Hacker News new | past | comments | ask | show | jobs | submit login

> There is no reason, with all the identities attached, and auditing procedures, and digital signing, and protocols we have today, that we shouldn't be able to know who owns the stock at any given point, and not have to rely on these super-slow resolutions.

Trillions of dollars are moved in these markets - you can't just hop in and do a quick rebuild with crypto, launch it, and call it good. Imagine rebuilding ythe core infrastructure at Microsoft - that takes years and years. But now you also have the entirety of the global economy dependent on your software. You also need to meet heavy government regulation and comply with oversight.

There is a whole shit ton of reasons this process hasn't been updated to T+0. It was t+3, now t+2, and I've heard (from the CEO of Webull on bezinga power hour yesterday) wanting t+1 this year.

Edit - to be clear I'm not saying dropping the current system for a new one is the only option. I'm saying improving the system at all (or rebuilding it) is arguably one of the most challenging software tasks one could undertake and there are serious reasons financial markets are not anywhere near what is "technically possible".




When one realizes that in January 2021, some major financial institutions are still settling transactions via CSV files sent on basic FTP, the scope of the problem becomes clear!


<< laughs with tears and a devilish look in eyes >>

Regardless of the fact that most of the banking sector is still stuck on IBM mainframes from the 60s-80s running COBOL:

I have seen accountants who's only job it was to come in every day and do the same sums on Excel.

I've seen people insisting on a calculator and a printout so that they could sum up columns of an Excel table and send the results back via email.

There are valid reasons to move slowly. Transition costs to new systems are usually immense and the process is a nightmare for banks, but none of what I described fell into that category. It was just people refusing to change their ways. A report for treasury could've been instantaneous with a super simple live updating dashboard. But no. Instead, the CEO got an Excel file emailed to them every week that was put together by 40 people – many of whom entered the numbers manually.

In the industry we have come to call people involved in these tedious processes "hamsters", because they might as well be going up and down the escalator all day.

I don't think many people understand how excruciatingly slow banks move and how inefficient they are.

/cathartic rant over

Edit: (I should add for context that this was a fairly large bank in continental Europe)


Certainly, there are many inefficiencies in banking (and other industries that have been around forever). And the mismatch between trading and settlement here obviously caused some problems (note, though, only at the hip fintech brokers, not at the old fashioned brokers that insist that your trades in a cash account settle before you trade again).

However, may I suggest that the better solution to this mismatch is not to speed up the settlement infrastructure (well, that too, to an extent), but to slow down trading??

A couple of auctions every day instead of continuous limit order book trading and a Tobin Tax on trades would eliminate a lot of rent seeking, simplify lots of things, and hardly affect the fundamental functions of the capital markets at all (maybe improve them).


It's actually TSV via SFTP and well over 50% of American wealth operates this way.


I'm pretty sure we broke one of our bank client by telling them that data import will be exclusively with SFTP. We ended up allowing FTP (and thus our security certification was voided).

It was a Euro bank though, but it was barely two years ago.


SFTP if they're "compliant." FTP is still in the wild.


Tons of businesses exchange data like this (e.g. drop ship product availability, etc). Platform independent and easy to read over if things go wrong. I don't see the issue.


So they do move with times. /s


Actually a large part of financial operations goes through Excel sheets automated by mountains of VBA code.

Hilarious. It's definitely out of the if it works then don't touch it idiom.


I still use wooden sticks for roasting marshmallows. That tech stack is ancient! Can you elaborate on your objection to CSV for record transmission? What would you suggest as a replacement?


SFTP, I hope.


[flagged]


Could you please stop posting swipes and flamey comments to HN? You've been doing it a lot lately and it's not what this site is for, no matter how annoyed you may feel about some of what you see here.

https://news.ycombinator.com/newsguidelines.html


> Trillions of dollars are moved in these markets - you can't just hop in and do a quick rebuild with crypto, launch it, and call it good.

OP talked about how 'this is absolutely possible', but you're responding to him by saying "but we can't just drop everything and move to the new system".

You're right but that doesn't make OP's point any less correct. Generally in a legacy system we migrate by building all new features onto the new system. For instance, if the company wants to move their legacy jQuery based banking app to Vue.js, they can start by building a more orthogonal component in the new technology, so it doesn't affect the other thing. Eventually once enough things have migrated (possibly years later), the benefits of the new system justify the cost of migration even more.

Stock of an existing company like GE is different than a company which is yet to launch (say Coinbase). The best way (perhaps the only way) to migrate is to start launching new IPOs on this new system. We did migrate from being on paper to computers, I'm sure we can do it again (and hopefully with better technologies in the future...again).


What are the “whole shit ton of reasons” why shorter settlement periods are so difficult?

In particular, I’m curious why we didn’t jump straight from T+3 to T+1. Even if T+0 is especially challenging, what would make T+1 substantially harder than T+2?


As with all these changes, it's mostly about coordination and legacy systems.

Just as Y2K was a challenge and just as the 2038 date will be a challenge, it's all about pushing changes through legacy systems. ascii->unicode, ip4->ip6, python 2->3. We all know the drill. We've all lived through these things.

It's rarely a technical problem. It's about coordination across firms, domains, people, and systems that may not be known ... until they break.


Yes. I used to work at an investment bank. When the market moved from T+5 to T+3 it was a major project for all internal systems to be adjusted. It wasn't just a matter of changing a SETTLEMENT_DAYS macro in a header file. And even afterwards, it wasn't that, because there were different systems written in different eras and in different languages.

Multiply this by however many thousands of firms, all with their bespoke back-office systems. and it takes time.


What language was this written in? Did it stay in the same language? If it's say, FORTRAN, I wonder if there's value in learning it soon.


Back office system in the financial sector? You're probably looking at COBOL, although I know Caché gets used as well.


Coordination problems can become technical problems - or be caused by technical decisions.

Moving a system/network of actors from one system to a new, incompatible system? You need to coordinate the switch so it all happens at once. Can’t coordinate? Then you need a compatibility layer between the two systems.


My understanding is that the delay is due to humans in the loop. Most of the clearing is automated, but if the numbers don't add up, humans still intervene to figure out how to reconcile the differences.


Other comments have claimed that the humans in the loop are because people insist on hand calculating everything. Not for when it goes wrong, but in the first place.


I find that hard to believe, the sheer volume would overwhelm a human.


>Trillions of dollars are moved in these markets

...and they jammed when a single subreddit decided to buy GameStop shares. Just think about what that means.


The market was not jammed, only Robinhood was jammed. I could still buy GME on Schwab if I wanted


"GameStop and AMC trading restricted by TD Ameritrade, Schwab, Robinhood others "

https://www.msn.com/en-us/money/personalfinance/gamestop-and...

"Some users are experiencing issues with trading platforms Vanguard, TD Ameritrade, and Charles Schwab due to heavy volume"

https://www.businessinsider.in/tech/news/some-users-are-expe...

Not to mention that the market is made up of all platforms and several of them were downright prohibiting GME buying. Some are listed in the article. Robinhood alone has more than 13 million users.


The restrictions at Schwab (and I'm pretty sure TD) were different than Robinhood, they only restricted shorting, buying on margin, and selling naked options. Buying and selling GME/AMC with cash were available the entire time (I could do it).


... or Fidelity or Vanguard or Wells Fargo or Citibank or ...


I had heard that Schwab and TD Ameritrade suspended GME trades as well, was that not true?


They did not. They did introduce restrictions: https://www.tdameritrade.com/td-ameritrade-trading-restricti...


Theres a site for that: https://www.wherecanibuygme.com/


I was unable to look up the GME quote on TDA Thursday. In their UI, the quote is the place with the Buy button.


I used TDA via the ThinkorSwim app and it had no issues on Thursday. If you are at a PC and using the site, I'd suggest getting the TOS desktop app as it seems to be a lot more stable. It is NOT user friendly though so it may take a tad to get used to.


Related but separate. Robinhood has an internal clearinghouse which halted trading for Robinhood.

Apex is a large third-party clearinghouse used by M1 and others that restricted buys for all (most?) of its clients. TDA and Schwab both have their own internal clearinghouses so far as I know.


What? A few (new) trading platforms that are deliberately trying to shake up the market 'jammed'. The market itself was fine.


>Trillions of dollars are moved in these markets - you can't just hop in and do a quick rebuild with crypto, launch it, and call it good.

Which is why I didn't say anything like that. Just, that settlement time should have improved with our protocols for validating ownership, not been frozen in time.


It has. It used to be T+5 [business day] settlement until 2004, then T+3 until 2017, and is now T+2.


I meant, improved proportionally.


To be fair, That's nothing. High frequency traders have been on the order of microseconds for years. everyone trades under the assumption that the system is faster than it is. This is going to be an unpopulair opinion but wall street is esentially a government-like entity built by the elite. troughout history any scheme that destroyed market assumptions have been met by the SEC under the guise of protecting the fair, orderly and efficient market. No thought is given to actually fix design flaws in the open market. Just prosecute anyone who manipulates it incorrectly.


Russia called it a modernization when they moved off T+0 to T+2 because people wanted more slack to complete their payments.


To be fair - the issue with demanding immediate settlement is that fund transfer settlements aren't instantaneous.


I mean you can still offer credit in T+0 settlements but the real challenge (I think) is setting up an industry-standard margin pledge facility for trades to happen without immediate availability of liquidity.

Here's DTC's whitepaper from 2018 discussing some issues with real-time gross settlement: https://www.dtcc.com/~/media/Files/downloads/Thought-leaders...

TL;DR: Market makers using a netting mechanism to settle trades at the end of the day, and they often don't have the funds necessary to settle trades in real-time.


>>Imagine rebuilding the core infrastructure at Microsoft - that takes years and years. But now you also have the entirety of the global economy dependent on your software.

That is a bad analogy, MS takes years to change the core because it require INSANE levels of backwards compatibility. Apps written for windows 2000 still work and need to work today...

This would not need to be the case with this type of system, there are a whole host of political reasons why the system is built the way it is, part of it is the desire to control it (and in this case the wrong people where doing things they were not suppose to)

I am sure you will call that "conspiracy", but the reality is that the investment system is setup to be slow and opaque not because of the need for backwards compatibility, or security it is that way so the "correct people" control it plain and simple


In fairness, stock exchanges care even more about backward compatibility and so it would be at least as much an issue for them.

Edit: Sorry, I say it way too much, but this merits a repeat of the one-liner: "The reason God was able to finish the earth in only six days is that He didn't have to worry about backward compatibility [or legacy system integration, or satisfying an installed userbase]."


Sure, but there's no real reason why ALL companies need to have their stocks compatible with each other on the same exchange.


> [T]here's no real reason why ALL companies need to have their stocks compatible

I think it would be way harder to trade a basket of stocks (e.g. pairs trading, going long one and short the other) if you had to worry about mismatched settlement dates across the different stocks; it would be like trading spot against a one-day forward.


You've just defined exactly what high frequency traders do to make money. They balance all stock exchanges in order to make tiny profits on the stock differences.


They don't, there are hundreds of exchanges and plenty of competitors for listings - CSDs & clearing are a pinch point.

But in some ways exchanges also function as a natural monopoly; especially for primary issuance. e.g. look at AMEX's IPO slate compared to the big two, where the market is right now - it's insignificant, and they are the #3 exchange in the US. People want to list to make money, and that means going where the liquidity is until there's a really good reason not to, like with NASDAQ's move to electronic trading in the '70s.


I think even Win95 apps work on Windows 10, not just Windows 2000 :-)

Think about all the flak they're getting for Settings/Control Panel. It's a multi-decade process to rewrite all of Control Panel, because Control Panel supports custom integrations plus it's such a central piece of software that looking at the code the wrong way probably breaks some client doing some crazy stuff with it :-)




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: