Computing power being what it was in the eighties, still the financial system was at the leading edge of technological prowess. Big money involved, huge money, had meant the ever-expanding global monetary processing footprint had itself become big business. Not just large volumes of real economy payments to process and clear, also increasingly large volumes of securities transactions. During the decade before, the seventies, both kinds had progressively moved beyond national borders, too.
One day in 1985, a computer system at the Bank of New York (BONY) found that it couldn’t handle a large increase in daily volume. Amounting to 32,000 transactions, a flaw in the software had meant that on November 20, this major “clearing” bank had rapidly run out of money. The fact that the firm “cleared” trades for US Treasury securities meant that it matched buyers and sellers in the secondary market, sending payments – from its own account – out to the sellers during each day while being repaid for deliveries from the buyers.
Each one settled via Fedwire.
This particular network processing problem, however, had meant that while BONY had been paying out funds to Treasury security sellers, it couldn’t figure out how to redeliver the bonds; cash had gone out, but little, very little, was coming back in.
The reason for this is Fedwire’s demand for DVP, that is, delivery-versus-payment whereby as soon as a trade order is received the security’s seller is immediately credited the funds from out of the agent’s account; the agent in this case BONY. The result is what’s called a daylight overdraft; as the name implies, it’s a form of unofficial intraday, interbank credit.
Because BONY wasn’t able to process incoming cash demands from buyers (while also not being able to redeliver the UST’s customers believed they had bought), the bank was, in essence, providing credit to them, paying sellers on their behalf – with funds it didn’t actually have available. And it was receiving intraday credit, unbeknownst to central bankers, from the Federal Reserve’s New York branch (FRBNY).
Thirty-two thousand trades piled up, rising to about $20 billion overdrawn by 1415 (ET) that afternoon. Since these transactions are settled by Fedwire, the Federal Reserve Bank of New York was expecting BONY to pay it back in a timely fashion – otherwise, the unofficial daylight overdraft becomes something officially more serious.
That didn’t happen, instead, struggling mightily to retrace those thousands of trades which the software glitch had somehow actually damaged its database, day turned to night so that by 2030 (ET) the overdraft topped $30 billion when BONY staffers working furiously began to settle at least a small trickle of redeliveries.
Still, FRBNY had to extend special circumstances including pushing back the Fedwire closing to 230 (ET) the following morning; and even then, BONY still ended up short by an enormous $22.6 billion. With no choice, that was the amount the bank was forced to borrow, on the books, from the Discount Window – a record-setting draw that was equivalent to one and one-half times the size of BONY’s entire balance sheet (on top of $1 billion in overdrafts left unsettled).
Interestingly enough, the huge loan could only have come about since, due to the same computer error, BONY had in its (temporary) possession about $23 billion in UST’s that hadn’t yet been redelivered to waiting customers (Discount Window loans must be collateralized).
What might have happened had these trades been for, say, junk bonds? It was 1985, after all.
Eventually, the next morning, according to later Congressional testimony from FRBNY’s President E. Gerald Corrigan, the Fed’s New York branch in frustration ordered a brief halt on all BONY securities transactions using Fedwire. Lasting about an hour and a half, this gave the clearing bank time needed to sort out the mess and start receiving payments for securities it had paid for on buyers’ behalf the day before.
There had been other ramifications which seemed to expose some growing issues. As Corrigan also testified, with the outage enforced upon BONY, “the result was a backup in the willingness and ability of some other market participants to transfer securities among themselves.” They saw what was happening, out of prudence (information asymmetry) they took a (mild) step back.
This wasn’t strictly a potential hazard for Bank of New York nor limited to potential failures due to computer processing errors. And it wasn’t just about the Fed and Fedwire, either, since there were other rapidly advancing places for settlements of securities transactions as well as high-value interbank payments.
Fedwire has been around since 1918, literally started using telegraph wires so that far-flung Fed branches could more easily transmit payment requests on behalf of member banks. With the rise of the eurodollar system especially in the sixties, there grew a desire to modernize and extend payment processing across that increasingly vast offshore space.
Other than Fedwire, much of this privately got done via Telex. Revolutionary when first introduced in the thirties in Europe, by the late sixties demands on the system were outpacing the human capacities to essentially read, understand, and then process what had been nothing more than nonstandard text messages. It was said that it required an average of ten such messages in order to completely and accurately clear and settle a Telex-based payment request.
To at least ensure banks from all over the world were just speaking the same language, a consortium of 239 banks located in 15 countries got together, setting aside competitive desires, and cooperated to form the Society for Worldwide Interbank Financial Telecommunication. We know it as SWIFT.
By 1979, SWIFT membership had grown and so had daily volume; rising to around 120,000 messages per day, on average, during that year (just imagine the back office costs had these still been done via Telex; meaning, how much global monetary movement and ability had been unlocked by rising efficiency).
Contrary to popular belief, though, SWIFT is not an interbank payments system – it is instead merely a messaging system by which institutional members can effortlessly send and receive payment requests to and from all over the world. In dollars.
Quite a lot of those SWIFT messages end up in a place called CHIPS. Though this, too, arose during the revolutionary decade of the seventies it unfortunately had little to do with Erik Estrada and his partner Larry Wilcox. Rather, in 1970 the Clearing House Interbank Payment System sprung forth from the desire of New York Clearinghouse Association (NYCHA) members to more efficiently process, settle, and clear these global dollar transactions into and out of the offshore eurodollar market where Fedwire was, shall we say, problematic.
Previously, this trade group of a specialized dozen NYC money center banks had employed something called the Paper Exchange Payment System, or PEPS. As the name already implies, large-value (as opposed to large-volume) interbank transactions had to be cleared and settled by paper notice.
While effective, it, too, was becoming inefficient as volume increased alongside the expanding eurodollar market. Since most dollar-based transactions occurred with these NYC money center banks, or at the very least their foreign subs operating in the offshore eurodollar market, it made sense that the NYCHA would take the lead on developing CHIPS to replace PEPS.
But like SWIFT, CHIPS isn’t really the money behind these interbank mechanisms, either; it was and remains, also as the name implies, merely the clearinghouse that allows the crucial monetary functions to take place. It is, basically, a messaging algorithm that sorts and nets payment messages – including those coming from SWIFT. Incoming SWIFT messages get “mapped over” onto CHIPS messages on behalf of a whole range of financial institutions including those who may not be direct members (some transactions are settled on behalf of non-settling members by bilateral agreement with about twenty NYCHA settling members).
These two essentially work together in many ways; it’s been estimated that 70% if not more of CHIPS traffic originates as SWIFT messages.
And since about 90% or so of all cross-border, dollar-denominated interbank funds transactions are handled by CHIPS, that’s why the vast majority of (non-settling) participants in it are US-based subsidiaries of foreign banks (under NYCHA rules before 2001, these also had to be located in New York so as to be under the supervision of the New York State bank regulating authority).
CHIPS itself neither extends money nor, really, has any; it begins and ends each day with a zero balance at…the New York branch of the Federal Reserve. Even as Fedwire’s primary competition, CHIPS uses Fedwire in its final settlement processes to distribute funds to participants, or demand payments from them.
Beginning at 2100 ET (9pm) the night prior, and before each morning’s deadline at 900 ET, every CHIPS user must deliver, via Fedwire, its opening position requirement to the CHIPS account at FRBNY. This is pre-calculated and no bank can send or receive payments requests until the funds are obtained.
At the open, messages are then sent and received through a central queuing system which begins to match and net requests for individual banks; if Bank A sends a request to Bank B to pay $100 on behalf of one of its customers at the same time Bank B sends a similar $100 request to Bank A, the CHIPS system nets the two together leaving both banks’ credit position (established by the opening position requirement) unchanged.
At 1700 ET, further messages are disallowed and the CHIPS algorithm begins multilateral netting to tally up who owes what; funds that are deducted from each bank’s opening balance should they owe, on net, credited to others when the tally of messaged incoming receipts are more than outgoing payments.
This final net position is then used to determine each bank’s closing position requirement; if positive, the bank receives the amount via Fedwire from CHIPS; if negative, the bank must transfer funds into the CHIPS FRBNY account using same. On this basis, CHIPS always begins and ends the day at zero.
Like BONY’s case, however, there are occasions for daylight overdrafts, or intraday interbank credit. For one thing, some banks permit receiving customers to reuse funds received via CHIPS before end-of-day settlement may transmit net funds to the receiving bank; in this case, the receiving bank is lending credit to its customer.
Mostly, though, intraday credit is extended by the receiving participant to the sending participant. Since the latter doesn’t necessarily settle up until the end of day, it receives the equivalent of a daylight overdraft drawn from the receiver’s participation via CHIPS. And since none of these transactions are tracked, it never appears on a balance sheet, meaning intraday interbank credit is largely guesswork.
Some studies conducted during the eighties (one by the Dallas Fed in 1989) concluded that intraday credit might average anywhere between $30 to $55 billion, often peaking at mid-morning hours. I haven’t seen subsequent studies, though regulators claim to pay very close attention (especially, for “some” reason, following the 2008 global dollar shortage crisis still attributed to subprime mortgages by approved convention).
To manage such risks, CHIPS employs several rules to limit the potential for banks overextending these hidden daylight overdrafts; including real-time monitoring of queued up payment demands.
Should a bank get stuck short of funds, maybe like BONY had, the system demands limited recourse via several options. If customers withdraw an unexpectedly large amount, meaning a bank gets hit with a greater payment demand than it was anticipating, extended a large amount of intraday interbank credit from the receiving bank, ultimately that bank is expected to clear it by either borrowing in money markets (fed funds or repo) or tucking its tail between its legs and hitting up FRBNY’s Discount Window.
Assuming it has the collateral for two out of those three.
To even participate in CHIPS, settling members are required to post US government securities with FRBNY in advance and maintain them. That balance can change, too, based on conditions. This means that, functionally, intraday interbank credit is not “free”; there are these costs as well as any additional charges which might arise when a large closing position requirement shows up.
On occasion, this has led to what has become somewhat of a puzzling nonformal, nonstandard arrangement between parties already informally extending this intraday interbank credit and those in demand of it. One study published by the Federal Reserve Bank of Chicago (Zhou) wrote of:
“Given that intraday credit is costly (either because of pricing or collateral requirements), while making a payment sending/withholding decision, a bank faces the tradeoff between sending the payment order promptly by borrowing costly intraday credit or delaying the payment and suffering the delay cost. In such an environment, if banks cooperate to maximize joint profit, there will be no delay (no payment order is blocked by other banks delayed payment). In a noncooperative equilibrium, however, a bank will delay a payment order to reduce its expected intraday-overdraft cost and wait for the incoming funds to arrive. By doing so, it transfers the intraday-credit cost to the payment-receiving bank. The negative externality generated by this delay is a dead-weight loss to the payment system, and it cannot be eliminated by the existence of the intraday money market because liquidity on the intraday market will also be costly.”
Regulators and central bankers assess the risks of these systems like CHIPS almost exclusively based on the potential for a bank failure to ripple through these markets and cause a cascading effect (one study in 1986 using a couple days in 1983 linked a possible bank failure to ripping Fedwire intraday credit apart to the tune of 30 to 40% interbank defaults downstream).
The most extreme disruption need not be the biggest concern; something like what happened with BONY at Fedwire. Illiquidity can happen when there are no failures, and the system simply finds itself short of cash – or collateral.
As Zhou pointed out, there are real-world ripple effects which banks will absolutely deploy hidden inside the incredibly darkened world of intraday credit (“negative externalities”). Behavior is altered to account for even the possibility of reaching and then sustaining daylight overdrafts and their equivalents (the lesson of Bear Stearns).
The Federal Reserve has made it a policy for intraday credit to be punitive (beginning 2003 when Primary Credit, what used to be known as the Discount Window, was revamped and set at some policy-determined spread above the federal funds target) in order to avoid a situation where the banking system as a whole builds up hidden intraday leverage (one key criticism of the BONY event was that it was allowed to do so at Fedwire, thus when something went wrong FRBNY was on the hook for potentially catastrophic losses from “free” intraday credit).
But in this condition, where the Fed stands back only to extend emergency credit, it doesn’t account for other, less-obvious forms of illiquidity (including, again, collateral concerns).
What that means is the Fed expects any CHIPS member who comes up short will have to meet this shortfall in wholesale money markets. Given that federal funds isn’t much of a marketplace any longer, not since August 2007, all that’s left is either repo or Primary Credit; while both collateralized, the latter a modern-day death sentence.
Banks will absolutely undertake whatever alterations to their behavior, negative externalities, creating potentially severe monetary frictions for the system along the way in order to avoid any chance of getting stuck in that situation – especially during periods when collateral is perceived less available, more costly, and therefore intraday credit that much more relatively expensive.
I believe this best explains the nuts and bolts of the September 2019 repo issue, rather than some sudden, tax-induced shortage of bank reserves unexpectedly locked up from the banking system in the TGA.
It's not just about September 2019 repo, either. Systemic illiquidity, we’ve been led to believe, is some thing about bank failures alone. Nonsense. If there isn’t another Lehman Brothers, that doesn’t mean illiquid, frictional behavior isn’t happening. Bank failures are instead only the most public – not necessarily the most extreme – form of it.
And if there isn’t a failure, that doesn’t mean these problems, these monetary frictions, might not be serious and therefore a very hindrance upon markets and the economy.
Via CHIPS, global economy.
Monetary policy dictates a focus on bank reserves and to a certain, specific extent extends only to the Fedwire parts of the “plumbing.” In many of these interbank pathologies, those Fedwire and thus Federal Reserve pieces are and have been largely incidental. There is so much more going on than you’ve been led to believe. They know it, they just don’t want you to because it undermines monetary policy “credibility” and the central bank’s ability to conduct money-less expectations policy which makes all these things sound as easy as the push of a button.
Just so long as it isn’t attached to a November 1985 BONY computer, I guess.