What's So Special About the Fed's Bank Term Funding Program?

A new par for the course...at least in a different crisis

The Fed’s Bank Term Funding Program (BTFP) that it rolled out post-SVB is the discount window on steroids…on steroids—for at least the safest collateral. For the Fed’s open market operations collateral—effectively Treasuries and agency MBS—the BTFP does a two-step from the discount window. First, it switched from fair market valuation to par valuation; second, even after that adjustment, it applies no haircut. A $100 Treasury trading at $95 will get you $100 of liquidity.¹ (The other step up from the window is that the BTFP offers one-year funding.²)

BTFP says BFD to MTM

A common refrain in the coverage about banks’ substantial mark-to-market (MTM) losses on their held-to-maturity (HTM) assets is that those losses don’t actually materialize unless the bank is forced to sell. That’s only true in the immediate term. A bank also realizes those losses over time if it has to use market-based funding—which includes the BTFP, the discount window, and the Standing Repo Facility. The MTM losses are based on the fact that market funding rates are above the assets’ yield.

Why MTM accounting sometimes feels/is “wrong” for banks is that it prices their assets against market-based funding rates. To the extent a bank has sticky deposits, MTM accounting is “wrong” as far as the story it tells to investors. It might be more correct to price the corresponding amount of assets to the deposit curve, which in all but tail cases will price the assets more favorably than market rates.³

Thus, while the BTFP’s “penalty rate” is just 10 bps above market funding rates, it’s hundreds of bps over the deposit curve, where banks were funding themselves before needing to turn to the Fed. If marking the corresponding assets to market seemed wrong before, it clearly isn’t anymore — even despite the Fed’s help:

Interest Rate Losses IRL

As long there is no default, the experiences of credit risk and interest rate risk are opposites as an asset approaches maturity. When an asset matures, its credit risk has disappeared; its interest rate risk has been fully realized.

Thus, despite the generosity of the BTFP’s par valuation, its market-based pricing means the Fed is effectively just helping banks term out their losses. Instead of being forced to sell today, the banks are just paying funding costs higher than the asset yield.⁴

Yet, with the BTFP’s par-valuation-zero-haircut help, the Fed is underwriting any loss of *quantity* of funding on account of MTM losses. And because it’s generally overfunding relative to the market even absent the MTM losses, it reduces demands on the bank’s capital from its funding structure—despite not insulating bank capital from realizing the MTM losses over time.

The fact that these losses are a product of risk-free rate changes means the BTFP won’t save the banks from them. However, in a credit crisis or dash-for-cash, eliminating haircuts or par valuation is a game-changer—and isn’t inherently crazy. We typically think of a central bank intervention in crisis as aiming to draw a line under a fire sale, essentially telling the market “this has gone on long enough.”

Yet, it’s not obvious that typical central banking practice—lending against haircut market values—should always stop a crisis. Indeed, it often doesn’t!

Dashes for Cash & Flights to Safety

Take the early-covid “dash for cash.” A rapid, mass move into cash meant a fire sale in the Treasury market. Despite falling interest rate expectations, Treasury prices rapidly blew out several percentage points amid mass sales. And the Fed offering effectively unlimited repos did not garner enough interest to sufficiently stem the tide. Offering repos is nice if the counterparties simply lost a funder overnight. But, if the problem is the asset side, the central bank isn’t doing any better than the market.

Imagine a Treasury trading at $100. And a bank is repo-ing it for $99, using $1 of its capital. Enter: a pandemic-esque dash for cash; now the Treasury is trading at $95. Even if the market won’t fund it or it increases haircuts and the Fed steps in, the Fed is still only gonna give the bank $94 for it. Aka, the bank has to have the capital to absorb the lost funding or lever up further.

Moreover, even if the firm (and, by extension, the system) has the capital to absorb that initial selloff, the Fed repos/discount window/BTFP will margin call the bank all the way down if the value of the collateral keeps sinking. That is, the Fed hasn’t guaranteed that there’s a floor under the assets. And that means there’s still risk being borne by bank capital—which is already in scarce supply in a crisis.

And the same goes for oversold credit risk. MBS and ABS were in fetching fire sale prices in the Global Financial Crisis—and still bore significant downside trading risk. Structures like the Fed’s Maiden Lane facilities, where it took assets off private balance sheets (AIG’s and Bear Stearns’s), with the expectation that it wouldn’t lose money as a “patient investor” in the assets, helped put a floor under those institutions’ risk profiles and capital positions.⁵

Lending against haircut fair values was largely effective at bringing credit spreads in during the GFC, but not without the system becoming massively undercapitalized—to a degree much more severe than was warranted by the actual condition and long-term value of the system's assets. Capital injections restarted the system. While raising/increasing capital was almost certainly unavoidable, central banking could have limited the damage. Where was par valuation and/or no haircut when every AAA-rated structured finance security was being sold off and seeing increasing market haircuts effectively just because ABS sounds like MBS?

This is a powerful tool that the Fed has broken the seal on, and the Fed may face a lot of pressure to do something similar, even via the normal discount window, in a future crisis—particularly one where the losses aren’t simply from rising risk-free rates.

Crises Future

Stigma isn’t the only reason a bank may eschew central bank facilities in favor of fire sales that yield less liquidity. The fire sales don’t come with the risk of being margin called or facing an increasing haircut⁶; they draw a line under the bank’s losses.

The central bank may limit its own effectiveness by haircutting assets and using market values. The banking system may simply not have sufficient capital to spread some even to the central bank emergency facilities. To the extent fire sales are already underway, the system has sustained capital losses. And to the extent those securities are being further haircut at the central bank, that’s more capital being required still.

At a certain level of crisis severity, haircuts, fair value, and marking to market may needlessly overprotect the central bank from a risk management perspective (if the assets are already oversold)—and at the cost of limiting the interventions’ effectiveness by putting too much pressure on the capital left in the system.

We know that if we want to use “penalty” rates effectively, they have to be set relative to normal times. Setting them relative to crisis-time rates will render central bank lending too expensive and ineffective. It’s important we think about collateral valuation the same way.

That said, these statements apply more clearly to dashes for cash or credit risk fire sales, where assets get oversold relative to fundamentals and the central bank can step in until the excess discount disappears.

But the risk-free yield curve is as fundamental as it gets. The BTFP is validating the truth of banks’ mark-to-market losses, if at least terming them out. Those unrealized interest rate losses won’t disappear unless rates reprice favorably and markedly; they’ll be realized.

Comments welcome via email (steven.kelly@yale.edu) and Twitter (@StevenKelly49). View Without Warning in browser here.

1. A Treasury purchased at a discount, say for $97, being carried on the books at an amortized cost of, say, $98, and trading for $95 will also yield $100 of liquidity. However, that’s also true of/offset by bonds issued at premium. However however, Treasury bills are always (absent negative rates) issued at a discount. So the first sentence is a realistic description for shorter-term Treasury holdings.

2. The discount window lends for up to 90 days currently, and its authorizing statute allows for up to 4-month terms. But discount window loans are renewable at maturity. So the one-year term isn’t that special.

3. Of course, the reliability of deposits is difficult to forecast. SVB’s customers were in many cases legally required to keep their deposits with SVB… but that covenant is good for little if those customers start burning through deposits altogether.

4. This statement is about banks who need the funding because they’ve lost non-market funding (i.e. deposits). For a bank that was already funding its collateral in the market or at the DW, the Fed is just being generous in giving banks more mileage out of their collateral—i.e. a higher valuation. This could help a bank replace funding on collateral that’s more expensive to fund. E.g., it can repo less of its corporate bond portfolio, instead funding it with the extra repo funds it’s getting from its Treasuries-backed borrowings.

5. AIG and JP Morgan (acquiring Bear) had to stick some capital into the structures to protect the Fed. But that amount was finite and known. So the effect was the same: limiting the institutions’ downside risks to their capital.

6. While the Fed marks to market, it thankfully doesn’t increase haircuts in parallel with the market. The FHLBs, however, do increase their haircuts in crisis.