This is the first chapter in a new series about close-out netting, and the tiresome, hideous problem of close-out netting opinions. If you haven’t already, you might want to check out the introductory prologue, already released. It is here. This is chapter one: we are still in scene-setting mode. Before we get started with the really boring stuff it is worth getting down to brass tacks and asking some really basic questions about how why we even have netting, why regulators care so much about it, what difference netting opinions make and why are they so long?
If you thought this was fun, or interesting, please pass it on to someone who you think might like it.
But before getting to that we need to go right down into the elephants and turtles of what banking is all about.
For to understand why we need close-out netting opinions we must understand what close-out netting does. To understand what close-out netting does, we must understand the point— the importance — of capital adequacy. To understand the imperative of capital adequacy, we must have our heads around the profound structural problem faced by every credit institution: “borrowing short and lending long”.
Fundamental intermediation
To understand that problem we need to understand the basic function every bank plays in the economy, of taking — with permission, of course — money from those who have more than they need, and giving it — for a fee — to those who don’t have enough.
This fundamental business of banking is intermediation. Depositors — those with too much money — lend it to the bank, in return for deposit interest. The bank then lends that money — or, at least, a sum equal to most of it, to borrowers — those of its customers who don’t quite have enough. The bank lends to them, also, in return for interest. I don’t imagine this will come as a great surprise to anyone.
As long as the borrowers are as good as their word and pay back what they owe, all is mostly well. If the borrowers can’t pay their money back, then the bank will have a harder time repaying its depositors. That much is obvious. But even if they can, the banks face an embedded structural problem. Generally, banks take in deposits “on call”. They are repayable to depositors on demand: Just stick your card in the ATM and you have have your savings back. But when banks lend money to their customers, they tend to do so on term loans: they can’t ask for their money back when ever they want it. They must wait until the end of the term.
This can present banks with a cashflow problem, especially if they happen to field a large amount of withdrawal requests from depositors at the same time. Usually, this doesn’t happen: a bank’s deposit base tends to be fairly stable, and across the bank inflows and outflows on a given day are more or less balanced. But occasionally circumstances can contrive to change that, and things can get pretty spicy, fast.
We can best explore this by the sad tale of Silicon Valley Bank.
Interlude: The sad tale of Silicon Valley Bank
SVB committed one of the most elementary errors in banking: borrowing money in the short term and investing in the long term.
—Larry H Summers, 14 March 2023.
There are any number of ways to measure the size of a bank: by number of customers, depositors or employees; by branches, by geographic spread or by aggregate value of customer deposits — but the conventional one is total asset value — the aggregate value of all the investments the bank has made — much of it with those deposits.
By that measure, in March 2023 Silicon Valley Bank, though founded only 40 years earlier, was the 16th largest bank in the U.S.
If you measured Silicon Valley Bank by its number of depositors, it wasn’t nearly as big. Whereas JPMorgan has something like 90 million customers, Silicon Valley Bank had about 40 thousand. And whereas the average customer balance at JPMorgan, across business and retail accounts is about $20,000, the average customer deposit balance at Silicon Valley Bank was a shade under $4,000,000.
I can save you a bit of maths here: Silicon Valley Bank was a tenth of JPMorgan’s size, but had one thousandth as many customers. And those customers were not diverse. They were not, like JPMorgan’s customers, a scattered distribution of miscellaneous butchers, bakers and candlestick makers: Silicon Valley Bank’s customers were all techbros. They all knew each other. They were all directionally positioned the same way. They all had lots and lots of spare cash.
If you are JPMorgan, the beauty of your retail banking portfolio is that it is not at all concentrated. There are millions of butchers, bakers and candlestick makers, and they all deposit and withdraw their tiny amounts over the month, based on their own unique needs and circumstances. Because they are, for the most part, independent of each other, the overall effect is to smooth out all the volatility in your overall balances. Your “deposit base” is stable. It’s predictable.
Silicon Valley Bank was very different. It had thousands, and not millions, of customers, and each one deposited a lot. Their deposit balances ran to millions. It was a really concentrated little bank.
What was special about SVB? What was special about its customers? What kind of freak keeps millions of dollars in a checking account? Well, the clue is in the name.
Silicon Valley Bank was founded in 1983 by a couple of Bank of America executives who spotted an opportunity to provide banking services to the then nascent “technology industry”. At the time, “Silicon Valley” was a fun name for a rabble of dope-freaks, dropouts and hippies mucking around with soldering irons and BASIC in Cupertino garages. These cats struggled to get accounts at Wells Fargo and JPMorgan. The new bank met their needs.
Roll forward 40 years and it’s a different story. Silicon Valley Bank is the go-to bank for the VC crowd. Anyone who was anyone banked there. An SVB card was as much a part of the techbro clobber as the black turtlenecks, immortality vitamins, slap-head anarcho-libertarian ideology, sack-hopping altruism and the simulation hypothesis.
But only techbros banked there. Silicon Valley Bank didn’t collect deposits from many butchers, bakers or candlestick makers. It didn’t really want them: when you have queues of unicorn techbros queueing up to stash their cash with you, who has time to onboard the little guys?
As time passed, Silicon Valley Bank steadily “lent in” to its niche: it provided startup advice. It made introductions. It had great VC contacts. It did off-the-cuff in consulting. It was part of the valley ecosystem. Some venture capitalists made banking there a condition of funding.
Techbros are joiner-inner types at the best of times: this was a network effect on steroids.
By 2020 software was eating the world, just like the VC snapperheads said it would, and money was abundant. There was the metaverse, crypto and nascent AI: as long as you had an account at SVB, VCs did not discriminate. They threw money at WeWork, WireCard and Theranos and it was fine. They would happily throw money at you. All you needed was a pitchbook, a hoodie and a compelling story about blockchain.
The name of the game was “first-mover advantage”: move fast and break things. Scale. Lever. Dominate the your space. Build a moat. Cash burn was a badge of honour. If you ran out, you just raised more.
So Silicon Valley Bank’s main customers, the techbros, had lots of cash. Like, lots. In the U.S., retail bank deposits up to $250,000 are government insured against bank failure by the Federal Deposit Insurance Corporation. For regular banks, most of their deposits are FDIC-protected: only 20% of retail deposits exceed the quarter of a million dollar threshold. You’d expect that: who puts quarter of a mill in a savings account?
SVB’s customers, that’s who. Ninety percent of SVB’s deposits were over the $250,000 FDIC threshold. There are two things to take from this: firstly, SVB deposits were stupefyingly big and secondly, they were at risk if SVB should get into trouble. There would be no government bailout.
Everyone at Silicon Valley Bank was depositing cash. Almost no-one was borrowing.
This is not normal for a bank. Borrowing and lending tend to balance out. When money is cheap, people feel good. They borrow to buy stuff, and pay the borrowing costs out of their income. They put things on the credit card. They buy houses this way. Cars. They put things on hire purchase. They put capital into their businesses. For this, they go to the bank. That’s what a bank does: matches depositors and borrowers. Yin and Yang.
But techbros aren’t like normal people. They don’t get money from banks: they get it from venture capitalists. Techbros are — were — structural lenders, not borrowers.
So, Silicon Valley Bank had an odd problem: what to do with all these excess deposits? Between March 2021 and March 2022 — in a single year, in the middle of the Pandemic — SVB’s deposit base doubled. It was being given deposit cash in a near-zero interest rate environment. Its interest costs were low, sure, but it had some, and overheads. Without borrowers, it had to find somewhere to park that cash and some income to pay its bills, return a profit, and yet be able to meet deposits if customers suddenly wanted their money back en masse. Usually, customers don’t do that — retail banking is a boring business — but SVB was special. If there could be a secular inflow in, there could be a secular inflow out.
SVB had a credible answer. It would invest the deposit cash in U.S. government bonds and highly-rated mortgage-backed securities. These were, it thought, a better bet than mortgages and business loans anyway.
The gig for normal banks is to lend that deposit cash to home buyers and businesses. These investments carry credit risk. Businesses can fail. If you’ve lent out on a mortgage, you run the risk your borrower can’t pay you back, and the property collapses in value. Even if it doesn’t, they are long-term commitments. You can’t get your money back until the term expires. And that term might be as long as 30 years.
By contrast, U.S. Treasuries are safe. They are as safe as it gets: Uncle Sam don’t go bust. They are reliable: they pay fixed, regular coupons. They are liquid: you can sell them in a day and get your money back.
For a commercial bank without borrowers, they looked like a sensible hedge. I dare say the managers and executives at Silicon Valley Bank congratulated themselves on their prudence.
Still, short-dated government bonds in a low interest rate environment don’t pay much interest.
This is not the place to talk about yield curves in detail, but to generalise outrageously, the longer you are prepared to invest your money, the higher your interest rate tends to be. The “yield curve” usually slopes up and to the right. Silicon Valley Bank put all its deposit cash in 10-year U.S. treasury notes. All seemed in order.
There was one extra little funny. Silicon Valley Bank was big — but not colossal. It was not a systemically important financial institution, or “Stiffy”. U.S. regulators did not think it was big enough to present a real threat to the financial system.
This meant it avoided certain capital regulations that applied to bigger banks. One was the “liquidity coverage ratio”. This requires a bank to keep a certain portion of its deposit obligations in “high-quality liquid assets” on hand at all times — cash or cash like instruments, which doesn’t include long-dated treasuries – to meet unexpected withdrawal requests. If market conditions suddenly change and customers suddenly all want their money, a bank can use its “liquidity buffer” to ride out the storm.
In 2019, Silicon Valley Bank had lobbied for an exemption from these coverage ratios for banks with less than $250bn in assets. For reasons that, no doubt, U.S. bank regulators would come to regret, they granted it that exemption. As a result, SVB kept very little liquid cash on hand. It ploughed it all into ten-year Treasury notes and highly rated mortgage-backed securities.
Then came COVID. At first it seemed a good thinrg. SVB’s deposits tripled—from $61 billion in early 2020 to $189 billion by 2022. It doubled down on its strategy.
The inflation rate that didn’t bark in the night-time
“Is there any other point to which you would wish to draw my attention?”“To the curious incident of the dog in the night-time.”“The dog did nothing in the night-time.”“That was the curious incident,” remarked Holmes.
— Sir Arthur Conan Doyle, The Adventure of Silver Blaze
All good things must end and the pandemic turned out to be the final straw that broke the zero inflation madness. Ever since the global financial crisis in 2008 governments everywhere had been helicoptering cash onto their bin-fire economies with, seemingly, no adverse inflationary effects.
The idea that low interest rates cause inflation seems baffling for a moment, but makes sense when you think about it. Inflation is prices going up. Prices tend to rise when demand outstrips supply. Demand rises when people spend. People spend when money is cheap. Money is cheap when interest rates are low. Low interest rates tend to push inflation up.
And yet for fifteen years money was cheap. With every new crisis, central banks firehosed their economies with it, but inflation barely moved. No-one had a good explanation. Over time there was a kind of tacit assumption that, somehow the link between cheap money and big inflation was broken.
So when the economic world ground to a juddering halt, once again, in March 2020, central bankers knew the drill. They went to their printing presses as usual.
But this time inflation shot up. This took central banks by surprise. Central bankers have one basic lever to control inflation: interest rates. Just as cheap money boosts inflation, expensive money kills it. Around the world central bankers hiked interest rates. Fast.
Duration mismatch: the fundamental risk of banking
Borrowing short and lending long is not — despite what the former secretary to the U.S. Treasury had to say about it on Twitter — an elementary error, but the very fundamental premise of banking. Banks cannot avoid it. That is what they do. By its fundamental nature, it is that exact risk — “duration mismatch” — that the service of “banking” presents to those who engage in it.
To see why we need to go one layer of turtles deeper, readers.
Banking is the business of reallocating capital. Banks borrow money from those in the community who don’t need it and lend it to those who do. A bank hoovers up all those for-now-unneeded deposits, loafing around in savings and cheque accounts, consolidates then and then lends them out, usually in sizeable slugs and for sizeable terms, to those who need investment.
In their most simplistic terms, banks intermediate between lenders — “depositors” — and borrowers.
The priorities of depositors are different to those of borrowers.
Depositors give up comparatively small balances they don’t need right now in return for interest, but on condition that they can have the money back whenever they want it. Monthly wage packets come in and then get salted away over the month. Generally speaking, a bank is obliged to return deposits on demand, and a failure to do so is a payment default.
In the normal run of things, banks are quite good at predicting when and how much money their depositors, en masse, will need to withdraw. Their “deposit base” tends to be stable over a long period. This means that banks can, with reasonable confidence, use a large portion of their deposit base to make long-term investments.
Which is just as well, because “long term money” is what bank borrowers want. They want to lock up the money they borrow. They want committed bank facilities to buy property and invest in working capital. These projects have a much longer term: mortgages may be twenty or thirty year arrangements. Borrowers need to know the money they borrow, and the price they have to pay for it, is committed.
So: banks borrow from depositors “short” — overnight, or on call — and lend to their customers “long” — for pre-agreed terms.
This creates what lawyers call “duration risk”. If all the depositors suddenly ask for their money back at once, and it is tied up for twenty years, the bank is “cashflow insolvent”.
In the ordinary course, banks manages these risks by diversifying: they have millions of depositors and tens of thousands of borrowers, and the behaviours and repayment schedules of both are predictable enough for the bank to manage its cashflows notwithstanding the duration mismatch, as long as it maintains a sensible “liquidity buffer” of free cash to cater for unexpected spikes in withdrawals.
If it’s that easy, then how does anyone get it wrong? Every now and then, banks do. Silicon Valley Bank did, badly.
On time being money
Hey, Rick? Time is money, right? I know “standing on the landing” may be a great song title, to me it’s just a tax loss.
—Mike, The Young Ones, 1984.
The general ideal of banking is to earn more interest on loans that you pay on deposits. This is, in essence, how banks make money. Deposits usually pay an interest calculated on an overnight rate. Bank loans are usually fixed for much longer periods.
While this creates “duration risk”, as we’ve seen, we’ve also seen that in ordinary times it can be managed: banks run a “bid/offer spread” between lending and borrowing rates for one thing, and for another, at any time long-term interest rates, at which they lend, tend to be higher than short term ones, at which they tae deposits. The interest rate “yield curve” at any time usually slopes up and to the right. Therefore, structurally, they should be profitable most of the time.
This not always true. And the whole yield curve can move upwards. Today’s overnight rate could be higher than last month’s ten year rate. This could create a problem: a bank could be stuck with a portfolio of low-interest investments which it has to fund with high-interest deposits. If the overnight deposit rate unexpectedly spikes, a bank will have to raise its deposit rates, to stop its borrowers shopping around for a better rate, but will be stuck with its lending rates as they are. This can create, at least in the short run, a problem for the bank, unless it has enough cash on hand to meet the interest shortfall until is sorts itself out.
Things would be even worse if the interest rate spike was accompanied by a surge in withdrawal requests.
At this point let’s go back to Silicon Valley Bank. It is March 2023.
Second interlude: the sad demise of Silicon Valley Bank
Meanwhile, in Silicon Valley
Meanwhile, in Silicon Valley, that weird low-inflation, low-interest-rate, cheap-money environment — which for a decade had prompted cone-headed VCs to swamp money at anyone with a pitchbook and a turtle-neck — ground to the same juddering halt. The funding dried up. No more Series D rounds.
Everything still might have been okay for Silicon Valley Bank had it been able to sit on its low-yielding treasury securities, as it intended to, until they matured, riding out the temporary blip in interest rates. But that depended on depositors being cool with leaving their money in the bank.
That, too, might have happened, had it been JPMorgan and its customers a widely distributed, unconnected bunch of butchers and bakers with chump change in their checking accounts.
But Silicon Valley Bank’s customers were not like that: they were homogenous, tightly interconnected and weird — as in “western, educated, industrialised, rich, and democratic” weird, but also as in “freaky tech-bro taking pills to be immortal” weird — and all of them had unfeasibly large balances that weren’t covered by FDIC insurance.
Techbros suddenly found they needed to raid their bank accounts to make payroll. And they wanted their money, in any case somewhere a bit safer.
Silicon Valley Bank started shipping lots of withdrawal requests. It had lots of liquid US Treasuries it could sell, if it needed to, to fund withdrawals. No great panic, except for one thing: being long-dated fixed rate bonds, their stated interest rate was low. Their “present value” was also significantly under water.
To fund the withdrawals Silicon Valley Bank had to sell part of its U.S. government bond portfolio at a $1.8BN loss.
It announced plans to raise capital in the market to shore up its balance sheet and reassure investors — but that announcement had exactly the opposite effect.
All hell broke loose. All those techbros scrambled to pull their money out.
On 9 March 2023 Silicon Valley Bank suffered a bank run. The next day regulators shut it down.
Coda: call me “Lucky”
As I sit, staring into this liquid amberRipples move out to the edge of the glass.Is that, really, your reflection in there?I just want to jump into the warm depthsAnd be there with youOne more time.Oh, alright.Hit it, boys —
—Blondie, Here’s Looking At You (1981)
Silicon Valley Bank’s sudden collapse is illustration enough of the non-linear, hard-to-predict, tightly-coupled nature of complex financial markets. But it wasn’t done. Ten days later, half a world away, there would a final twist.
Ripples were moving out to the edge of the glass. On the 12th of March, Signature Bank — another darling of the tech crowd, with a surprisingly similar idiosyncrasy — also failed. Fear spread. The credit market had some kind of acid flashback to 2008. Was this the start of another banking crisis? Eyes fell on the weaker banks around the world.
Attention fell on one. The old warhorse Credit Suisse — battle-scarred, woozy, punch-drunk and unsteady on its feet after repeated drubbings — finally hit the canvas. Granite-chinned old “Lucky” had taken a beating in pretty much every catastrophe of the prior decade: Archegos, WireCard, 1MDB, Mozambique Tuna bonds, Greensill, Abraaj, Madoff, to name a few and got up every time.
This time, for once, the blame lay elsewhere. But this time, the old dog wouldn’t get up.
The social contagion problem
Hamlet: Why, then, ’tis none to you, for there isnothing either good or bad but thinking makes itso. To me, it is a prison.
—Hamlet, II, ii
Fulfilling the important community role that they do, banks are vulnerable to social contagion. If depositors form the opinion that they might not repay their deposits, they are likely to withdraw them. Depositors — even unweird ones — talk. They listen to the radio and read the newspaper. They are on Twitter. Word can spread quickly. Very soon there can be a line of depositors queuing up outside the bank.
This is inevitably a bad thing, not just for the bank, but also the depositors. No-one wins out of a bank run.
For the longest time, bank regulators have required banks to organise their balance sheets to reduce the risk of duration mismatches — and reduce the risk of customers being worried there might be a duration mismatch, which is just as important.
As long as they aren’t worried about duration mismatch the “ordinary” balance of loans and deposits will tend to be fairly stable.
A very simple way which banks, and bank regulators, can reduce this risk is by requiring banks to hold a certain amount of cash in reserve. If a bank keeps some of that deposit cash “on its balance sheet” to cover for unexpected outflows when horrible things happen. One horrible thing that can happen is the failure of a derivative counterparty. Counterparties can have surprisingly large “exposures” to single swap counterparties — see the series on Archegos — so they need to have money on hand to cover that contingency.
But how much? How would you calculate it?
It is one thing to be owed money by a bankrupt counterparty — that’s the oldest problem in banking — but what if you say your counterparty owes you, but its insolvency administrator looks through your “single agreement” to all the transactions, and tells you that you you owe the bankrupt counterparty?
That would be bad, right?
And what if it said it added up all your “out of the money” transactions, and you owed it twenty times what you were claiming from it?
This is the risk that close-out netting addresses. We will look it closer in the next instalment.
The music
The audio version of this piece makes lots of extra in-jokes (if you enjoyed this please do listen to it — it took ages to edit!) and to do that borrows some of JC’s favourite tracks. So you can listen to them properly, and the artists are suitably remunerated (thanks guys!) I have put a playlist together. Here it is:
