A fractional reserve crisis

The crisis that has engulfed crypto in the last year is a crisis of fractional reserve banking. Silvergate Bank and Signature Bank NY were fractional reserve banks. So too were Celsius Network, Voyager, BlockFi, Babel Finance and FTX. And still standing are the crypto fractional reserve banks Coinbase, Gemini, Binance, Nexo, MakerDAO, Tether, Circle, and, I would argue, every one of the DeFi staking pools. All of these are doing some variety of fractional reserve banking. Custodia Bank and Kraken Finance claim to be full-reserve banks – but 100% reserve backing for deposits is both hard to prove and not a guarantee of safety.

What do I mean by “fractional reserve banking”? My definition might surprise you. For me, fractional reserve banking simply means that the composition of a bank’s assets is less liquid than that of its liabilities.

Fractional reserve banking is not dead

The well-known fractional reserve banking model taught to economics students proposes that: 

• banks keep 10% of deposits in reserve and lend out the rest (no they don’t)
• for every $10 a bank receives in deposits it can create $90 of new loans (no, this is not how bank lending works)
• every $1 of bank reserves becomes $10 of bank money (no it doesn’t).

This was never more than a toy model, a wildly over-simplified description of banking that bears little resemblance to what banks actually do. It conflates lending with payments and thus creates a wholly fictional relationship between lending and reserves. And after a decade of QE, the relevance of this model of fractional reserve banking, based as it is on the mistaken notion that banks “multiply up” their reserves by some percentage set by the central bank, is extremely doubtful. 

But though the model may be past its sell-by date, this does not mean the concept of “fractional reserve” is dead. Far from it. The current crisis has once again brought to the fore the fundamental risk of fractional reserve banking – namely that banks, and other financial institutions doing bank-like things, can literally run out of money.

Classically, a fractionally reserved bank has demand deposits funding a portfolio of long-dated commercial and household loans, much of it secured on real estate. In the Diamond-Dybvig model of bank runs, when depositors all pull their deposits at the same time, the bank is assumed to foreclose loans to obtain sufficient money to pay them. But loans are legal contracts: unless there is a clause in the contract permitting instantaneous no-fault termination, they can’t simply be foreclosed without warning. So a bank whose assets consist entirely of illiquid loans is at serious risk of defaulting on its obligations to its depositors. 

In practice, banks keep a certain amount of liquid assets to ensure they can meet a reasonable level of payment requests. The rest is – frankly - a Hail Mary, though for regulated banks, deposit insurance and central bank lender-of-last-resort support mitigate the risk of payment requests exceeding available liquidity. 

The quantity of liquid assets banks maintain to meet normal payment requests used to be determined by the reserve requirement. But in these days of excess reserves, Basel liquidity regulations, and "living wills", few central banks now have reserve requirements. More importantly, central banks now recognise that the primary purpose of reserve requirements and other liquidity regulations is not to control lending, but to ensure that banks have enough liquidity to make payments. 

Reserves are the essential liquidity needed to ensure the smooth operation of the payments system. They help to ensure the independence of outgoing payments from incoming ones. We could call them "float". 

For regulated banks with central bank reserve accounts, reserves are the balances in those reserve accounts – what we call “high-powered money” or “outside money”. For other financial institutions, reserves come in various forms, but they are always about ensuring the institution has the money it needs to make payments in full and on time. So, a British credit union that pre-funds its clearing account at Barclays is holding sterling reserves. A Caribbean bank that pre-funds its US dollar nostro account at Citibank is holding US dollar reserves. And a crypto exchange that keeps U.S. dollars in omnibus accounts at U.S. banks “for the benefit of” its customers is holding dollar reserves.

Holding some funds “in reserve” in anticipation of payments helps to maintain smooth operation of the payments system. But if financial institutions try to hold reserves in anticipation of payments that may never be made, the system becomes illiquid. We saw this happen in September 2019, when the U.S's large banks hoarded reserves in anticipation of very large payment requests from corporations and the U.S. government. As the banks stopped lending liquidity, the cost of market funding shot up and repo markets froze. The Fed was forced to step in to provide liquidity and thus maintain the flow of funds through the financial system.  

Some payments systems are fully pre-funded. The US correspondent banking system, for example. And here I must give a shout-out to Ripple, which says that locking up dollars in nostro accounts is inefficient. I don’t agree with their solution, but they are absolutely right. Dollar hoarding makes the system sticky and expensive. The only reason we don't notice this is that we've never known anything else. 

However, if foreign banks, small banks and non-banks keep reserves in banks that are themselves fractionally reserved, they cannot credibly claim to be fully reserved. The stablecoin issuer Circle painfully discovered this when Silicon Valley Bank failed. Its supposedly “safe” stablecoin, USDC, fell off its peg when it was revealed that Circle held some of its reserves in SVB.

Fractional reserve, solvency and creditworthiness

If a financial institution is “fractionally reserved”, it does not have enough actual money to pay out all its deposits at once. This does not necessarily mean it is insolvent. What it usually means is that there is a liquidity mismatch between its deposits and its assets. If deposits are wholly liquid, and assets wholly illiquid, the institution may be unable to meet its obligations even if it is in balance sheet terms solvent, by which I mean the value of its assets exceeds its liabilities. Silvergate, SBNY and Silicon Valley Bank were all, on paper, solvent, but were nevertheless unable to raise the money they needed to pay depositors.

How can an ostensibly solvent institution become unable to pay its way? It’s all down to creditworthiness. As long as a fractionally reserved financial institution can borrow, it can meet its obligations. But if no-one will lend to it, then it cannot. And in these days of collateralised lending, creditworthiness is all about the quality of your assets.

A bank can’t borrow more than the market value of the securities it is pawning (lending cash against valuable collateral is pawnbroking). And if the assets are risky, the amount it can borrow will be much less. I am constantly amused by crypto people who talk glowingly about “overcollateralization” as if this is something new, innovative and exclusive to crypto. Tradfi has been doing overcollateralization for hundreds of years. But it prefers reciprocals. What crypto people call overcollateralization, tradfi calls a “haircut”.

If a bank’s balance sheet consists mainly of degraded loans and junk securities, it’s going to face large haircuts on collateralised borrowing. This applies whether the institution is borrowing from markets, from other financial institutions, or from central banks. There’s an apocryphal story of a high street bank offering the Bank of England its staff travel loan scheme as collateral against borrowing. “We will accept it,” said the Bank, “but you won’t like the haircut.” But even a staff loan scheme is better collateral than highly volatile cryptocurrencies and stablecoins. In its guidance on the prudential treatment of crypto-asset exposures for banks, the Bank for International Settlements cheerfully imposed a 100% haircut on cryptocurrency collateral. That is infinite overcollateralisation. (You see why tradfi prefers reciprocals?).

But even if a bank’s balance sheet has substantial quantities of “safe” assets such as government bonds, falling bond prices can reduce its borrowing capacity to the point where it cannot raise enough money to meet its obligations. This is what happened in the Eurozone crisis: the so-called “doom loop” between sovereigns and banks, which meant that whenever a sovereign’s bonds fell in value (usually because of a Moody’s downgrade), its banks ran out of money. And it is also what happened to Silvergate Bank, Silicon Valley Bank, and SBNY, and to other US medium-size banks too. These banks were solvent on paper because they were not marking their securities portfolios to market. But because of the Fed’s interest rate rises, they were taking heavy unrealised losses on those securities. When they had to sell those securities to obtain money, the losses were crystallised. And for both SBNY and SVB the market value of the securities was insufficient to raise enough money to meet their obligations. Having to pledge securities at market value rendered them insolvent. This is why the Fed’s new lending program for banks accepts securities pledged at face value, not market value. It is a “lender of last resort” intervention to ensure that fractionally reserved banks don’t fail because Fed interest rate rises make it impossible for them to meet their obligations.

It's not just banks that are fractionally reserved 

Taken to its logical conclusion, “fractionally reserved” can be taken to mean any balance sheet structure in which the average duration of liabilities is shorter than the average duration of assets. This is known as “maturity transformation”, and is characteristic of banks. However, it is an extremely broad definition which applies not only to banks, but many other types of financial institution. Money market funds, for example. When Reserve Primary MMMF “broke the buck” after the fall of Lehman Brothers in 2008, it was experiencing a fractional reserve crisis. Its shareholders were selling shares, but its assets were falling in value, eventually rendering it unable to obtain the money needed to pay the shareholders par value. In the last decade we also saw real estate investment funds suspending withdrawals when falling commercial property prices created a large gap between the value of their assets and their liabilities. And the bellwether for the current crisis was, of all things, defined-benefit pension funds.

Defined-benefit pension funds should be the last things on earth to experience a fractional-reserve crisis, except possibly the Danish mortgage market. They have long-dated liabilities matched with long-dated assets. They are about as liquid as the Gobi desert, but their payment obligations are stable and forecastable. Yet last September, they were suddenly faced with demands for large payments for which they were wholly unprepared. They had to sell assets to raise the money to make the payments. The assets they sold were UK 30-year gilts, of which they were almost the only holders. As a result, the 30-year gilts market crashed. As the price fell, the funds were forced to sell even more assets. This created a “death spiral”, a dangerous positive feedback loop that threatened to spill over to other markets too and even to the real economy. The Bank of England stepped in as buyer of last resort, setting a floor under the 30-year gilt price and thus enabling pension funds to raise the cash they needed. Crypto aficionados perhaps should note that setting a floor under a falling asset price can stop it becoming a death spiral. At the macro level, QE prevents economy-wide deflationary death spirals by setting a floor under asset prices.

But why did the pension funds need to raise money? It turned out that they hadn’t been fully matching their liabilities with assets of similar duration. Instead, they had been relying on interest rate swaps to cover their deficits – in effect, betting on interest rates staying low. They obtained cash collateral to post against these swaps by pledging long-term assets as collateral for short-term borrowing. They knew they were at risk of margin calls if the value of their collateral fell. But it was worth it for the potential returns.

Deliberately taking on asset-liability duration mismatch, and the associated liquidity risk, to earn a return is the essence of fractional reserve banking. Pension funds are not banks, but they were taking bank-like risks. And the disruption their crisis caused threatened to trigger a wider financial crisis. Hence the Bank of England’s intervention – and, eventually, the UK government’s change of fiscal direction. Pension funds were, on this occasion, “shadow banks”.

The return of shadow banking

We’ve come to associate the term “shadow bank” with the special purpose vehicles, money market funds, insurance companies and the rest of the alphabet soup that played such a large part in the 2008 financial crisis. Lehman Brothers was a shadow bank – an unregulated financial institution that did bank-like things and took bank-like risks. Its failure caused a world-wide financial and economic crisis. Other giant banks that are now household names were also “shadow banks” at that time; Goldman Sachs, for example. After the crisis, these too-big-to-fail institutions converted themselves into banks and submitted to regulation and supervision so that they could gain access to public safety nets.

But where there are shadows, there is always shadow banking. And few areas of the financial system are as shadowy as crypto. So it should come as no surprise to discover a massive outbreak of shadow banking in the crypto ecosystem. Of the crypto-related companies I listed at the start, only Silvergate and SBNY were regulated banks with central bank reserve accounts. Kraken and Custodia are regulated banks, but as the Federal Reserve has rejected their applications for reserve accounts, they must rely on other U.S. banks for dollar clearing, and they probably have to pre-fund their clearing accounts (which they don’t want to do). The rest of the companies do – or did - bank-like things but are not regulated like banks and do not have access to the public safety nets provided to regulated banks. They are, or were, “shadow banks”.

Things that grow in the shadows go unnoticed until it is too late. Pension regulators failed to spot the risk posed by duration mismatches and illiquidity in defined-benefit pension funds. Bank regulators failed to spot the risk posed by duration mismatches and illiquidity in U.S. mid-tier banks such as Silvergate, Signature Bank NY, and Silicon Valley Bank. And crypto regulators – to the extent that they exist at all – failed to spot the gaping balance sheet holes and catastrophic shortage of liquidity in crypto shadow banks such as Celsius, Voyager, BlockFI and FTX. In the 2000s, unregulated financial institutions blew up a giant bubble in mortgage-backed securities and derivatives. In the 2020s, unregulated crypto institutions blew up a giant bubble of crypto-related securities and derivatives. It’s the same old jive.

And it’s the same old snake oil, too. Turns out, if you put “Fully reserved, unlike banks” or “Unbank yourself” on a crypto shadow bank, people rush to lend it money, believing it will deliver high returns without risk - especially if the shadow bank sports an FDIC logo to which it is not, and never will be, entitled. Convincing people their funds are completely safe even though you are lending them out to hedge funds and ponzi schemes and have no insurance is a helluvan earner. And if you are really good, you’ll build yourself an escape hatch so when the scheme blows up, it is only your customers that get burned.

Now, of course, there’s a flurry of regulatory activity “to protect the public”. New rules for pension funds and their advisors; new rules for bank capital and liquidity; and above all, a raft of new rules to bring the crypto industry to heel. The stable doors are not only being shut, but reinforced, and an array of shiny new padlocks is being installed. But the horses have disappeared into the shadows, where they are already building the next generation of shadow banks.

Related reading:

Proof of reserves is proof of nothing

Anatomy of a bank run

Rediscovering old economic models

After FTX: Explaining the difference between liquidity and solvency - Coindesk

Caveat Depositor: How safe is your stablecoin? - Coindesk

Photo courtesy of Brian Lucey.