The variable is rate-sensitivity by self-selection. Money market customers accepted minimum balances and transaction limits for incremental yield; the 96% outflow rate is what they do when a higher-friction, higher-risk yield channel is available. Savings customers are lower-balance and less rate-sensitive, almost by definition — if they were rate-sensitive, they’d be in a money market.
The customers producing the directional bias are identifiable: rate-sensitive money market holders, moving $3,232 at a time, repeatedly, in a pattern that hasn’t shifted in 17 months. Today’s behavior remains bounded by crypto-asset volatility. The next question is what happens as that volatility declines.
Where It Goes From Here
Today’s behavior is driven by exposure to volatile crypto assets — the risk that keeps the average ticket at $3,000 rather than $30,000. The question is what happens as the asset on the other side changes in three ways.
Stablecoins as payment infrastructure (already underway). Visa, Mastercard, and PayPal have integrated stablecoin settlement. Stripe acquired Bridge. PayPal’s PYUSD circulates at a meaningful scale. Cross-border B2B increasingly settles in stablecoins because the alternative — correspondent banking with two-day settlement — is operationally worse. None of this requires the stablecoin to pay yield. Customers begin holding balances for utility, and the cash often comes from a bank account.
Yield-bearing stablecoins (conditional on Section 404). If yield-bearing products become legally permissible, the customer already comfortable holding a stablecoin balance has a reason to grow it. The 96% outflow rate from money market accounts is what happens when rate-sensitive customers have a high-friction, high-risk yield channel and use it anyway. Replace the volatile asset with a stable instrument paying 4–6%, and the same behavior scales materially. The trade groups are right to fight: the carve-out for “duration, balance, and tenure” rewards is broad enough to permit substantial gateway-level evasion if implementing rules don’t tighten it.
Tokenized investment products (happens regardless). Once a customer holds a stablecoin balance, they’re on rails that connect to tokenized investment products. BlackRock’s BUIDL, a tokenized fund holding short-duration Treasuries and money market instruments, grew from launch to over $2 billion in under a year. Franklin Templeton’s BENJI is scaling similarly. These are not stablecoins; they pay yield through investment returns and sit outside Section 404’s stablecoin yield prohibition. As stablecoin platforms integrate them natively, the customer holding a stablecoin balance is effectively one click away from a tokenized money market fund position.
The combined trajectory: stage one (today — speculative trading from rate-sensitive customers); stage two (stablecoins draw operating and working capital); stage three (yield-bearing stablecoins, if permitted, scale the existing pattern); stage four (tokenized money market funds and Treasuries compete with bank deposits at institutional pricing, 24/7). Each stage broadens the customer base. The Section 404 fight is the contest over stage three. Stages one, two, and four happen on their own trajectories.
The 1977 Parallel
The trajectory has a precedent. In April 1977, Merrill Lynch introduced the Cash Management Account, combining brokerage, money market fund, checking, and a Visa card into one product that paid significantly higher yield than any bank could legally offer under Regulation Q. With short-term Treasury yields running 8-10%, the gap created a structural arbitrage. Money market mutual fund assets grew from roughly $4 billion at the start of 1977 to $235 billion by the end of 1982 — overwhelmingly from bank deposits, particularly higher-balance customers. By the time the Garn-St. Germain Act authorized competitive bank money market accounts in October 1982, five years of migration had already happened. The customers who moved did not come back.
The structural parallel: Then, the CMA was the gateway, money market funds the destination, and Reg Q the friction. Now, stablecoins are the gateway; tokenized money markets and Treasury funds are the destination; and the friction is the operational and regulatory difficulty of offering seamless 24/7 cross-product integration. Then, the regulatory framework caught up after five years. Now, Section 404 implementing rules take effect roughly two years after enactment, in an environment where the destination ecosystem is already developing.
The most consequential difference between then and now is what community bankers can see. Banks in the late 1970s could observe deposits leaving in aggregate, but couldn’t identify which customers were leaving or how migration was accelerating within their own base. Community bankers today can. The data here is institution-level transaction data — the kind sitting in every community bank’s core system but historically buried beneath aggregation. The query that produced the 122-bank dataset can run at any individual bank.
The Question Worth Sitting With
The data identifies the leading edge: rate-sensitive customers repeatedly moving funds from community bank deposits to crypto platforms across market cycles. Today, the behavior remains bounded. But as stablecoins evolve from speculative instruments into payment infrastructure — and eventually into gateways to tokenized yield products — the competitive dynamics around deposits change.
Where institution-specific data does the most work is in implementing rulemaking. Legislative text gets negotiated by senators, staff, and lobbyists; implementing rules get drafted by career regulators who need empirical grounding. Trade group letters carry institutional weight; academic submissions carry analytical rigor. Institution-specific transaction data, on the other hand, carries ground-truth visibility into how actual bank customers behave at the level of individual accounts.
Community banks may not control where market structure goes from here. But they still control whether they understand what is happening inside their own customer base before the migration becomes visible in aggregate.
