Banks borrow from the Fed to earn more from the Fed

Banks found a profitable arbitrage in an obscure corner of the Federal Reserve's plumbing. After Silicon Valley Bank failed, the Fed set up the Bank Term Funding Program (BTFP), which lent against collateral valued at par. By late 2023 an unintended gap had opened. Banks could borrow from the BTFP at roughly 4.9 per cent and leave the money on deposit at the Fed, where reserve balances earned 5.4 per cent.

A facility built in a panic

The programme was announced on the weekend after Silicon Valley Bank collapsed in March, when depositors were pulling money from regional lenders across the country. Its central feature was generous. Banks could pledge Treasuries and agency mortgage bonds and borrow against their face value, not their market value, for up to a year. Many of those bonds had lost a large part of their value as interest rates rose, so the par valuation let banks raise cash without selling at a loss and crystallising it in their capital.

The Treasury put up money from its Exchange Stabilisation Fund to cover potential losses, and the facility worked as intended. Deposit flight slowed, and after the failure of First Republic in May the panic subsided. Borrowing under the programme settled at a steady level through the summer, used mostly by banks that wanted insurance against a renewed run.

How the gap opened

The flaw lay in the pricing formula. The rate on a BTFP loan was set at the one-year overnight index swap rate plus ten basis points and fixed for the life of the loan. When the programme began, markets expected rates to keep rising, so the one-year rate was above the rate the Fed paid on reserves and the facility was expensive. By late in the year markets were pricing several rate cuts, which dragged the one-year swap rate below the policy rate. The BTFP rate followed it down while the rate paid on reserves stayed put.

The Fed shuts the tap

That risk-free spread of 50 basis points pushed BTFP borrowing to a record above $160 billion, and an emergency backstop became a subsidised carry trade for commercial banks. The Fed, realising it was handing banks free money through an emergency facility, moved to change the BTFP borrowing rate before letting the programme expire. Banks, as ever, had taken whatever an administrative pricing error offered them.

The sums were real, though modest by banking standards. A spread of 50 basis points on $160 billion comes to $800 million a year, if every dollar were used for the trade, which not all of it was. Because the rate was fixed for a year, a bank that borrowed near the low point locked in its margin for the full term, whatever the Fed did next. Some banks had every reason to repay older, more expensive BTFP loans and borrow again at the lower rate.

Who paid for it

The cost fell on the Fed and, through it, on taxpayers. Since the autumn of 2022 the Fed has been running an operating loss, because the interest it pays on reserves and reverse repos exceeds the income from its bond portfolio. It records the shortfall as a deferred asset to be repaid out of future profits before it resumes sending money to the Treasury. Every dollar a bank earned from the BTFP spread added to that loss and pushed back the day when the Fed again contributes to the public finances.

The banks that gained were mainly regional and community lenders holding bonds bought at low yields during the pandemic. For them the arbitrage was a small offset to the damage higher rates had done to their securities portfolios and their deposit costs. Larger banks, with easier access to other funding and more attention from supervisors, were less likely to use the facility at all.

An old rule ignored

The design broke a principle that central bankers have followed since the nineteenth century. Walter Bagehot, in his account of the Bank of England's role in a crisis, argued that a lender of last resort should lend freely, against good collateral, at a penalty rate. The penalty keeps the facility for emergencies and pushes banks back to private markets once calm returns. The Fed's ordinary discount window lends at the top of its policy range, 5.50 per cent, precisely for that reason. The BTFP, by tying its rate to market expectations, lost that penalty as soon as those expectations turned.

The facilities of 2008 offer a partial comparison. Several of them lent against a wide range of collateral, but most were priced by auction or at a margin over market rates, and borrowing fell away naturally as markets recovered. They were designed to become unattractive in normal times. The BTFP became more attractive as conditions normalised, which is the opposite of what an emergency facility should do.

The case for the Fed

The strongest defence is that the facility did its job in March, when speed mattered more than precision, and that nobody at the time expected markets to price rate cuts within a year. The cost of the arbitrage is small next to the cost of a wider banking panic. Much of the borrowing was also legitimate, since banks that fear deposit outflows are entitled to lock in funding while it is available. Treating every dollar of BTFP use as abuse would be unfair.

That defence explains the error without excusing it. The pricing formula could easily have included a floor tied to the rate on reserves, and the gap was visible for weeks before the Fed acted. A central bank that wants its emergency tools to carry no stigma needs them to carry no subsidy either.

What to watch

The programme is scheduled to stop making new loans on 11 March, and the weekly Fed balance-sheet release will show how quickly borrowing runs off. The key question is whether banks that relied on the facility shift to the discount window, which still carries a stigma, or find private funding. Their first-quarter results will show how much of their recent earnings came from the spread.

The BTFP episode turned a lender of last resort into a source of easy profit for banks. The Fed has closed the gap, and the lesson for its next emergency facility is to price it above the rate it pays on reserves. Expect the programme to lapse in March without a successor and borrowing to fall steadily through the year as loans mature.