Blockchain for Banks Probably Can't Hurt
A share of stock is a piece of paper covered with fancy embossing and pictures of eagles and things. The way you sell a share of stock is, you and I meet under a buttonwood tree, and I say I will pay you 10 gold dollars for one share of Amalgamated Pennyfarthings stock, and we shake hands, and then a day or two later you have your clerk bring over the piece of paper to my clerk, and I have my clerk bring your clerk the 10 gold dollars in a little bag labeled "$," and your clerk bites my gold dollars to make sure that they contain the right amount of real gold, and the clerks make the exchange, and I put the fancy piece of paper into my vaults and bask in my new part-ownership of Amalgamated Pennyfarthings.
Or that is sort of how it used to go, and it is also sort of how it notionally goes today. But the way it actually goes today is that you go to your online brokerage account and put in a market order to sell one share of Universal Hoverboards, and I put in an order to buy one share, and a computer matches us up, and then my broker automatically adds one share of Universal Hoverboards to the electronic file representing my account, and your broker automatically subtracts one share from your electronic account, and vice versa with the money. And then a thing called the Depository Trust Company, which is owned jointly by your broker and my broker and all the other brokers, does the same thing for the brokers, subtracting one share of Universal Hoverboards from your broker's electronic account and adding one to my broker's electronic account, at the brokers' electronic instructions. There may or may not be an embossed piece of paper -- probably not -- but if there is, it sits in DTC's vaults and no one's clerk needs to move it anywhere. All of this could happen, in round numbers, instantly, though in practice it mostly takes three days.
This is a huge improvement over the clerks and the bags and the embossing, and everyone basically loves it, though every so often it goes weirdly wrong because its glossy efficiency is built on top of the former rickety 18th-century infrastructure. But it is not especially rocket science. The idea is pretty simple: A finite number of brokerage firms intermediate most public stock trades and have a mutual interest in getting those trades right. And those firms all own computers. So if they get together and agree on who owns what, and on who sold what to whom, then they can keep track of that on the computers and make it much faster and more efficient and more reliable than it was when it was all kept on scraps of paper.
Here is a fun Bloomberg Markets article about Blythe Masters. Masters, who is beloved in financial circles for previously sort of inventing credit derivatives and electricity-market gamesmanship, is now marketing ways for brokerages to get together and agree on who owns what, and on who sold what to whom, so that they can keep track of that on computers and make it much faster and more efficient and more reliable than it currently is. Not for public stocks. Those are fine, DTC already did that. But, for instance, syndicated loans are pretty lame:
When investors buy and sell syndicated loans or derivatives or move money around the world, they must cope with opaque and clunky back-office processes that rely on negotiated contracts between buyers and sellers, lots of phone calls, lots of lawyers, and even the occasional fax. It still takes almost 20 days, on average, to settle syndicated loan trades.
There are various reasons for this, some having to do with the specifics of how loans work, and some having to do with tradition and inertia. But of course it is tempting to rationalize the process. Get a finite number of loan traders -- big dealer banks, at least, and maybe some hedge funds and other buy-side loan-market participants -- together, have them agree on some standardized terms and a process for acknowledging trades, and then put all the ownership and trade information in some central computer database rather than relying on faxed paper trails and borrowers' ledgers to figure out who owns what.
So that's what Masters -- and Sunil Hirani, Don Wilson and their firm Digital Asset Holdings -- will do:
In March, Masters joined Digital Asset as CEO. She, Hirani, and Wilson set to work developing blockchain-based software for three inefficient markets they deemed ripe for an overhaul: syndicated loans, U.S. Treasury repos, and equity shares in private companies.
Oh right sorry "blockchain." Blockchain. The blockchain is the cryptographic whatsit that processes bitcoin transactions. Its bit of magic is that it allows transactions to be verified in a completely decentralized way, without trusting any central authority. Bitcoin transactions are processed by a decentralized web of computers. Those computers agree on which transactions are legitimate, and on who owns bitcoins. The record of their agreement is the "blockchain," and it is stored on each of their computers. The people running the computers are paid for doing this transaction processing work: The work is called "mining," and miners are rewarded with bitcoins. The mining rewards provide an incentive to people to process transactions and to ensure that the blockchain is correct, all without any central trusted overseer.
This is a delight to libertarians, but it is not what is going on at Digital Asset:
At the same time, Masters recognized that the open structure of the bitcoin process—no one controls who does the mining—would be anathema to an industry in which client confidentiality is sacrosanct. So in July, the company acquired Hyperledger, a San Francisco software firm that’s developing the technological equivalent of gated communities. Its system is designed so that users will be able to process transactions themselves rather than depend on the open bitcoin blockchain.
“With private chains, you can have a completely known universe of transaction processors,” Masters says. “That appeals to financial institutions that are wary of the bitcoin blockchain.”
So rather than a trustless decentralized blockchain, you get a trust-based blockchain jointly owned by the transaction processors. (That is, by recognized market participants. The analogy is to the brokers who are "participants" in DTC.) This is in part because the bitcoin process tends to attract miners who are scruffy and/or anonymous, which is awkward in the heavily regulated world of financial services. But it is also because the bitcoin process attracts those miners by paying them in bitcoins. If you don't want your blockchain to rely on mining -- if you don't want to pay your transaction processors in bitcoins -- then you need some other way to incentivize them. "Come join our network because it will cut down on your transaction-processing headaches" is a perfectly reasonable way to incentivize transaction processors who were already processing transactions, like banks and brokers. (It's more or less DTC's pitch!) It seems unlikely to appeal to anyone outside that ecosystem.
In its most straightforward form, you could build a blockchain that is decentralized (among the participants) in sort of the same way that bitcoin is. Each participant could run the software simultaneously and keep separate copies of the transaction records, much the way that brokers currently keep their own accounts of what shares they own at DTC, but the "official" copy would be the one that participants agree on, rather than a centralized copy kept at one official arbiter like DTC. This has some advantages over the centralized database (what if DTC's computers crash?), and some disadvantages (uses more electricity), but it is not ultimately wildly conceptually different. It's just some brokers getting together and agreeing on who owns what and who sold what to whom. "There is an ongoing debate among cryptocurrency enthusiasts over whether a private blockchain is a sustainable model and anything more than a traditional database."
- It is dumb and bad for reasonably standard market transactions to take 20 days to settle because archaic procedures require market participants to fax each other documents and perform ancient incantations.
- Market participants should get together and agree on a way to fix that.
- That way should probably involve computers.
But do you care about exactly how the computers determine and record the market participants' agreement on who sold what to whom? You could have a centralized model where the participants get together, set up a DTC-like entity, and let that entity keep track of ownership and transfers. Or you could have a semi-decentralized model where the participants get together, agree to run the same blockchain code, and keep track of ownership and transfers by consensus. There might be good technological or practical reasons to prefer one or the other, and certainly the blockchain excites many technologists. But the point is: Either of those models seems much better than waiting 20 error-prone days for a trade to clear. If clunky old procedures haven't been replaced by computers by now, why would they be replaced by blockchain computers in the near future?
Part of the answer is that things have to be computerized eventually, so putting your money on any sort of computerization has a reasonable chance of paying off, and the current vogue happens to be for blockchains. Part of the answer is that the blockchain vogue has a certain universality right now, at least among financial-technology types, that makes it an appealing pitch across markets. "We need to find a way to reduce processing time in syndicated loan markets" is a pretty niche pitch, perhaps appealing to the back-office guy who handles syndicated loans. "We want to use the blockchain to trade syndicated loans" might get you a higher-level audience, perhaps with the chief technology officer or the head of loan trading. And "we want to use the blockchain to trade syndicated loans and Treasury repos and private company shares and whatever else you've got" could get you in front of the CEO.
But surely part of the answer is Masters herself. Her undying fame comes from figuring out, some two decades ago, a way to abstract away from trading syndicated loans. JPMorgan was making syndicated loans that, for reasons of custom and tradition and market structure, it couldn't sell. Masters helped invent credit derivatives as a way to transfer the credit risk of the loans without transferring the loans themselves. Credit derivatives abstracted away from the clunky mechanics of actual loans, allocating credit risk not based on who had a particular pile of documents but rather on who had agreed to take that risk. That abstraction looked radical at the time, but it has had a lasting effect on how the financial industry works. It's not crazy to think that Masters might be able to do that again.
Obviously don't hold me to this in any details. This is not 18th-century investing or legal advice.
Ha, no come on, what actually happens is that you sell immediately to a wholesaler, and I buy from a wholesaler, and the share passes between like six different high-frequency trading firms in between. But I am simplifying.
Also a fun picture of Masters with her three dogs (update: four, in the print edition), who look like good dogs.
The description that follows is, obviously, simplified. A good source for understanding bitcoins and blockchains is Nathaniel Popper's "Digital Gold." Here is the bitcoin.org introduction, which links to more resources.
Or whatever. Like, you could have a syndicated loan processing network that pays processors in syndicated loans. But where would you get the new syndicated loans? The point of bitcoin is that the bitcoin network creates bitcoins ex nihilo. That's harder with securities.
And I should say, I have no idea if Digital Asset is up to this form or some other variant.
It may be the case that blockchains do a better job of keeping in one place both current ownership and transfer-history information than most centralized repositories do. That is potentially useful. From the article:
Anyone with access to the ledger can read the contract with a click of a mouse. That means regulators, who depend primarily on self-regulatory organizations to police the markets, could easily verify that a securities transaction didn’t violate anti-money-laundering rules or other laws.
But of course you could build a centralized database that did that too.
Incidentally, in Treasury repo markets there is some computerized centralization, in GCF and tri-party repo. But those are not just about trade processing; they are also about the centralization of credit risk, and create their own worries.
Also, there is recent history of a centralized database getting some pretty bad press. I am thinking of MERS, the Mortgage Electronic Registration System, which is sort of like a DTC for mortgages. Like DTC, it was premised on the idea that you keep the antiquated paper certificates in one place (or, in the case of mortgages, many many county property records offices) and just move around electronic entitlements to those certificates. But -- in part because it was sometimes a bit sloppy with the paper certificates -- it has been criticized as a way for banks to evade property law, and its transfers have sometimes not held up in court (though that decision was reversed on appeal).
So that is a bad recent precedent for banks building their own centralized databases. Of course, the problem isn't quite the centralization of MERS's database, but rather with how mortgages got into the database in the first place, and it's not clear that a blockchain architecture would have helped.
As Gillian Tett puts it:
Blythe Masters fervently hunted for a way to make credit derivatives work, and eventually she spotted an opportunity at Exxon. In 1993, after Exxon was threatened with a $5 billion fine as a result of the Valdez oil tanker spill, the company had taken out a $4.8 billion credit line from J.P. Morgan and Barclays. When Exxon first asked for the credit line -- which is a commitment to provide a loan, if needed -- J.P. Morgan was reluctant to say no because Exxon was a long-standing client. The loan epitomized the twin problem of capital requirements and internal credit limits. Like so many of J.P. Morgan's corporate loans, it would produce little, if any profit, yet would gobble up credit, pushing the limits, and would require a large amount of capital reserve. In theory, the bank could have dealt with that headache by selling the loan to a third party, since a market for selling such loans did exist. But that would have violated its commitment to client loyalty.
Masters thought she could see a solution.
The solution was credit defaults swaps, and the rest is history.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Matt Levine at firstname.lastname@example.org
To contact the editor responsible for this story:
Zara Kessler at email@example.com