When long-term yields rise, bank lending tightens through a channel the textbook never mentions: falling bond prices eat the capital that actually limits how much banks can lend. The quantity of reserves, the variable the money multiplier treats as decisive, barely matters at all. The best evidence for this is a six-year experiment that ran in plain sight.
Four trillion dollars that went nowhere
Between late 2008 and 2014, the Federal Reserve bought Treasuries and mortgage bonds on a scale never seen before. It paid for them by crediting banks with reserves, the deposits commercial banks hold at the central bank and use to settle payments with each other. The monetary base, which is reserves plus physical currency, more than quadrupled. Excess reserves, the balances above any regulatory minimum, peaked at roughly $2.7 trillion.
The money multiplier is the textbook claim that a bank receiving a reserve dollar lends out most of it, the loan is redeposited at another bank, that bank lends again, and the chain ends at a multiple of the original reserves. That multiple is one divided by the reserve requirement. With a 10% requirement, $1 of reserves becomes $10 of deposits. Apply that arithmetic to 2008 through 2014 and you predict an explosion in broad money and, soon after, in prices.
Neither arrived. Broad money growth stayed in the single digits, inflation spent most of the following decade below the Fed's 2% target, and the ratio of narrow money to the monetary base fell below one. That outcome is impossible if the multiplier describes a causal process. Trillions in reserves sat on bank balance sheets because the multiplier gets the direction of causality backwards. In 2014 the Bank of England said so directly: its Quarterly Bulletin article "Money creation in the modern economy" stated that the multiplier is not how money is created. In March 2020 the Fed cut reserve requirements to zero, and lending did not become infinite.
What this means for you: a headline announcing that the central bank is "adding reserves" tells you very little about how much credit the economy will actually get.
What a loan actually does to a balance sheet
Start with the accounting, because the accounting is the mechanism. When a bank approves a $500,000 mortgage, it does not pull $500,000 from a vault of existing deposits. It records two new entries at once: a $500,000 loan on the asset side and a $500,000 deposit in the borrower's account on the liability side. The deposit is new money, created by a keystroke. Nobody else's balance went down.
Reserves enter only afterward. The borrower pays the home seller, who banks elsewhere, and the lending bank must settle that payment by transferring reserves to the seller's bank. If it is short, it borrows reserves overnight from another bank or from the central bank. So the order runs loan, then deposit, then reserves. The textbook runs it as reserves, then loan, then deposit.
This is why the more useful definition is endogenous money: the money supply is determined inside the economy by borrowers' demand and banks' willingness to lend, rather than set from outside by the quantity of reserves. The bank's real questions at the moment of lending are whether the borrower will repay, whether the loan earns more than it costs to fund, and whether the bank has enough capital to hold it. How many reserves it currently has is not on that list.
What this means for you: when you want to forecast credit growth, look at loan demand surveys and bank lending standards rather than reserve balances.
Why the central bank cannot say no
The economic mechanism that makes reserves passive is the way central banks run policy. The Fed, the ECB and the Bank of England all target a price, the overnight interest rate, rather than a quantity of reserves. That choice forces their hand.
Picture the alternative. Suppose banks collectively need more reserves to settle the day's payments and the central bank refuses to supply them. Banks bid against each other for the fixed pool, and the overnight rate shoots above target. A central bank that has publicly committed to a rate cannot tolerate that, so it supplies the reserves through open market operations, repo lending or a standing facility. The actors here have clear incentives: the bank treasurer needs to settle payments by the end of the day, and the central bank needs to keep its policy rate credible. The only way both are satisfied is for reserves to arrive on demand at the target price.
The September 2019 repo episode shows the mechanism under stress. Reserves had been drained to the point where banks hoarded them, and overnight repo rates briefly spiked toward 10% intraday, far above the Fed's target range. The Fed's response was not to defend a quantity. It flooded the market with repo funding and then began buying Treasury bills to rebuild reserves. Scarcity of reserves produced a price spike, and the central bank removed it within days.
What this means for you: the lever that constrains banks is the policy rate, the cost of reserves, and you should read central bank decisions as changes in that cost rather than changes in supply.
Where the constraint really bites
If reserves are not the limit, two things are: capital, meaning the shareholders' equity that absorbs losses, and creditworthy demand, meaning borrowers the bank actually trusts to repay at a rate that covers its costs. Capital is where the cross-asset transmission happens, and it is the part most macro commentary skips.
Here is a worked comparison. Take a bank with $10 billion of reserves and a 10% reserve requirement. The multiplier says it can support $100 billion of deposits. Now look at the constraint that binds. Regulators require the bank to hold capital equal to some share of its risk-weighted assets, loans and securities scaled by how risky supervisors judge them. Assume a 10% capital ratio and a 100% risk weight on business loans. Each new $1 billion loan then requires $100 million of capital. A bank with $8 billion of capital and $75 billion of risk-weighted assets already uses $7.5 billion, leaving $500 million of headroom. That headroom supports $5 billion of new lending. Its reserve balance has nothing to do with the answer.
Now add the rates channel. Suppose the same bank holds $20 billion of bonds with a duration of 6, meaning prices fall about 6% for each one percentage point rise in yields. A one-point rise in long rates produces roughly $1.2 billion of losses. Where those losses count against regulatory capital, that $1.2 billion wipes out the entire $500 million of lending headroom and pushes the bank below its target ratio. At a 10% ratio, $1.2 billion of capital would have supported $12 billion of loans. A bond market move has shrunk the economy's credit capacity without the central bank touching a single reserve.
One link here deserves care. Accounting rules differ: large banks recognise losses on available-for-sale securities in regulatory capital, while many smaller US banks can filter them out. Even where rules shield reported capital, depositors and markets do not ignore the losses. Silicon Valley Bank in March 2023 had ample reserves access and failed because rate-driven losses on its bond portfolio became visible and its deposits ran.
What this means for you: the move in the 10-year yield matters for credit as much as the policy rate does, because it reprices the capital base that every loan sits on.
When the plumbing does matter
The endogenous view is not a claim that reserves never bind. It is a claim about which side of the transaction moves first, and it has failure modes.
The first is collateral. Central banks supply reserves on demand against acceptable collateral. When a bank runs out of eligible collateral, the tap closes. Greece in June 2015 is the sharp example: the ECB declined to raise the emergency liquidity ceiling for Greek banks, deposits could not be honoured, and banks shut for three weeks under capital controls. Reserves were available in principle and unavailable to those banks in practice.
The second is a central bank that stops targeting a price. The Volcker Fed from 1979 to 1982 shifted toward targeting reserve quantities, and overnight rates became wildly volatile as a result. Under that regime, reserve scarcity transmitted directly into funding costs and then into lending.
The third is mistaking accommodation for stimulus. A floor system with abundant reserves removes the reserve constraint entirely, but that does not make banks lend. In 2010 and 2011, banks sat on reserves because demand from creditworthy borrowers was weak and capital was being rebuilt.
What this means for you: treat reserve quantities as a warning signal only when collateral is scarce, the policy framework changes, or funding markets show price spikes like 2019.
Reading the transmission today
The current yield curve shows where the pressure sits. As of October 2, 2026, the 10-year Treasury yield stood at 5.28% and the 2-year at 4.83%, a 2s10s spread of +0.45 percentage points. Two channels follow from those numbers.
On capital, a 10-year yield above 5% means bonds bought during the low-yield years carry heavy unrealized losses, and every further rise in long rates erodes lending headroom by the duration arithmetic above. On margins, banks fund themselves with deposits that reprice toward short rates, roughly the 2-year segment, and earn on loans and securities priced further out the curve. A positive but thin 45 basis point spread gives them a modest net interest margin cushion, the gap between what they earn on assets and pay on funding, which is enough to keep lending profitable but not enough to rebuild capital quickly.
The causal chain runs: long yields rise, bank bond portfolios lose value, effective capital shrinks, lending standards tighten, and credit growth slows. Reserve balances do not appear anywhere in it. Of all the links, the final step from tighter standards to slower lending is the one that depends most on how individual banks manage capital, so the lag is uncertain even where the direction is not.
The mechanism in one sentence: banks create money when they lend, the central bank supplies reserves afterward at its chosen price, and so the binding limits are the cost of funding, the stock of capital, and the supply of borrowers worth lending to. For anyone allocating capital, that shifts attention from the size of the central bank balance sheet toward long-end yields and bank capital ratios, because that is where a rate move turns into a change in credit.
