In March of 2008, the Fed started lending under its emergency authority provided under Section 13(3) of the Federal Reserve Act, which allows the Fed to provide liquidity to essentially any type of entity, in response to the Global Financial Crisis. It was the first time the Fed had provided liquidity under a 13(3) authorization since the Great Depression.

The Fed would go on to use the authority for 15 separate programs during the GFC. It would use it for eight programs during the pandemic. Most recently, the Fed used it for one program—the Bank Term Funding Program—during the Banking Crisis of 2023. The Fed can only invoke the Section 13(3) facility when a supermajority of the Fed Board determines that “unusual and exigent circumstances” are present and—since 2010—the Treasury secretary concurs.

Former Fed Chair Ben Bernanke noted that the March 11, 2008 press release for the Term Securities Lending Facility “lacked any mention of emergency powers or 13(3) authority. (We worried that trumpeting the invocation of emergency powers last used in the Depression would deepen the panic.)”

Indeed, this is how the emergency authority is often described: as not having been used between the Great Depression (when the authority was created) and the Global Financial Crisis. Occasionally, a slightly more correct version of this history will prevail and mention that 13(3) was invoked, but not lent under, in 1966, 1969, and 1980.

Prior to a 1980 law change, banks that were not “members” of the Federal Reserve System (so-called “state nonmember banks”) did not have direct access to the discount window. National banks and state member banks (those state banks whose primary federal regulator was the Fed) had direct access to the window. The Depository Institutions Deregulation and Monetary Control Act (DIDMCA) extended discount window access to state-regulated banks and thrifts that were not members of the Federal Reserve System.

In 1966 and 1969, higher interest rates put funding pressures on nonmember commercial banks, mutual savings banks, and savings and loan associations. Because these firms were not Fed members, they did not have direct discount window access. The Fed Board invoked 13(3) to allow the Federal Reserve Banks to lend to these firms and replace the lost deposit funding. The 1966 and 1969 invocations were effective for several months, but no loans were made. These programs are mentioned in some of the public historical documents of the Fed. (See, e.g., Howard Hackley’s Lending Functions Of The Federal Reserve Banks: A History.)

To my knowledge, there is just one public source that mentions there was also a 1980 invocation of Section 13(3). A 2009 speech from the then-general counsel of the New York Fed, Tom Baxter, said that in 1980, “the authority was activated – but not used – to grant a loan to a Michigan nonmember bank to pay for cash letters presented to it.” (This speech has largely disappeared from the internet, but a copy lives in the Yale Program of Financial Stability Resource Library here.)

A FOIA request for Fed Board materials related to the 1980 invocation fills out the picture—though the bank in question was in Kansas, not Michigan. The documents provided include the minutes of the March 21, 1980 Fed Board meeting (now accessible here) and of the March 27, 1980 meeting (now accessible here).

On March 21, 1980 the Fed Board invoked Section 13(3) to enable the Federal Reserve Bank of Kansas City, per the latter’s request, to lend to the Mission State Bank and Trust Company (“Mission”)—which was a state nonmember bank, and thereby not eligible for the discount window. The Fed Board authorized the Kansas City Fed to lend up to $5 million until March 27, when it could be extended. On March 27, the Fed extended the authorization to April 2. The DIDMCA became law on March 31, 1980, providing discount window access to the bank and obviating the need for the 13(3) lending.

A new threshold?

This is a little weird. While 13(3) programs have been relatively common in the 21st century, the bar for invocation of unusual and exigent circumstances has yet been pretty high: systemwide banking stresses and a pandemic. The 1960s invocations were also related to systemwide banking stresses (among nonmember banks and thrifts), and the Great Depression was the Great Depression.

Mission was the 13th largest bank in the Kansas City metro area and had $100 million in assets, hardly significant on a national scale. Moreover, no other banks were targeted by the 13(3) invocation, suggesting limited concern over contagion. While the Board minutes are slightly redacted, it seems the bank was losing deposits, and a restructuring of its balance sheet was ongoing. The minutes did note “possible adverse effects on other financial institutions in the community if the bank failed.”

There have been plenty of much more destabilizing moments in markets since 1980, for most of which I’d bet 13(3) invocation wasn’t even discussed. Given the flexibility around usage of 13(3), it’s not obvious that an invocation is any way binding precedent for the Fed (aside from what the market comes to expect). The Fed invoking the 13(3) authority for a similarly situated group of firms today (the Fed can no longer target an individual firm) seems inconceivable.

The March 21 Board minutes noted, “Board members were concerned that granting this authority might set a precedent for approval of similar requests that might be made in the future.” The minutes went on to say, however, that the authority should be invoked given the DIDMCA “legislation was expected to be approved within the next few days. Under that law, the bank would gain immediate access to the discount window on the same terms and conditions as those for member banks. As a compromise, it was suggested that the authority might be extended for only a few days.”

The minutes go on to emphasize that the Board did not want this to viewed as precedent:

However, it was reiterated that the Board would prefer that this authority not be used, and that this action did not set a precedent for other institutions that were not similarly affected by the pending legislation.

The March 27 minutes, which make clear that the loan had yet to be drawn on, reiterated:

At today’s meeting, members of the Board expressed the same reservations as they had at the previous meeting. However, in view of the proposed membership and reserve requirement legislation that was expected to be enacted within the next few days and under which the bank would gain immediate access to the discount window on the same terms and conditions as those for member banks, the Board extended the authorization through April 2, 1980.

All told, this is really just more unusual than exigent. Basically, a new law was reasonably expected in short order, under which the Fed would be able to lend to Mission under its standard discount window authority (Section 10B of the Federal Reserve Act). The Fed used 13(3) as a bridge to the ultimate passage of that law, despite the bank in question not really being one of much import. There’s not an obvious takeaway here, and the Fed was clear that this use of the authority was not to set any precedent except for for other banks similarly affected by the pending legislation. But, at the very least, this piece of Fed history should live on. Mission State Bank and Trust Company failed later that year.

Comments welcome via email (sjkelly6@outlook.com) and Twitter (@StevenKelly49). View Without Warning in browser here.

The Bloomberg article underlines the limited role the unrealized losses played in 2023 when it came to bank failures and near-failures (though certainly it was a component of contagion)—as discussed at length with Jonathan Rose here—and discusses when those losses may matter more for financial (in)stability.

Full article: https://www.bloomberg.com/news/newsletters/2025-06-26/when-should-we-care-about-unrealized-bank-losses

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

1. Narrow banking

To truly be narrow banking, stablecoins must have reserve accounts at the Fed. Fed reserves are the only asset with no liquidity, credit, or duration risk. The financial stability risks here are well-known: Stablecoins would effectively become the safest banks in the world, incentivizing depositors—particularly those over the $250,000 FDIC insurance limit—to move their deposits to stablecoins, who would, in turn, place them at the Fed. This would disintermediate the banks and force the Fed to buy substantially more assets, or provide banks with substantially more funding directly.

When the (attempted) upstart bank The Narrow Bank (TNB) tried to build a whole business around simply gathering deposits and placing them at the Fed, the New York Fed rejected its application for an account at the Fed on financial stability (and monetary policy) grounds. An excerpt from the rejection letter:

Master accounts remain a sensitive issue in the stablecoin/crypto debates, and the swing voters in Congress seem unlikely to allow master account access to even be tacitly allowed in potential stablecoin legislation.

2. Treasury bills

Ok, but what about direct ownership of Treasury bills; doesn’t that allow for effectively narrow banking without disintermediation? Some weird things can happen here.

• One goofy, but probably less important, one: the debt ceiling. Short-term Treasuries set to mature around possible “X dates” can often sell off relative to Treasuries that bookend them on the yield curve. See, e.g., Circle reallocating in 2023:

  

• More important: Much as the use of narrow banking would lead the Fed to need to do substantially more printing, material Treasury bill demand would strain the existing supply of Treasury bills. The Treasury Borrowing Advisory Committee’s recent materials (p.11) explore stablecoin demand for Treasury bills going from the current $120+ billion to an estimated $1 trillion in 2028, citing Standard Chartered research. (That probably sounds high to most readers, but it’s worth exploring anyways since—in some sense—none of this matters if stablecoin market cap never grows that much.)

  Shortages of Treasury bills relative to organic demand for such can lead the financial sector to manufacture private substitutes—which are prone to breaking down under financial stress. Prior to the financial crisis, the shortage of T-bills led the front end of the Treasury curve to price very richly, incentivizing both supply and demand of private-sector-manufactured AAA-rated short-term paper (ABCP, CP, repo, etc). From Stein, Hanson, and Stein (2016):

  

  In present day, the hedge fund basis trade can be seen simply as one manifestation of the shortage of Treasury bills, as long-term Treasury debt is funded by hedge funds manufacturing T-bill substitutes—short-term, Treasury-backed repos—for short-term cash funds to invest in.

  As Greenwood, Hanson, and Stein noted in their paper, the Fed’s overnight reverse repo facility (ON RRP)—which allows banks and money market funds to deposit funds overnight at a rate approximately equal to the monetary policy target rate—can help limit this dynamic by elastically providing money assets and putting a floor on the front of the curve. However, if unchecked, this would simply bring us back to the narrow banking world described in #1. Indeed, while the Fed ostensibly places a counterparty limit on the ON RRP, it continually raised that limit to accommodate its QE, to the limit’s current $160 billion. (For MMFs, this applies at the fund level; for instance, Fidelity had ≈$600 billion at the ON RRP in mid-2023.)

  Yet, with respect to stablecoins specifically, the Fed updated its ON RRP counterparty policy in a somewhat subtle way that seems to prevent a stablecoin backdooring into a narrow bank at the ON RRP. While BlackRock specifically created a MMF—the Circle Reserve Fund, in which USDC is the only investor—to invest some of Circle’s USDC reserves beginning in November 2022, the New York Fed’s counterparty policy for the ON RRP added this bit in April 2023:

  

  In other words: Hey Circle Reserve Fund, don’t even try.

If Treasury bill supply expanded dramatically, this would likely change the shape of the aggregate deposit franchise, but would not directly disintermediate on net, as the Treasury funds are spent into the economy as they normally would be. There’s some disintermediation in the sense of the shift in the curve of total demand for Treasury assets.

3. Treasury-backed repos

While short-term repos against long-term Treasuries likely don’t transfer much of the duration risk of long-term Treasuries to the stablecoin, some of it could remain—such as when 30-year bond prices fell by ≈8% in a week in March 2020 (particularly if stablecoin disclosure requirements are such that markets can see the collateral exposure).

The greater financial stability risk here is tying the supply of funding in money markets to crypto market cap. When cryptoasset market cap goes up, the supply of stablecoins needs to grow—and vice versa.

That means that when crypto assets fall, money market funding is at least rearranged. This is likely manageable in “normal” times for, though less so if the stablecoin market does indeed grow substantially. And certainly less so if it happens alongside a systemic event.

4. Bank deposits

“Deposits in insured depository institutions” is a common, unfortunate phrase, because it implies the institution is insured—rather than deposits up to $250,000. (Yes, in a systemic crisis, there are mechanisms by which larger amounts can be protected.)

We saw the risks to the stablecoin itself of holding reserves in banks when USDC was trading at less than 90 cents on the dollar while SVB’s depositors’ fate was uncertain; when they were rescued, redeeming USDC still faced backlog delays. Circle’s monthly disclosures of USDC’s bank deposit partners prior to March 2023 is surreal in hindsight1 (my highlighting):

However, this dynamic only turns stablecoins into a financial stability risk if they grow tremendously, such as along the lines of the Standard Chartered estimates.

The greater financial stability risk here, even at current stablecoin market caps, is akin to that noted for Treasury-backed repos: the tying of senior financial system funding (deposits in this case) to crypto market cap. That it’s typically been mid-sized banks servicing crypto means stablecoin (and other crypto) deposits can represent a material part of the funding profile (and the market’s view of their business model)—again, particularly if the stablecoin market grows.

And, to the extent the stablecoin market grows by coming into its own as a retail payment product, stablecoins would be effectively gathering insured (stable) deposits and transforming them into flightier, uninsured deposits concentrated in the banks servicing these clients.

5. Other risky stuff

Bitcoin, commodities, foreign commercial paper, whatever. Ironically, this is the one with the least financial stability risk if the stablecoin market remains around its current size and with current transaction purposes, which are, so far, limited beyond participating in the crypto ecosystem. Risky assets forming the backing reserve has the least financial stability risk because stablecoin market cap can ebb and flow dramatically in this world and only affect the markets for risk assets, rather than systemically important funding markets.

If, however, nonbank stablecoins become a $1T+ market used for mainly real-economy transactions, then there’s financial stability risk in the holders losing any money, and you don’t want this risky stuff backing any par liabilities that aren’t issued by a regulated bank. Avoiding the risks in holding these kinds of assets has seemed to be Congress’s primary motivation since it began drafting stablecoin bills. (If stablecoins stay small, however, avoiding such assets is great (stablecoin) consumer protection but may actually be a net negative for financial stability proper.)

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

While emergency liquidity support is often introduced with the real or communicated intention of preventing an “illiquid” bank from tipping into insolvency, we show that the liquidity crisis should instead be viewed as the manifestation of the market’s assessing the firm as nonviable as a going concern. For that reason, authorities should provide AHEL assistance only to institutions that they have deemed viable or that they have committed to make viable through additional interventions, most commonly through a government capital injection or merger with a stronger institution.

Despite AHEL assistance, which the authorities in several cases sized to meet all potential funding outflows from the troubled firms, in no cases did liquidity provision alone prove a “cure” to the run on the institution. We also suggest, perhaps counterintuitively, that crisis-fighters should not use the terms of an AHEL intervention to manage moral hazard. Moral hazard can be addressed in the more structural policy responses that will need to follow AHEL assistance. AHEL programs should focus on providing sufficient liquidity to get the institution through its acute crisis phase.

The full paper can be viewed here.

1. Silvergate Bank, rather than SVB, was the first bank to fail and the first bank to have a run.

2. Bank business models that materially focused on the venture capital and crypto sectors were at the center of the crisis.

3. Unrealized losses and uninsured deposits are inadequate explanations of which banks experienced runs in 2023.

4. While the crisis affected midsize “regional” banks in general, the most severely impacted banks were geographically clustered around the tech- and crypto-heavy West Coast.

5. Affected banks had been under considerable prior market pressure in 2022 and were not suddenly scrutinized for the first time in March 2023.

6. The historically unprecedented speed of the 2023 bank runs is not well explained by social media or digital banking applications.

7. The crypto and VC sectors had a special role in triggering the 2023 runs because the affected banks had business models focused on them. Once the runs were initiated, other large, sophisticated depositors ran with roughly the same speed and severity.

Read the full paper, which concludes with its implications for the post-2023 policy reform debate, here: https://www.chicagofed.org/publications/working-papers/2025/2025-04

As with any good Rorschach test, there are merits to both views, and FDIC inspector general report from October 2023 gets at the crux of why both views carry water: namely, crypto-related activities can be very risky for banks, but the FDIC has also lacked a clear policy for providing supervisory feedback in a timely manner. As the FDIC OIG wrote:

The FDIC issued letters (pause letters), between March 2022 and May 2023, to certain FDIC-supervised financial institutions asking them to pause, or not expand, planned or ongoing crypto-related activities, and provide additional information. However, the FDIC did not (1) establish an expected timeframe for reviewing information and responding to the supervised institutions that received pause letters, and (2) describe what constitutes the end of the review process for supervised institutions that received a pause letter.

Notably, the FDIC is the one in the scopes on this for its FIL-16-2022, but the OCC and Fed have essentially equivalent supervisory letters out—IL-1179 and SR 22-6, respectively. (Will Jonathan Gould and, say, Miki Bowman follow Travis Hill and release those agencies’ supervisory correspondence on crypto as well?)

While the debanking fire-and-brimstone crowd see the letters as a silver bullet in favor of their argument, that’s a hard sell. Any view on these letters requires some interpretation. So, rather than another article trying to read between the lines, I just wanted to mention some examples where we know supervisory debanking wasn’t happening. At the very least, this forces the debanking narrative to contend with some very public exceptions—particularly, for the largest banks.

For instance, while Coinbase’s Paul Grewal has effectively become the chief spokesman for the advocates of the debanking narrative, Coinbase has very publicly banked with JPMorgan—*checks notes* the largest and most reputable bank in America—for years. (Coinbase also has a stake in, and derives significant profits from, Circle and its USDC stablecoin, a company that is also a main player in this story—more on that below.)

March 2023

In the debate over debanking, those who contend it was all above-board supervising always have the checkmate move of pointing to Signature and Silvergate, which faced massive runs following the failure of FTX. Those were real safety and soundness risks that ex post justified the heightened supervisory scrutiny, if not the lack of the rigorous and timely process that the FDIC OIG noted.

But there’s two other things to note from the March 2023 episode.

1. Many of the debanking accusers like to remind the debate that the supervisory scrutiny wasn’t just over things like investing in crypto or lending against it, but about even disallowing crypto—and VC (after all, this debate started in earnest thanks to Marc Andreessen)—firms from accessing basic banking services like checking accounts and payroll.

   • Yet, the same officials allegedly implementing Operation Chokepoint 2.0 also invoked the FDIC’s “systemic risk exception” to ensure that SVB’s venture capital depositors and Signature’s crypto depositors did not get their deposits haircut. In making this determination, Fed and FDIC staff specifically noted that their actions on systemic risk would also prevent SVB’s and Signature’s failures from disrupting their clients’ ability to make payroll. As the GAO wrote (emphasis added):

     FDIC and the Federal Reserve reported observing that many of the uninsured depositors of the banks were corporate enterprises. Therefore, losses to these firms or an inability to access their funds for even a short time could put these firms at risk of not being able to make payroll and pay suppliers, potentially causing disruptions to U.S. market and industrial operations. The regulators cited examples of disruptions and losses. The Federal Reserve indicated that several depositors of SVB were unable to make payroll payments at the end of the week leading to the failure. In addition, several payroll companies contracted with SVB to process paychecks, which led to delayed payroll for companies that did not bank at SVB. The Federal Reserve also reported that some companies that held deposits at SVB were forced to sell their uninsured deposit claims at 90 cents on the dollar on March 10, 2023, to make payroll.

     While a large part of this decision process was clearly about contagion to other banks and the real economy, banking regulators made the decision to make these depositors whole—rather than draw a firebreak on the first bank after SVB and Signature and just focus on stopping contagion. Or, better yet, they could’ve waited a little bit longer and—in addition to forcing losses on Signature’s crypto depositors and SVB’s VC depositors—could’ve let First Republic, PacWest, and Western Alliance melt down before creating the firebreak at the next bank. Point being: the rescue actions lean against there being a strongly held political vendetta against these industries’ having checking/payroll accounts.

2. Circle’s USDC stablecoin briefly broke the buck as its industry-leading transparency1 came back to bite. Here was a footnote from its February 2023 reserves examination report:

   

   Circle disclosed it had $3.3 billion stuck at SVB despite attempts to withdraw. USDC traded at less than 90 cents on the dollar that weekend — until the government announced it would stand behind the uninsured deposits of the failed banks, and it continued to loiter at a discount to until USDC announced it had cleared the backlog of redemptions:

   

In the week before the SVB failure, Circle had moved $5.4 billion to to BNY; by the Monday following the failure, had its ~$10 billion deposit pile exclusively at BNY (in addition to other invested assets):

Also as soon as Monday, Circle had two new banking partners for minting and burning stablecoins:

Taking a microscope to the March 2023 petri dish, you’d still need to squint to find a debanking story.

The Debanking Years Broadly

More generally, if press releases and headlines are any indications, big banks’ crypto projects marched on throughout the years that supervisors were allegedly choking off all things crypto-related.

Indeed, many of the released FDIC “pause letters” were asking banks questions about their proposed projects that involved internal blockchains or partnering with outside firms on crypto-technology projects. Yet, contra the debanking narrative that these questions were just a way for supervisors to infinitely play for time, the big banks (non-exhaustively):

• Decrypt, May 17, 2021:

  

• Decrypt, May 17, 2022:

  

• Wall Street Journal, August 22, 2022:

  

• New York Magazine, March 16, 2023:

  

• Bloomberg, May 9, 2023:

  

• Bloomberg, July 6, 2023:

  

• Bloomberg, September 18, 2023:

  

• Bloomberg, December 19, 2023:

  

  

Big Banks Make Crypto Look Small

“Debanking” of the kind that’s currently in the zeitgeist seems to be particularly right-wing-coded. Ditto for a desire for regulatory and supervisory “tailoring.” Usually, that desire for tailoring is a desire to resist the (left-coded) urge to extend the regulatory/supervisory standards GSIBs down to midsized and small banks. Yet, appropriate “tailoring” of crypto banking risks probably means providing much greater scrutiny to midsize and small banks, for at least a couple reasons.

• New technology, with new money laundering risks, makes for huge fixed costs of compliance. GSIBs may already be there, or can get there with marginal costs. Small/midsize banks may need much further investment. E.g.: Silvergate’s internal real-time payments network, SEN, did not have functional automatic transactions monitoring for at least 15 months up to the FTX failure.

• The largest banks don’t need to take the risks of banking the margins of the cryptosphere to source their low-cost deposits. Banks like JPM and BofA already have huge franchises of offering 0% interest checking accounts, which customers accept on account of the convenience, other services, etc that these large banks offer. If you’re a smaller bank looking for a big pot of zero-interest deposits, offering to bank the crypto world looks like a pretty tempting way to do so. But the math assumes you can manage the franchise, avoiding the AML/KYC and financial stability risks. One way to read the supervisory letters is that they were asking banks to show their work on those fronts first, rather than the banks diving into the crypto industry and developing these management capabilities thereafter.

  • Crypto banking isn’t the only siren-song business model that suffers from this. SVB was also effectively paying nothing on its venture capital deposits; there’s certainly much less AML risk there, but the financial stability risk came home to roost. Maybe there’s a National Bank of Commercial Office Real Estate that will be next.

  If Circle fails, no one will worry about BNY’s stability. If Coinbase fails, JPM won’t be in more than one headline. These banks can take on these clients without crypto (or VC) becoming their new-and-unstable business model—and without making supervisors unduly nervous about their AML programs. The same can’t be said for smaller banks. As the January 2023 joint statement from the FDIC, OCC, and Fed said, “the agencies have significant safety and soundness concerns with business models that are concentrated in crypto-asset-related activities or have concentrated exposures to the crypto-asset sector.” That concentration would just hard to reach when you’re a bank with a massive balance sheet.

Distinctions between the underlying banks’ sizes may help explain the apparent discrepancy between the possibility of debanking in one interpretation of the pause letters and the many crypto-positive announcements of big banks.

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

Abstract

Despite the often-messy politics of Federal Reserve actions, particularly in crisis, the US Treasury Department has shown substantial unity with the Fed during modern crises. During the Global Financial Crisis of 2007-2009 (GFC), the COVID-19 pandemic in 2020, and the Banking Crisis of 2023, the Fed and Treasury cooperated to implement substantial and novel policy responses. This cooperative relationship has led to the increasing use of what this article calls the Treasury-Equity Model of Fed emergency lending. In the Treasury-Equity Model, fiscal resources available to Treasury are combined with money-creation powers of the Fed, with the intent of expanding what the Fed can accomplish with its emergency liquidity authorities provided by Section 13(3) of the Federal Reserve Act. This Model was conjured in the GFC, intended to alleviate the legal risk of a 13(3) program for the Fed—and later expanded to be a tool for enabling additional financial risk as well. Implemented for the final 13(3) program of the GFC, this Model became the standard during the pandemic response, covering seven of the nine Fed facilities rolled out in 2020. When the Fed invoked 13(3) in 2023, it implemented the Model once again. Before this Model becomes an institutional norm, to the detriment of other structures, it should be reevaluated in light of its history.

While theoretically a way to expand the Fed’s liquidity provision by supplementing it with loss-bearing fiscal resources, the history of the Treasury-Equity Model shows the politics of using fiscal resources lead to Section 13(3) programs with either the same or lower risk tolerance than the Fed could have accomplished without Treasury assistance. Even for programs that end up with the same risk tolerance, Treasury funding only slows program design and implementation. Moreover, the legal risk that functioned as the genesis of the Model appears to have been alleviated by post-GFC legislative reforms to the Federal Reserve Act. Thus, this article concludes, the Fed should be much more selective about accepting funding from the Treasury into its Section 13(3) emergency liquidity operations. Without an explicit understanding with Treasury that any fiscal support is meant to be spent—not lent—accepting Treasury funding support offers only drawbacks to a Section 13(3) program.

Full paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5075729

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

The machine-translated version reads less romantically, but is more decipherable. There are some interesting new bits on the Fed’s (mini) role, which are laid out below. (It’s not an exhaustive list, just the ones I found notable.)

As a reminder, Credit Suisse had a business model in peril for years—which ultimately led to a preseason run that started in October 2022, continued slowly for months, and culminated in fatal run in March 2023, a week after “SVB weekend” in the US. The death knell for Credit Suisse was its largest shareholder saying the words “absolutely not” on Bloomberg TV when asked if they’d put any more capital into Credit Suisse; this was on Wednesday, March 15. The merger with UBS was announced that Sunday, March 19.

Prior to this parliamentary report, we knew the Fed lent—between March 15 and March 22—at least $60 billion to the Swiss National Bank (SNB) through the Fed’s FIMA Repo Facility, for the benefit of their lending to Credit Suisse. I say “at least” for two reasons:

• SNB Governing Board Member Antoine Martin said that “almost half” the USD 182 billion of liquidity assistance extended was provided in dollars—and that the SNB relied mainly on the FIMA, which lends dollars against US Treasury holdings, to source dollars.

• The SNB reported that it had to post $75 billion of collateral to the FIMA facility, which would make little sense for a $60 billon loan from a Treasury facility with minimal to zero haircuts.1

Okay, onto the new report’s interesting Fed bits, with the pdf screenshots courtesy of Google Translate (and the pursuant asterisk that nothing should be seen to be verbatim):

When the first exodus of funds from Credit Suisse began in October 2022, the Fed got uncomfortable with Credit Suisse’s US subsidiary drawing down its liquidity buffer and heightened its liquidity reporting requirements:

In cross-border discussions in November 2022, the Fed spurned the idea of Credit Suisse’s savior being a US bank:

The footnote cites concern about contagion to the US:

Come that fateful March 15, 2023, after the largest shareholder said no to the idea of injecting more capital, the NY state regulators apparently looked to ring-fence the US Credit Suisse operation, with the Fed having to talk them down:

The report says that on Thursday, March 16, the Fed said it could not rule out the default of the US subsidiary if no liquidity support was forthcoming,2 and it (perversely?) asked for a higher liquidity buffer:

Indeed, that weekend, the Fed raised the US subsidiary’s liquidity requirement by $8 billion (about 30%):

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

Treasury Secretary nominee Scott Bessent has suggested, however, that he will take the opposite tack — which potentially sets up an even greater relative scarcity of bills (and surfeit of duration).

Taking that as given, it is even more incumbent upon the Fed to not also be chasing the limited supply of bills. Its balance sheet “normalization” should be neutral to the Treasury’s duration profile, not aim to recapture its pre-GFC model of being focused on bills.

Or so I argue in a new piece over at the Official Monetary and Financial Institutions Forum (OMFIF): The Fed Should Stay Out of the Fight for T-Bills (free to read)

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

It had four answers for us:

1. Scarce use of leverage.

   There is limited leverage in these structures. This reduces the risk of fire sales, illiquidity, and other insolvency outcomes. (Private credit also asked the banks why they had never thought of that.)

2. Matched funding.

   Private credit investment funds have term funding. They lock up long-term funding for several years, and thus avoid the redemption risk associated with maturity transformation. (They again asked why banks hadn’t tried this.)

3. No volatile marking to market.

   These are private assets, with no tradable prices. This allows private credit managers and their investors to report more stable prices based on “fundamental” values.

4. No retail allowed.

   Private credit need not be feared because the ecosystem would only be opened to sophisticated, institutional investors that understood and could bear the risks.

As the market has matured, let’s check in on these characterizations. None has been wholly compromised, but each is slipping.

1. Leverage.

   While data is scarce, leverage is increasing in private credit by all accounts. Private credit is increasingly making use of net-asset-value (NAV) financing, subscription financing, collateralized fund obligations, seller financing, and more. These various structures have different designs/uses, but they all allow private credit funds to do more financing on less client funding. That is, they raise leverage.

2. Redeemability.

   From an investor’s perspective, locked-up funding is a drawback. New private credit fund structures have thus arisen to move towards solving that problem. “Evergreen funds” and “interval funds,” while not offering unlimited redemption on demand, are growing experiments in providing investors more opportunities to redeem a portion of their funds. They thus run some risk of maturity-transformation-related losses and/or needing to impose gates (the latter of which we’ve already seen in analogous structures).

3. Secondary trading.

   Many firms—e.g., JPMorgan, Golub, Apollo—are building out loan trading desks for private credit loans. This will add liquidity and, by extension, more regular market-based marking of assets and any pursuant volatility. Several asset managers are pursuing private credit ETFs, with State Street and Apollo together filing for one last month.1

4. Retail onboarding.

   The ETFs are an example of the many attempts to bring retail investors into the private credit fold. As private credit’s distributions to investors have slowed (which is also part of the reason for the increased use of leverage), it has found more limited interest/ability from institutional investors to provide fresh funds. While countless investors are likely still under-allocated to private credit, the low-hanging fruit increasingly looks to have been harvested. As such, the retail demand is next market for private credit to turn to. Early examples abound.

   Per Bloomberg, emphasis added: “But private credit funds are finding it harder to raise capital from the largest investors as still-high interest rates weigh on financial assets. With quarterly inflows from the biggest investors near multi-year lows, according to data provider Preqin Ltd., small investors are becoming an increasingly important feeding ground for private credit firms.”

   Per the Wall Street Journal last week: “Now many institutional investors are full up on alternatives, and money managers are building products aimed at individual investors instead.”

It’s now a well-trodden joke to note that as the crypto markets grew, they relearned every lesson that traditional finance had learned decades and centuries ago—with the ultimate lesson coming with the “runs” and failures in 2022. As private credit hits its own growing pains, it seems to be increasingly understanding why we have to structure banks the way we do.

Other recent notes on private credit:

The bank lobby uses this risk to suggest lighter bank regs and lower capital requirements: This would, their argument goes, at least keep the risks in supervisors’ sights. The anti-bank lobby underlines the same risk to argue for widening the regulatory net: If raising capital requirements pushes risk out of the system, they argue, then regulators should also get more authorities over nonbank financial institutions (NBFIs).

They’re of course both right and both wrong; it would depend on the specific circumstances. But one thing worth noting is that, to the extent bank regulators are worried about the risks of shadow banks like private credit/equity, hedge funds, etc.—and particularly that those risks might boomerang back to the banks during times of stress—they can influence such risk-taking in the normal course of supervision/regulation, and increasingly are.

Because there is a fixed amount of money willing to take on risky investments, risk can shift to nonbanks, but there’s a limit to how much funding can move absent the growth of bank leverage. And indeed, we’re seeing increasing bank-provided leverage in private credit/equity—while banks’ prime brokerage businesses become their crown jewels. Banks are re-tranching: moving the capital risk to an external balance sheet that can be funded with a market-determined amount of capital rather than a (higher) bank-regulation-determined amount—and taking a protected stake in that risk by providing the senior funding.

Regulators are increasingly eyeing these businesses for potential risks though, and can always rein them in via their bank supervisors. As IFR interview with one of Citi’s credit executives reported, “banks get far more information about the exposure in the portfolios and funds that they finance compared to if they were buying public securities.” Supervisors can ask for this information and frown upon what they see.

I’m making no attempt to take a systematic look across the world, but here’s some recent transatlantic developments on this front—all of which come in addition to the usual flagging of such risks in official reports:

The Fed

Published alongside the Fed’s 2024 stress test of large banks was the Fed’s “first exploratory analysis.” Of the four exploratory tests, two “market shock” scenarios applied only to the GSIBs and included the simulated failure of each GSIB’s five biggest hedge fund counterparties. While these tests did not affect regulatory ratios—though it seems like that could be coming in the future (unless there’s a party change in the White House)—the Fed said that “the use of additional market shocks offers insight into how the trading books and counterparty concentrations of the largest and most complex banks would change under a range of different market conditions.” The Fed is clearly drilling down on the risks to banks from increasingly funding these shadow banks.

In June, the Fed published a draft Federal Register notice requesting comment on its plans to expand its data collection via the FR Y-14 forms—which collect supervisory information on large banks that is used for, among other things, the Fed’s stress testing and its setting of capital requirements. The information collection proposal seeks to sharply expand the level of detail the banks report on their NBFI clients/business. As just some examples, the Fed is seeking expanded reporting from banks on:

• Their NBFI borrowers’ balance sheet metrics;

• What specific kind of NBFI is borrowing form the bank—“(e.g., credit fund, broker-dealer, special purpose entity, etc.)”;

• Whether or not their loan clients have a “financial sponsor”—i.e., whether a borrower is a private-equity-backed firm;

• Reported “collateral market value,” so that the reporting includes updating the value of collateral that has no market prices.

  • This would presumably cover the assets that private credit/equity funds are posting for net-asset-value (NAV) loans. The increasingly popular NAV loans are used by private credit/equity funds to lever up and to fund distributions to investors. That is, they represent bank leverage being used to maintain and scale shadow banks.

Across the pond

Bloomberg, which is apparently getting really good at guessing bank supervisors’ passwords, has reported on several similar supervisory developments in the UK and Europe.

It reported the ECB is probing banks’ lending to private credit firms and their underlying funds. The Bank of England is similarly asking banks “to provide more information about their offering of net-asset-value loans to buyout funds,” inclusive of how much capital they’re allocating and how much borrower leverage they’re allowing.

Bloomberg also reported that the Bank of England’s on-site supervisors have begun asking banks “how much information their hedge fund clients are disclosing about their business and whether it was enough for [the banks] to understand the risks involved in the prime brokerage business.” The report added that the BoE was also examining banks’ rates and commodities prime brokerage business lines—and that the Federal Reserve and other foreign supervisors have also been involved. Moreover, it said:

After their review of banks’ businesses in the aftermath of the collapse of Archegos and the LDI crisis, the BOE sent letters ordering banks to clean up their equity financing and fixed income finance businesses. The central bank may do the same after its latest round of on-site reviews wraps up, the people familiar with the matter said.

Financial distress events often get described as something being the “spark” and something being the “tinder.” Trying to prevent shadow banking risk buildup by using bank supervision to squeeze shadow banks’ banking partners would be akin to simply taking away the oxygen.

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

Previous notes on shadow banking/private credit:

Most of what he discussed was not new. As he outlined in May, the Fed is considering:

• Establishing a supervisory expectation of discount window readiness, inclusive of prepositioning collateral at the discount window such that the sum of lendable discount window collateral value + the bank’s reserves = some sufficient proportion of uninsured deposits.

• Limiting banks’ inclusion of held-to-maturity assets in their calculations of regulatory liquidity ratios.

• Ratcheting up liquidity regulations’ deposit outflow assumptions for some types of depositors.

• And, broadly, revisiting the scope of liquidity regs for large banks. Presumably, this is about bringing more banks into the fold of the full LCR, etc.

(A previous Without Warning note discusses that speech, and particularly the shortcomings of the Fed’s designs for that third bullet point; see The Fed Is Hunting Bank Runners (May 30).)

In May, he said these were for banks “of a certain size.” Yesterday’s speech clarified marginally to say “community banks would not be covered, and we would take a tiered approach to the requirements.”

The non-update

In another section of the speech, however, Barr outlined a recent Fed update to its published FAQs on Regulation YY rules around banks’ internal liquidity stress tests (ILSTs). ILSTs are bank-run tests and are largely black boxes—i.e., not straightforward rules with published metrics à la the Liquidity Coverage Ratio (nor should they be).

Barr said the following (emphasis added):

We had been hearing that some were confused about how banks could incorporate ready access to the discount window and the SRF into their contingency funding plans and internal liquidity stress tests. Supervisors have a role in assessing the viability of large banks' plans to meet stressed outflows in their stress scenarios, and we have been asked whether the discount window, the SRF, and also Federal Home Loan Bank advances can play a role in those scenarios. The answer to this question is "yes."

We provided clarity to the public in August on permissible assumptions for how firms can incorporate the discount window and the SRF into their internal liquidity stress-test scenarios.

Some analyst coverage has interpreted this to mean the Fed has started giving banks credit in their ILSTs for assets prepositioned at the discount window/Standing Repo Facility.

That is not the case and, in any case, wouldn’t be that important given the assets included. Barr simply described what the August FAQ clarified, which was that banks’ ILSTs are allowed to assume that the discount window (and SRF and FHLBs) are viable monetization channels (only) for high-quality liquid assets (HQLA) in a stress scenario.

That is, banks with Treasuries and agency MBS sitting at (or at least nearby) the Fed can point to that as viable channel for monetizing those assets in a crisis. The FAQ specifically says that banks should not exclusively rely on these channels, however:

That is, a bank with all its Treasuries sitting at the Fed still can’t assume the Fed as a monetization channel for all those Treasuries in a stress scenario.

But this updated clarity is important for the banks that were hearing from some supervisors they couldn't use the SRF or discount window, even for Treasuries/etc, as part of their ILSTs—despite those being standing facilities. Anecdotal reports have suggested different banks were receiving different messaging (after all, Fed supervision spans 12 regional Reserve Banks) with regard to assuming access to the SRF/DW in liquidity tests.

But this is not a terribly meaningful change to the outcomes of these ILSTs, particularly in the cases where supervisors were already allowing it. The monetization value of USTs is likely similar whether at the Fed or in the market. There may be a slight numerator impact to the extent ILSTs were assuming higher haircuts on repos/sales of Treasuries in the private market during stress events; while the Fed will still rely on the securities’ market values at the discount window, it doesn’t raise haircuts in a crisis.

The updated FAQ specifically says this guidance is *not* an expansion of the assets that count towards a firm's liquidity:

And it underlines that HQLA (“HLA” in the excerpt above) still has to be HQLA. That is, the Fed has not assigned some liquidity value to any collateral held at the window. This change is not on the order of the Fed giving some “numerator” credit to a bank’s ILST for, say, a corporate loan that is pre-positioned at the Fed. The Fed has just clarified that banks can, in fact, assume they can take HQLA to the Fed in a stress scenario.

This FAQ update does nudge on a couple things:

• It clarifies that banks’ ILSTs need not be able to also point to a different channel of monetization for things they plan to monetize at the Fed. This reduces some planning burden on proving access to additional funding sources that they do not intend to use. (It probably matters most for the SRF and FHLBs, given the stigma banks aim to avoid around the DW).

• It improves the fungibility of reserves and Treasuries.

  • This was really the context in which Barr was discussing the FAQ change—as opposed to a change in discount window policy. As Barr put it,

    We have heard over the years, however, that the degree of substitutability among these assets has been limited by concerns about capacity in stress for the market to turn securities into reserves immediately; these concerns are valid. This constraint can be addressed in part by the appropriate incorporation of Federal Reserve facilities into monetization plans in firms' internal liquidity stress tests.

  • The ILSTs, by rule, “must include an overnight planning horizon” among other time horizons. Monetizing HQLA at the Fed is a t+0 operation; whereas, private market channels may be t+1 (or longer), have certain time-of-day deadlines (which the SRF also has…), or have other administrative hurdles. If a Treasury is sitting at the Fed discount window, it can become reserves in an instant.

Thus, the FAQ clarification may help incentivize banks to shift their reserves demand into reserves/Treasuries demand—which may allow the Fed to shrink its portfolio further, at least in theory (non-bill Treasuries may be more imperfect substitutes with reserves).

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

It’s thus worth examining the private credit boom in the context of its timing. The rapid increase in rates in 2022 saddled banks with “hung debt”—loan deals intended to be syndicated that lost substantial value after being written. Some $80 billion in hung debt in the year following the start of the Fed’s rate-hike cycle meant that a combination of nursing losses and avoiding selling at a loss left banks’ leveraged lending balance sheets occupied. For instance, in the first quarter of 2023, private credit underwrote 94% of buyouts by quantity and 70% by dollar amount.

As the Bank for International Settlements wrote in its 2024 Annual Economic Report (emphasis added),

The restrained lending by banks during the recent tightening cycle and the temporary drop in syndicated leveraged loan issuance created the opportunity for private credit funds to make further inroads into areas traditionally dominated by banks.

While the famously hung Twitter loan continues to plague banks, bank balance sheets are otherwise clearing up. Bank-syndicated leveraged loans (LLs) are trading at a 22-month high in the US. While private credit direct lending (DL) picked up the banks’ slack in 2022 and covered their refinancings in 2023, private credit deals are now increasingly being refinanced by the banks:

As the banks became competitive again and high rates persisted, private credit began facing an erosion of “illiquidity premiums” and increasing defaults. As a result, private credit is also having to do more loan modifications, such as payment-in-kind transactions and “extending and pretending.” This is impacting the ability to return funds to investors and to put money into new loans. Available private credit funding is increasingly going to “secondaries”—funds that buy up stakes from other investors at a discount—or are being rolled into “continuation vehicles,” in which investment managers create a new fund to hold onto the assets of a retiring fund and try to get most of the investors to stay on.

While private credit is clearly not “so back,” that’s also not to suggest that it’s “so over.” Investors have increasingly accepted that it is a diversifier for a public markets portfolio—particularly insurers. And insurers are increasingly part of, or partnered with, private credit firms. For instance, Apollo CEO Marc Rowan recently said it can raise more than enough annuity funds (at least in this higher rates world)—through its life insurance and retirement savings “Athene” unit—than it needs for the private credit deals it can find.

Private credit is here to stay, but we may have already seen the passage of its golden age.

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

“How the FDIC Sourced Crisis-Time Fed Funding Through the Failed Banks of 2023” is out this morning.

The FDIC wanted liquidity in 2023 as relatively large regional banks failed, but it also had to navigate the institutional stigma of drawing on the Deposit Insurance Fund or its credit lines — plus, the debt ceiling...

Two particularly novel things happened:

1) The FDIC freed up collateral encumbered at the discount window by replacing it with an FDIC guarantee as security—keeping many of the Fed loans open.

2) The FDIC used its guarantees to get the bridge banks partially *uncollateralized* loans from the Fed.

The piece is necessarily byzantine, but tries to unpack a novel crisis-time liquidity channel for the FDIC—and where it comes up short.

Full article again is here.

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

The reforms consist of few initiatives for banks “over a certain size”:

• Establishing a supervisory expectation of discount window readiness, inclusive of prepositioning collateral at the discount window such that the sum of lendable discount window collateral value + the bank’s reserves = some sufficient proportion of uninsured deposits

  • The WSJ has reported the Fed is hovering around a requirement of 40% of uninsured deposits.

• Limiting banks’ inclusion of held-to-maturity assets in their calculations of regulatory liquidity ratios—due to the drawbacks/challenges of monetizing HTM assets.

• Ratcheting up liquidity regulations’ deposit outflow assumptions because: “Observed deposit withdrawals from high-net-worth individuals and companies associated with venture capital or crypto-asset-related businesses suggest the need to re-calibrate deposit outflow assumptions in our rules for these types of depositors.”

• And, broadly, “revisiting the details of the application of our current liquidity framework for large banks.” Presumably, this is about bringing more banks into the fold of the full LCR, etc.

Assuming higher outflows for crypto and high-net-worth individuals

While the discount window stuff has gotten much of the focus, the third bullet has perhaps gone a bit under-scrutinized.1 Frankly, it’s a bit surprising the crypto diehards haven’t yet latched onto it as another “operation chokepoint” meant to squeeze crypto out of the system. (This is not a suggestion!)

High-net-worth individuals

Regarding high-net-worth individuals, it may make sense to assign them their own assumptions in liquidity regulations. As of now, the US liquidity coverage ratio (LCR), which mandates enough liquidity for a hypothetical 30-day stress period, distinguishes between “stable” and “less stable” retail deposits—which is roughly insured and uninsured deposits, respectively.

Stable deposits are assumed to have a runoff rate of 3% in a 30-day stress window, while less stable deposits get 10%. (The Basel LCR guidance suggests 10% and up for less stable deposits, and they give specific mention to high-net-worth individuals.) This compares with assumed outflows of 20%-100% for other various categories of institutional or brokered deposits.

Indeed, it may make sense to consider deposits from high-net-worth individuals as more akin to financial/institutional deposits. Much like institutional depositors, these “retail” depositors:

• Lack deposit insurance coverage after $250,000;

• Are more likely to have accounts at multiple banks already established, which hastens transfers;

  • This proved a delay for many firms trying to transfer out of the likes of SVB; they had to first establish an account at another bank.

  • First Republic, meanwhile, got deposit inflows on March 9, 2023, the first day of SVB’s run—suggestive of the overlapping clientele with SVB. (Massive outflows for First Republic, however, followed on March 10.)

• Are likely more attuned to financial news; and

• Related to the previous two bullet points, may be receiving institution-level financial advice from their advisors/brokers.2

  • For instance, regarding First Republic (the main reason we’re talking about high-net-worth depositors): The FDIC inspector general reported that FDIC staff said “several large money center banks began advising their clients to pull their funds from First Republic” on March 10, 2023 (the day regulators seized SVB).3

Thus, it may make sense to think of these individuals less as “less stable” retail depositors and more like institutional ones.

Venture capital and crypto deposits

Notwithstanding that stablecoin deposits have special characteristics that add to financial instability (see, e.g., here), the proposal to specifically target venture capital and crypto firms with higher deposit outflow assumptions seems like overengineering to March 2023—when VC and crypto banks failed.

No doubt, SVB and Signature Bank failed in no small part because of their VC and crypto deposit bases and business models. However, it’s not at all clear that those depositors ran any faster than other institutional depositors would have (and did… see below).

SVB and Signature saw deposit runoff from VC and crypto well prior to their run. SVB’s deposits fell about 12.5% from Q1 to Q4-2022 amid VC cash burn.4 Signature lost a slightly higher percentage in Q4-2022 alone, the quarter when FTX blew up.5 These runoffs were related to the macroeconomic sensitivity of these deposits.6 With the Fed hiking rates swiftly and severely, the cash drain in the economy’s most bubbly sectors was on. The interest rate sensitivity was in their liabilities as much as it was in SVB’s assets (Signature did not have material mark-to-market duration losses).

But, that’s a problem for supervision—or, perhaps in the future, an input into how much collateral these banks should have been required to preposition at the discount window. It’s not what’s captured in the LCR’s focus on a 30-day stress period.

Indeed, as the New York Department of Financial Services noted in its postmortem of Signature: “the percentage of digital asset customer withdrawals on March 10 was relatively proportional to the percentage of digital asset customers in the deposit base overall.” That is, the run by crypto clients was as extreme as for the other 82% of Signature’s (largely uninsured) deposit base.

The NY DFS added that, “The bigger issue for Signature was that the Bank had a high concentration of uninsured deposits and was perceived as a crypto bank” (emphasis added).

It’s not as if mortgage firms ran any faster in 2008 than everyone else. Institutional runs are institutional runs are institutional runs.

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

1

Necessary caveat to this whole discussion: The liquidity solutions to March 2023 can’t be to just ratchet up self-insurance liquidity requirements until every bank is effectively a coffee can. Banks are already substantially less bank-like given post-GFC liquidity regs:

I’m certainly not suggesting undoing those requirements, but we still got the fastest bank run ever. This is where things like giving regulatory “credit” towards (potentially increased!) liquidity requirements for prepositioned discount window collateral can strike the right balance of carrot and stick without disintermediating banks.

The full panel—with BPI’s Bill Nelson, Yale’s Susan McLaughlin, and the ECB’s Luc Laeven—can be viewed here.

I sought to cover a few big-picture issues, which have gotten less focus as we’ve zeroed in on a handful of increasingly consensus technical reforms to the discount window:

1. What should we actually expect from the discount window? Can it be expected to save a bank?

2. The elephant in the room: how to distribute sufficient emergency liquidity when collateral market value has fallen as a result of monetary policy. The BTFP has been characterized as a policy failure of the discount window. But, would a perfect-world discount window really have been able to accomplish what the discount window did? There are parallels to “dash-for-cash” episodes, where market values fall irrespective of monetary policy or credit risk.

3. Collateral values are partly endogenous to the central bank’s financial stability function—certainly for credit risk, but also particularly for banks who derive franchise value from funding themselves more cheaply than implied by mark-to-market pricing. This has implications for crisis facility design.

The speech is also heavily footnoted to hopefully function as a one-stop-shop document and provide further reading to those interested. Link again to the remarks here: https://www.atlantafed.org/-/media/Project/Atlanta/FRBA/Documents/news/conferences/2024/05/19/financial-markets-conference/presentations/kelly-remarks.pdf

Comments also welcome via email (steven.kelly@yale.edu) and Twitter (@StevenKelly49).

Read the note here: https://som.yale.edu/story/2024/weekly-fed-report-still-drives-discount-window-stigma

The Treasury’s normally sleepy “quarterly refunding announcement” on November 1 of last year was beloved by markets. Fear over how much Treasury debt was likely to come on offer at the long end had contributed to huge run-up in long-term rates. Slightly surprising markets, the Treasury signaled on November 1 it would moderate its increase long-end issuance and was comfortable continuing to be slightly above its typically targeted range of 15%-20% of Treasury supply being in bills. The slight tilt towards the short-term end of the curve at least contributed to, if not got too much credit for, the subsequent fall in long-term yields and market rally. See, for example, this graphic from the WSJ:

Basis Loaded

Everyone’s talking about the basis trade again: the BIS, the BoE, the Fed (and Fed researchers), the Financial Stability Oversight Council (which is, uh, all the US regulators), the SEC, the New York Fed, and the Treasury.

As readers know, the basis trade exploits the arbitrage spread between Treasury futures and the cash Treasuries that are “deliverable” into those futures contracts. A hedge fund (typically) performing the basis trade will go long the Treasury bond, and short the corresponding futures contract that is trading relatively richly. And, typically, asset managers take the other side of the futures:

The true size of the trade is unknown, but indicators like the above suggest it’s bigger than its pre-pandemic peak. The short position of “leveraged funds” (typically hedge funds) and the long position of “asset managers” (institutional investors like pension and mutual funds) have approached $800 billion.

Because delivering the Treasury itself upon expiry of the futures contract satisfies the hedge fund’s obligations under the contract, the hedge fund is, well, hedged. It’s free to collect that spread—“basis”—and then deliver the cash asset when the futures contract expires.

A recent presentation of the Treasury Borrowing Advisory Committee (TBAC)—comprised of market participants that advise the Treasury on its borrowing plans and other Treasury market and economic issues—conservatively estimated about 45 bps of annual carry in this trade from 2018 through 2023:

As the TBAC presentation noted, a hedge fund using 20x leverage—which the presentation calls “anecdotally…a good approximation” of leverage in the basis trade—can generate 900+ bps of annual excess return with this seemingly risk-free trade:

(Mind you, the 45 bps estimate from the TBAC presentation was conservative, and a recent Fed note put average hedge fund Treasury repo leverage at 56x.)

Basis Clearing

As readers remember from the start of the covid crisis, the trade isn’t actually risk-free; it can break down before expiry if price moves are severe and margin calls can’t be met without fire sales. This is a particularly acute risk when, as we saw in March 2020, there is a “dash for cash” and the basis diverges: At the outset of the pandemic, rates were headed lower for longer so futures prices rose, but Treasuries nonetheless sold off as investors sought money.

As the recent TBAC presentation showed, even with just 20x leverage, the basis trader’s margin was quickly blown through—exacerbating the dash for cash as hedge funds were forced to sell assets and close positions:

Stolen Basis

The hedgies running this trade might tell us they’re providing a public good because the UST demand from the basis trade improves liquidity and lowers the US cost of financing. Fact check: True. At least statically true anyways; dynamically, we have to consider the costs/risks of the basis trade implosion, but it’s reasonable to expect that, arithmetically, the average cost of financing might be lower (holding the Treasury’s average issuance maturity constant, that is…).

But this trade, funded as it is in the repo market,1 tells the Treasury something else: There’s a ton of structural demand at the short end of the issuance curve. All else equal, the funders of Treasury repo would likely be happier with Treasury bills—no counterparty risk, no collateral price risk, no legal/operational risk with respect taking possession of collateral. (i.e., better money.)

Thus, in addition to providing demand to the Treasury’s long-end issuances, the basis traders are creating a money asset (short-term, Treasury repo) from a long-term asset, helping satisfy institutional money demand on the short end. The Treasury itself is thus missing an opportunity, the size of which can (at least) be proxied by the basis trade, to issue substantially more bills with limited market indigestion—and without the need to rely on the basis trade machine. Indeed, it would crowd out the trade.

And Asset Managers Kick Rocks?

But would short-circuiting the repo/basis loop mean the asset managers on the other side of hedge funds’ futures short are just eating the costs of their futures counterparties disappearing? It looks unlikely.

The recent TBAC presentation to the Treasury on the basis trade analyzed asset managers’ motivations to use so many long Treasury futures. What the TBAC found is that active fixed income managers are stuck benchmarking to aggregate bond market indexes, which have increasing amounts of Treasuries in them.

These active managers, meanwhile, prefer to pick up their extra yield by higher credit concentrations—which, for various reasons (pp. 17-23), have tended towards having lower duration contributions than the index’s Treasuries they’re supplanting. To avoid this leading them to have a duration basis relative to their benchmark, they enter into long Treasury futures.2 Effectively, the asset managers are paying the 45 bps of carry to the hedge funds in exchange for the opportunity to invest in their preferred credit assets without worrying about straying from their benchmark on duration risk—a trade they can easily pick up yield on:

The TBAC presentation shows that as bond market benchmark indexes flood with Treasuries, asset managers tend to just pick up this duration exposure with Treasury futures, rather than reallocating their portfolio away from their lower-duration-risk credit investments:

So, to the extent the Treasury shifts its issuance toward bills, aggregate bond indexes will also have a shorter duration. Thus, bond funds will find that active credit management (which tends toward lower duration contributions) won’t leave the funds short duration relative to the aggregate bond market index—and thus they won’t have such high demand for long futures. (This also means the funds would be free of the implicit leverage in futures contracts!)

Making Money

There is insatiable demand at the short end that the Treasury has seemed unwilling to satisfy. Prior to the GFC, this demand was met with private sector replacements like commercial paper, ABCP, and repo—which of course proved to be poor substitutes. Deficits (and the Fed’s reverse repo facility) have helped solve that problem since, and long-term-Treasury-backed repo is a step safer than MBS-backed repo. But, even Treasury-backed repo trades unwind in crisis when institutions want to move up the moneyness scale.

Treasury prides itself that its debt issuances of the various tenors are “regular and predictable,” so any shift towards a shorter term structure would have to be managed carefully. Though, notably, issuance already shifts shorter during crisis episodes and other shocks, indicative of the limited price elasticity at the short end.

Too, it’d be remiss to not note that the only terming out of debt the Treasury has done in the last 20 years has been at times the Fed was doing QE:

The Treasury was thus offsetting to an extent the Fed’s efforts to pull duration from the market; the Fed’s subsequent unwindings of QE was followed by growth in the basis trade.

And yes, the Treasury aims for its debt issuance patterns to minimize the total cost of borrowing over time. Despite there being much less price sensitivity to front-end supply (particularly given the reallocation potential from Treasury repo to direct bill purchases), the extent to which the Treasury misses out on instances of a negative term premium on longer-term debt implies a theoretical—if likely nominal—cost. The tradeoff here, though, is less Treasury market blowup risk.

The Treasury can also be much less fearful of “rollover risk,” the typical concern about short-term debt, given it issues the world’s safest (and money-est) asset, and practically no one is constrained from buying short-end bills—unlike long-end bonds which may require the machinations of the basis trade.3 As noted, short-term debt already functions as the Treasury’s “shock absorber” exactly because of the market’s ability to absorb short-end supply.

Big Shorten

Financial engineering can provide substantial value, but, in the case of the basis trade, the Treasury doesn’t actually need its help. The market’s demand for institutional money assets exceeds what the Treasury has been issuing. Treasury can skip over the risk-creating basis machinations and issue directly to the ultimate funds-provider for the basis trade: short-term money market investors.

The acclaim Treasury got for its November refunding announcement can be just the beginning.

Related: Refitting the Standing Repo Facility to backstop Treasury trades like the basis trade

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

(FT Alphaville is free to read. Of course free with an FT subscription, but also with just an email address: How to read FT Alphaville for free)