Market markers and other institutional participants don’t want instant settlement, as it its a liquidity risk. It’s why you’ll see T+2 go to end of day settlement eventually (imho), but will never see real time settlement from centralized finance. Decentralized finance would improve upon this, with the trade off of having a low volume ceiling (due to network transaction limitations).

https://www.dtcc.com/~/media/Files/downloads/Thought-leaders... (pages 8 and 9)

The problem with this position is that it's false. I welcome corrections, but there is a wealth of literature on this topic. For the past two to three decades there has been a small cottage industry of research in Liquidity Savings Mechanisms that are designed to prevent this scenario through clever system design.

These mechanisms have been evaluated, simulated, and tested by central banks at greater volumes than DTCC's system. The difference here is in orders of magnitude. The Feds handle $2+Tn./day. They handle $500Bn/day at most. If the Feds can introduce these systems and maintain liquidity, then what reasoning does DTCC have?

https://www.newyorkfed.org/medialibrary/media/research/epr/0...

I am loathe to cast aspersions, but the aspersion I'm casting here is that they're dragging their feet, as it would effectively end an entire sub-sector of the industry. Or, at the very least, substantially re-order it. This inefficiency is someone's margin and profit.

Maybe this is an attempt to force that change to be faster? It really should be faster today, clearing sped up.

The lag for bank account transactions makes sense to wait to make sure money is there and no fraud. However, once money is in an account at an approved FDIC/SIPC insured (preferably both, Robinhood is only SIPC) brokerage it should be much faster, at least up to the FDIC/SIPC limits. There is really no reason why it shouldn't except to help market makers have the upper hand.

This could be the root of what may need to change or was desired to change with this black swan event.

Faster settlement means less time to discover and correct mistakes.

At a high level, nobody really likes trades being broken, but it's kind of a necessary "evil" that helps stabilize the entire system.

Let's say settlement happens in one minute, one morning a pension fund fat-fingers a price and loses 10 million USD to some student in her dorm room. That afternoon, the student then donates 1 million USD to an orphanage, pays off her grandparents' mortgage, buys 1 million USD worth of municipal bonds to finance a new baseball stadium, and buys several million USD in long-dated ETF call options. Some market-maker delta-hedges those options by buying the ETF. In response to the price move generated in the ETF, a large institution buys a basket of stocks and participates in the creation process with the ETF's issuer.

Now, what happens when the fat-finger is discovered and disputed? Too many things have changed for the SEC to break all of the trades and other transactions after the settlement. The student no longer has enough money to make the pension fund whole (even if everyone involved could agree that the most fair thing to do would be to limit the student to 100k profit and transfer the rest back to the pension), so I guess they're stuck with the pensioners having to live with that 10 million USD loss. "Oops, next time, make sure your pension hires a more careful trader, I hope you've learned your lesson" isn't very helpful advice to a retiree who's now out on the street.

> "Oops, next time, make sure your pension hires a more careful trader, I hope you've learned your lesson" isn't very helpful advice to a retiree who's now out on the street.

It's really not, isn't it? Almost like the dependency on human carefulness compromises the integrity of the whole system.

We've seen plenty of trading problems due to direct human error, and we've seen plenty of trading problems due to software bugs.

If we got rid of breaking trades, it would presumably force everyone to use carefully specified systems and use avionics-like development practices. Maybe we'd even see machine-checked proofs being compiled to programs and smart contracts via the Curry-Howard correspondence. Long-term, it would probably be good. In the shot term, lots of fallible counterparties going bankrupt would be very destabilizing to the system.

> to make sure money is there and no fraud

"make sure there is no fraud" takes a LOOOONG time.

It is the reason why a car bought from a dealer drives around without plates for a month. That's how long it takes the dealer to be totally sure the payment wasn't fraudulent.

At least for cash and bearer instruments, you can acquire them in good faith, even if they turn out to be stolen. Well, in Germany, that is:

https://www.gesetze-im-internet.de/englisch_bgb/englisch_bgb...

While cash is the settlement vehicle of choice for dealer-less used car sales, certified checks (only legally issued by the federal bank) are preferred or even mandated for in-person government (surplus) auctions.

If you talk to the dealer in advance, to make sure they know about these certified checks, I'd expect them (over here, in DE) to hand you the title in direct exchange for the check, within an hour of you first setting foot on to the lot (assuming you called in advance so they have time for you and the car you want).

Sorry, I should've said "in the USA".

Our banking system here sucks ass.

What you say about certified checks is, in theory, possible here. In practice the sales drones who handle these transactions have never heard of this and will treat you like you're trying to scam them out of a car somehow.

Also the title is never physically at the location where the car is sold from; the processing of titles for dealers has been outsourced to a handful of large companies who effectively act as warehouses for the physical title documents. Even if the sales drone wanted to go along and had enough clout to make it happen, it would still take an overnight-shipping delay to get the title to the buyer.

Yeah, hence me specifying Germany.

I know, it's kinda sad, tbh. Get yourself a taste of SEPA to see what you could have/use instead of ACH.

Nice to hear. And yeah, that's why I mentioned needing to call ahead to make sure that someone who can legally sign the contract is gonna be there and has sufficient understanding of how these certified checks look like/work.

Fair. Is that only for new cars, or does that include (independent) used car dealerships?

> A substantial amount of trades in the market are executed by liquidity providers that provide efficient two-sided markets. These important trading firms trade on both sides of the market and are largely risk flat at the end of the day. Real-time settlement would require these liquidity providers to have significant sums of capital and securities on hand to make trade-for-trade deliveries. The current infrastructure relies on the efficiencies of trade netting to accommodate substantial amounts of trades. Wholesale market-making is important to the infrastructure as it provides the capital and balance sheets to back customer trading. Significantly increasing those capital requirements by requiring pre-funding of all trades without the benefit of recognizing the offsetting trades would substantially increase the costs of trade execution to the end user. Real-time settlement could also introduce operational inefficiencies. In a typical low-volatility trading day, NSCC and DTC process over one million shares per second on average, valued at over $16 million. During peak trading hours, such as market open and close, this number spikes to over 300 million shares per second, valued at over $25 billion. Settling these amounts throughout the day in real-time would introduce substantial financial and operational risks to the equity markets. Accelerated settlement that abandons the significant capital and operational efficiencies gained through centralized multilateral netting would be a step backward for the world’s most liquid markets. Striking a balance between the capital efficiencies of netting and the risk mitigation benefits of moving settlement closer to trade date should be the goal of the industry.

Thanks! Replacing market makers seems difficult.

Edit: Why the downvotes? I'm quoting the article for easy access.

> Significantly increasing those capital requirements by requiring pre-funding of all trades without the benefit of recognizing the offsetting trades would substantially increase the costs of trade execution to the end user.

I don't see how this follows, if the trades are settled faster? Wouldn't you have to keep less capital on hand at any given time if you're settling trades faster? I'm clearly missing something, not sure what.

You aren't, there are netting, batching, and matching algorithms that have been developed for larger systems. The Feds, who process $2Tn.+/day, BoE, ECB etc have found these systems to hold up through trials, simulations, and implementations. DTCC, on the other hand, handles $500Bn.

Here's the Fed's review, https://www.newyorkfed.org/medialibrary/media/research/epr/0...

DTCC doth protest too much.

I'm not an expert but how I read it is: market makers currently can sell a stock without owning it. All is well as long as they buy it back by the end of the day as settlement takes 2 more days.

If settlement was instantaneous then the market makers would have to already own a bunch of the stock in order to sell any. Owning stock requires capital. Similarly if they wanted to buy more stock they'd need cash on hand.

Would they though? If settlement was instantaneous for market makers, then obviously settlement with the brokerages would also have to be instantaneous. Then it's no longer an issue of verifying anything, the risk of proving ownership would fall squarely on the brokerages or whoever is initiating the trade with the market maker in the first place.

I don't understand your question.

If settlement was instantaneous then to sell a stock on an exchange you'd have to have the stock. To buy you'd have to have cash.

Currently market makers don't have either.

Most defi is on Ethereum, which is making large scaling improvements. Zkrollups right now do several thousand tx/sec without security compromises, and data sharding will expand that by 23X in about a year. This is still probably a low volume ceiling by some definitions, but it's at least in the ballpark of major systems like the VISA network and NASDAQ trades.

Further down the road, more efficient data witnesses give another 10X. Also, the power of individual nodes affects both the number of shards and the capacity of individual shards, so future scaling is the square of whatever Moore's Law has left to give us.