Banca Valle del SiliconeNon era Lorenzo MediciIo ho cedole fissateTu le hai variateVogliamo la coperturaFissiamo la procedura
When we left off last time Silicon Valley Bank had collapsed. It burned through its “available for sale” portfolio of financial assets — mainly long dated U.S. treasury notes — to meet the surge in withdrawals it suffered when, in early 2023, the U.S. Federal Reserve — the central bankers’ central bank — hiked interest rates to tame inflation, and that caused the cheap money venture capitalists had been hosing at anyone with a pitchbook, a turtleneck and a compelling blockchain story, to suddenly dried up.
Suddenly all the bank’s customers needed their deposits at the same time.
That’s the deal with deposits: if customers want their money back, you must give it to them. If you don’t have cash on hand, you must liquidate assets. Silicon Valley Bank had ploughed its deposits into safe, long-dated, government-guaranteed bonds — good — but paying what were now uncomfortably low fixed interest rates — bad.
It could only sell them at a significant loss. A fixed rate bond’s “discount” to its redemption value is a function of two things: how far its coupon sits below current rates, and how long remains to maturity. The bank’s portfolio was well below market rates and, on average, very long-dated. This added up to a whopping “discount”.
You might have questions about that. For one, seeing as “borrowing short and lending long” has been the basic risk of banking since the Medicis, aren’t there tools for managing that risk? Couldn’t Silicon Valley Bank have, like, hedged themselves somehow, to avoid falling down this manhole?
Secondly, war stories are all well and good, but what do they have to do with close-out netting on an ISDA? Silicon Valley Bank was just a normal commercial bank, wasn’t it? Did it even have a swap portfolio?
Those are good questions and they are related.
Duration risk is something the Medicis would have understood and dealt with. It is the ultimate “known unknown”.
And there are things you can do to manage it. In fact, until 2022, Silicon Valley Bank did them: it hedged with “over-the-counter” contracts under which it would pay sums broadly equal to its fixed rate bond coupons, and receive floating rate coupons back. The bank announced its strategy in 2021, expecting rate rises from the Federal Reserve. As the market priced in rate rises, those contracts became more profitable.
This Substack is reader-supported. To receive new posts and support my work, consider becoming a free or paid subscriber.
These hedging contracts were, of course, swaps.
So, Silicon Valley Bank did have a swap portfolio. A big one. By the end of 2021, it was $15.2 billion in notional amount.
But over the course of 2022, Silicon Valley Bank risk managers believed rates had peaked. They systematically unwound almost all their interest rate swaps to realise mark-to-market profits.
This locked in one-off gains, but left the bond portfolio exposed to further rate rises. If, as the bank expected them to, rates eased the bank would realise even more profit. But if they rose further, the bank would be in a jam.
It is easy to be wise in hindsight. In Q2 2022, the Fed was only getting started with its rate rises. You can see the timing of SVB’s hedge unwinds below. It is the region in red.
So swaps did not cause Silicon Valley Bank’s downfall. To the contrary, the lack of swaps contributed by removing the main buffer that could have absorbed the coming rate shock.
You could say the lack of swaps was the proverbial “curious incident in the night time”.
Hedging interest rates: the boy and his shadow
She returned to the nursery, and found Nana with something in her mouth, which proved to be the boy’s shadow. As he leapt at the window Nana had closed it quickly, too late to catch him, but his shadow had not had time to get out; slam went the window and snapped it off.
You may be sure Mrs. Darling examined the shadow carefully, but it was quite the ordinary kind.
— J.M. Barrie, Peter and Wendy (1904)
This may sound trite, but it is still profound: interest is a consequence of lending. Interest doesn’t — until recently, couldn’t — exist without a corresponding disposition of principal from a lender to a borrower. Interest is to a loan as a shadow is to a boy: term loans cast one sort of shadow —fixed — and deposits cast another — floating — but both depend fundamentally on indebtedness. No Peter, no shadow. No loan, no interest.
Duration risk is the problem of trying to match fixed assets that throw one kind of shadow with callable liabilities that throw another.
If you could sell a fixed rate shadow, and buy a floating rate shadow, you would be covered. But how do you trade rates without borrowing and lending the principal sums they are derived from?
You might, if you lent fixed and borrowed floating over identical terms and for identical principal amounts to and from the same person: then the principal payments would offset perfectly. On the same day you borrow $15.2bn on a fixed rate, you lend $15.2bn on a floating rate, on terms that they will both be repayable simultaneously. No money needs to change hands.
This is an excellent solution: principal liabilities wash out to zero: the parties are paying and repaying identical amounts to each other on the same date — leaving only the residual interest rate exposures. The shadows remain; the boys cancel out.
But, alas, banks were left with a formal problem, occasioned by the way defaulting companies’ assets and liabilities are usually resolved in bankruptcy: it there were two standalone loan contracts one borrowed, one lent, each supported by valuable consideration and standing on its own as a sensible economic transaction entered at arms’-length, they would be treated for all practical purposes as distinct: one would be an asset and the other a liability, and they would sit on opposite sides of the balance sheet.
An insolvency administrator would naturally regard them as different things, having different consequences on the defaulting party’s bankruptcy.
A bankrupt’s liabilities are, of course, subject to recovery and resolution contingencies of the bankruptcy process. A creditor might get some or all of its money back; but it might not.
A bankrupt’s assets, on the other hand — its claims against its own debtors — are not. They are undiminished. Indeed, the more the bankrupt recovers from its debtors, the more it will have to repay its creditors.
So the insolvency process — which is there to protect creditors — would naturally prise apart loans owed by the bankrupt from those owed to it, even if certain pairs of loans were entered together and with a unitary purpose. An offsetting lender/borrower would have to repay one, but file a creditors’ claim in bankruptcy for the other.
This rather mucks up the clever idea of cancelling out the two “boys” and being left with their offsetting “shadows”. A bank counterparty would have to hold regulatory capital against its whole loan asset. It would not getting any relief for the value of its offsetting loan liability.
In 1981, bankers at Salomon Brothers — who have since passed into JC mythology as the First Men — had two clients, IBM and the World Bank with exactly this problem, only in the converse to each other.
The First Men had a bright idea. They called it the “swap”.
Swaps
“I wonder,” Mr. Darling said thoughtfully, “I wonder.” It was an opportunity, his wife felt, for telling him about the boy. At first he pooh-poohed the story, but he became thoughtful when she showed him the shadow.
“It is nobody I know,” he said, examining it carefully, “but he does look a scoundrel.”
— J.M. Barrie, Peter and Wendy (1904)
Swaps are a really neat idea. They solve that duration risk — the one that has been with us since Lorenzo Medici — in a novel way without blowing up the balance sheet the way offsetting loans would.
The essential insight of a swap is twofold: first, find a single counterparty who has the opposite duration problem to you, and secondly, remove the principal exchange altogether.
Now you can both hedge your position. Neither of you needs to needlessly inflate your balance sheet.
Economically, an interest rate swap is identical to offsetting loans with the same person. The difference is a formal, but important, legal description: rather than the principal exchange being resolved by set-off, it is removed altogether. A swap is in no sense two inverse transactions: it is a single unified Transaction with no principal exchange at all.
There is therefore no claim for an insolvency practitioner to make for a “return of principal”. The “boys” are gone. We are left with only their “shadows”. The present value of the offsetting payment streams — the shadows — will fluctuate, but they will always offset each other.
So, is this where we talk, finally, about netting, of those offsetting rate flows, and implied exchanges of principal? Is that why we need close-out netting opinions? To reinforce the argument, somehow, that a swap should not be recharacterised as a pair of disguised, offsetting loans?
No.
That bit is settled: as there is no principal exchange inside a swap, there is nothing for insolvency administrators to claim. Unlike the “offsetting loans” structure, the individual “legs” of a swap do not make economic sense by themselves. You cannot view them in isolation. Rather, the fixed leg is consideration for the floating one, and vice versa. You don’t need enforceability opinions at this level, because here there are no cherries to pick.
The “cherry-picking” problem arises only when we try to boil down exposures between different swap Transactions between the same parties under the same ISDA.
The Single Agreement
In most financial instruments, the interest is pinned permanently to a disposal of principal. It is therefore always clear who is the creditor and who is the debtor, and they don’t change.
Loans, for a bank, are always assets: once drawn, the bank is always creditor and the borrower always debtor. There are no circumstances — not even negative interest rates — in which a loan asset could turn into a debt liability.
In the same way, a bank deposit is always a bank liability. A depositor, as depositor, never owes the bank.
Swaps are genuinely bi-directional: they can be either assets or liabilities. You cannot predict who will owe whom at any point in the future. If rates fall, the floating rate payer will be “in-the-money”. If they rise, the fixed rate payer will.
This presents a challenge, or an opportunity, when it comes to aggregating swap exposures under an agreement. Some will be positive, and some will be negative. Ideally, the counterparties should want them to offset.
And now, finally, we can talk about close-out netting opinions. This is the offset they address.
In its day the ISDA Master Agreement was a revolution. The printed form has all kinds of arcane and clever tricks but the “single agreement” concept, set out in elegant and simple language in Section 1(c), is its masterstroke.
(c) Single Agreement. All Transactions are entered into in reliance on the fact that this Master Agreement and all Confirmations form a single agreement between the parties (collectively referred to as this Agreement), and the parties would not otherwise enter into any Transactions.
What this does is to claim the same status for all Transactions as the “swap” concept did for offsetting loans: to say that each of these Transaction “Exposures” is not an independent asset or liability, but is entered into in reliance on the agreement that all Transactions are to be summed down to a greater whole, being the total aggregate exposure under the whole ISDA. There is to be no “cherry picking”.
If this was as far as it went, it might be a bit lame: this does not, of itself, remove a “central offset” to render the separate Transactions co-dependent, the way cancelling the principal exchange converted two offsetting loans into a single swap. It just says, hey, we go into this on the understanding cherry-picking is not allowed, okay? Insolvency laws have a habit of overriding inconvenient contractual provisions. Bankruptcy is an alternative universe: normal rules to not apply. See “Bankruptcy is a phase transition”.
But the ISDA has a second trick that goes some way to achieving that effect. It is buried in Section 6(c)(ii):
(ii) Upon the occurrence or effective designation of an Early Termination Date, no further payments or deliveries under Section 2(a)(i) or 9(h)(i) in respect of the Terminated Transactions will be required to be made, but without prejudice to the other provisions of this Agreement. The amount, if any, payable in respect of an Early Termination Date will be determined pursuant to Sections 6(e) and 9(h)(ii).
See what’s happening there? Can you see the magic?
Kind of buried, right? What is is saying is that upon close-out, the Terminated Transactions are just terminated and nothing is required to be paid under them. The Transactions just vanish in a puff of smoke.
The Close-out Amounts are mystically conveyed somehow to the main Master Agreement, where they are determined not as final amounts due under the Transactions — by this stage, the Transactions have vanished, remember — but as calculation inputs to an amount due under the ISDA Master Agreement itself.
The intention here is to defeat arguments that, “for this set of out-of-the-money exposures, you are an unsecured creditor in the queue, but for this set of in-the-money exposures, you must pay up right here, right now.” Instead, all those amounts are treated simply as inputs to a single termination calculation for the whole agreement due under the “architectural” Master Agreement, not the individual Transaction. This, by the way, is why you cannot unilaterally terminate a Master Agreement — you need it, at all times, for this purpose.
That is where we leave off this episode — we will return next week to look at the mechanics of netting.
Thanks for reading! This post is public so feel free to share it.
