Wednesday, April 26, 2023

A quick defence of digital substrate agnosticism (or, why banks should be able to issue stablecoins)


In principle, I'm pro banks-issuing-stablecoins.

I think banks should be able to issue dollar IOUs on whatever digital substrate they see fit, whether that be an Azure SQL database, an Oracle database, or a blockchain. And they should be allowed to get deposit insurance for any of those dollars, regardless of the substrate on which those dollars are recorded. The medium should be irrelevant.

Banks should also be able to run their blockchain-based dollars on a fractional reserve basis, just like they already do with their Oracle and Azure-based dollars. That's right. Bank stablecoins shouldn't be required to operate on a full-reserve basis.

So when I see the following recently-announced initiatives, I sort of shrug and say, sure, that's fine. In Japan, three banks are going to test stablecoins while in Europe, Societe Generale wants to issue a stablecoin called EUR CoinVertible.

The Japanese example is notable, because it stems from recent amendments to Japan's Banking Act, the Payment Services Act, and the Trust Business Act allowing banks, fund transfer companies, and trusts to issue stablecoins.

Meanwhile, in the U.S., the Fed and its sister regulators have taken a different path than the FSA, recently suggesting that issuing tokens on a blockchain is "inconsistent with safe and sound banking practices." That doesn't mean that U.S. banks can't issue blockchain-based dollars, but I suppose it's a stern enough put-down that they won't ever bother.

I'm not a fan of the U.S.'s approach. Banking authorities don't say: hey, Bank of America, we'd prefer you use an Oracle database instead of an Azure one for your IOUs. Likewise, they shouldn't get to say: hey, Bank of America, don't use an Ethereum database instead of an Oracle one.

Mind you, unlike the crypto idealists, I don't think blockchains are the revolution that they are often made out to be. They're just another database, one with some quirky characteristics, so let banks figure out on their own whether they are a worthwhile medium or not. There's a high likelihood that blockchain-based dollars won't be successful with customers, but banks won't really know until they experiment.

One last point on this topic. In the same vein of substrate neutrality, if banks are going to use some sort of blockchain to issue dollars, they should be required to subject their blockchain-based dollars to the same anti-money laundering checks to which their non-blockchain based dollars are beholden. 

That means identifying all their users. This would be a departure from current practice among blockchain-based dollar issuers (like Tether and Circle) whereby they do KYC on just some users. A defence of substrate agnosticism suggests that the current KYC-lite touch isn't enough.

Monday, April 24, 2023

Zelle vs Interac e-Transfer, or why it's so difficult to kickstart a payments network in the U.S.

It's difficult to grow a payments product to universality in the United States, and that's partly due to the fact that the U.S. has a stunning 4,127 banks, 4,760 federal credit unions, and 579 savings & thrifts institutions, for a total of 9,466 depository institutions.*

Let's compare that to Canada. The rule of ten applies to most Canadian/U.S. comparisons. That is, the U.S. has around 10 times the population, so to get Canadian equivalents just divide by ten. (For example, there are 13,515 McDonald's restaurants in the U.S. Meanwhile, Canada has 1,363. That's almost perfectly in-line with the rule of ten's prediction.)

The rule of ten suggests that if the U.S. has 9,466 depository institutions, then Canada should have 946. But that isn't the case. Canada has 81 regulated banks and around 208 credit unions, for a total of just 289 depositories. (I am counting the 213 credit unions belonging to the Desjardins co-operative federation as one entity.)**

The rule of ten particularly fails with respect to banks. Canada has just 81 banks, not 412 as suggested by the rule. Banks are more influential than credit unions because they tend to be much larger.


So why is this data relevant to payments? A payments network is really only useful if it has a lot of participants on it, but a lot of participants aren't going to be on it in the first place if it isn't useful. That's the chicken-and-egg problem of payments networks.

To solve the chicken-and-egg problem, it helps to have a few large actors a vanguard commit to using the network at the outset, which kickstarts its usefulness, and then everyone else gets dragged into joining up. Voila, universal payments.

When you've got 9,466 depository institutions, it's hard to build a strong vanguard group in order to drive quick adoption of a new payments network.

Take The Clearing House's Real-Time Payments (RTP) network, for instance, a U.S. payments network which was launched in 2017. Out of the U.S.'s 9,466 depositories, RTP has attracted just 285 participating institutions, effectively limiting RTP's reach to 65% of all U.S. checking accounts. (The 65% number is from RTP's website.)*** That's not bad, but it's not great.

Another example is Zelle, a U.S. bank-owned person-to-person payments network that was introduced in 2017. By 2021, Zelle boasted 1,700 banks and credit unions on its bank-to-bank payments network. That's better than RTP, but according to Zelle, this still only represented 74%, or 577 million of all U.S. checking accounts, in 2021. (As of early 2023, Zelle reports having 1,900 financial institutions on its network, so it probably now connects 75-80% of all U.S. checking accounts.)

In Canada's case, with just 81 banks and 208 credit unions, it's much easier to build a vanguard group to drive a payments network forward.

For instance, Interac e-Transfer is the Canadian equivalent to Zelle, providing instant person-to-person transfers via bank and credit unions. As of 2023, Interac e-Transfer has 250 participating banks and credit unions. (It lists Desjardin Group, a federation of 213 credit unions, as a single entity). That's almost all of Canada's 289 depositories, and effectively 100% of all Canadian chequing accounts. That's ubiquity for you.

Admittedly, Interac e-Transfer has been around a lot longer than Zelle and RTP, having debuted in 2003, and so it has had more time to spread into all the cracks. (I wrote about Canada's big head start in instant payments a few years ago.) But even at the outset of the adoption process, e-Transfer enjoyed buy-in from Canada's five biggest banks (Royal, TD, Scotiabank, CIBC, and Bank of Montreal), which together owned 86% of all Canadian banking assets at year-end 2003. That's a huge vanguard group. The chart below, which uses 2022 data, gives a good feel for how significant this is.

The above chart also illustrates how small any U.S.-equivalent vanguard group will ever be. Zelle's 2017 group of 30 first-adopters may have seemed large on the face of it. After all, it included America's largest banks: JP Morgan Chase, Bank of America, Wells Fargo, Citibank, US Bank, PNC, and Capital One. Yet this vanguard still only constituted 52% of total U.S. banking assets, much less than the 86% committed to Interac e-Transfer on day one.

The diffuse nature of U.S. banking (and the concentrated nature of Canadian banking) will play into the upcoming launches of FedNow and Real-Time Rail (RTR), two instant retail payments system belonging to the Federal Reserve and Bank of Canada, respectively. I'd expect RTR usage to amp up quickly, given that Canada's big-5 banks will likely help sponsor it. FedNow adoption will lag. It's just not that easy to get 9,466 institutions on the same page.


* Number of US banks and savings/thrifts is from FDIC. Data on credit unions is from the NCUA
** Number of Canadian banks is from OSFI. Number of credit unions is from CCUA
*** A tweet where I list my data source for RTP data

Thursday, April 20, 2023

How stablecoin opacity and sloppy reserve management paid off (for now)

[I wrote this article for CoinDesk earlier this month and am republishing it here.]

USDC Boasted Transparency but It Didn't Help When Silicon Valley Bank Got Into Trouble

The weekend of March 10, 2023, was a profound test of how well stablecoins hold up under pressure. Now that everything has settled, some odd lessons have been passed on, namely: Transparency doesn't seem to be a good thing. And forget about prudent management of reserves – it's just not worth it.

Opacity and sloppy reserve management win the day. Or, at least, so it would appear.

Leading up to Friday, March 10, the issuer of second-largest stablecoin USD coin (USDC), Circle, was probably the industry's most transparent issuer.

It provided daily updates to investors through its BlackRock-managed money market fund, which backstops the stablecoin. On top of that, Circle had just adopted the New York Department of Financial Services’ guidance for stablecoin transparency, which required two attestation-of-reserves tests each month.

In contrast, Circle’s arch-competitor, Tether, which publishes attestation reports on a less-frequent quarterly basis, lagged far behind on transparency.

To boot, in its attestation reports Circle disclosed all sorts of useful information to users, such as each individual Treasury bill’s CUSIP number and where it banks. On that list was Silicon Valley Bank.

It was the last bit of data – Circle's banking relationships – that seems to have caught everyone's attention that Friday. After experiencing a run through most of the week, Silicon Valley Bank shares were halted at 9:30 a.m. local time after plunging 62% in premarket trading. Just before noon, the Federal Deposit Insurance Corporation (FDIC) announced that it would be shutting the bank down.

Keen-eyed social media commentators scanning Circle's disclosures noticed the mention of deposits held at Silicon Valley Bank. Tweets were issued.

They certainly had cause for concern. When a bank fails and the Federal Deposit Insurance Corporation (FDIC) takes it over, depositors are only protected up to $250,000 per account. Anything above that amount is at risk. The implication was that if Circle had funds stuck at Silicon Valley Bank, it could suffer big losses. That meant potentially being insolvent. And that raised the possibility that USDC holders might not be made whole.

Social media began to demand a statement out of Circle. CoinDesk picked up on Circle's Silicon Valley Bank problem after lunch. Curve's massive 3pool, an important source of stablecoin liquidity, began to be drained as fearful traders swapped their USDC for USDT. By that evening, 3pool was empty and Binance, the world's largest crypto exchange by trading volume, suspended 1:1 conversions between Binance USD (BUSD) and USDC, indicating a significant amount of stress in the market.

By 7 p.m. ET, a slight USDC depegging from the U.S. dollar occurred on trading markets, and after 10 p.m. Circle finally issued a statement. It revealed that $3.3 billion of USDC’s reserves was in limbo at Silicon Valley Bank. The market was stunned. USDC's price began a sickening plunge to below 90 cents.

The irony of this is that neither of Circle's competitors, Tether and Paxos, disclose where they bank. And so commentators on social media didn't have enough dirt on Tether and Paxos to start asking questions. While USDC collapsed on exchanges, the prices of the Tether and Paxos stablecoins held solid.

Had Circle been as opaque as its competitors, no one would have known that Silicon Valley Bank was its banker and the weekend run on USDC probably would never have occurred.

The lesson would seem to be: Don't be transparent or, if you need to be transparent, don't be transparent about your shortcomings.

How might Circle have managed its reserves differently?

Paxos, which issues the Paxos dollar (USDP) stablecoin and, until recently, BUSD, offers some cues. According to the Paxos attestation reports on USDP, Paxos keeps hundreds of millions worth of deposits with banks, but all of those deposits are insured.

It bypasses the $250,000 limit in two ways.

First, some of Paxos' deposits are invested through placement networks. The way this works is that Paxos' bank farms money out to other partner banks in $250,000 blocks. Each of these blocks is completely covered by FDIC insurance. Although there are 4,333 FDIC-insured banks in the U.S., providing a theoretical coverage ceiling of $1.08 billion, in practice Paxos only uses deposit networks for part of its deposit balance.

For the remaining unprotected part, Paxos has contracted with an insurance company to be covered by private deposit insurance. Of the $270 million in cash reserves used to back USDP going into the crisis, $72 million was privately insured.

And that, folks, is how to prudently manage large cash balances. It's a pain. You can't just casually stash your billions at a bank; you've got to go out of your way to properly secure it.

Which leads us into the second irony. If it didn't pay for Circle to be transparent, it also didn't pay for Paxos to be prudent.

On Sunday evening, March 12, the FDIC announced that the $250,000 limit on insurance would be waived. All deposits held at Silicon Valley Bank would be extended a blanket guarantee. Circle's $3.3 billion was safe. In moments, the price of USDC rocketed back up to its $1 peg.

The crypto sector had just benefited from its first federal bailout, and not a small one at that. According to FDIC, the 10 largest deposit accounts at Silicon Valley Bank held a combined $13.3 billion, implying that Circle was the largest beneficiary of the bailout.

The moral of this part of the story is that the government's official cap of $250,000 was never very serious; unofficially, FDIC protects everything. Paxos' careful deployment of private insurance and deposits networks seems to have been a waste of time and resources, and its competitors' "don't think, just deposit" strategy the right one.

In the aftermath, Circle now advertises USDC as "a stablecoin with GSIB cash." That means no longer keeping a big chunk of its cash reserves at mid-size banks like Silicon Valley Bank but lodging most of it at Bank of New York Mellon, a global systemically important bank, one almost certain to benefit from a bailout should it fail. That also means Circle probably won’t bother going through the process of negotiating private deposit insurance.

As for Tether, which issues the least transparent of the big stablecoins going into March 10, it wasn't terribly prudent, either. Its public attestation reports give no indication that it routes its $5 billion or so in cash through deposit networks, nor does it resort to non-FDIC insurance. It fully absorbs the risk of its bankers going bust.

Yet, the amount of USDT in circulation has exploded by around $9 billion, or 11%, since that weekend.

On a long enough timeline, another significant stablecoin test, like the one that faced USDC, is inevitable. The reasons for the next one will be difficult to predict, and likely different from the last one. While opacity and a nonchalant approach to reserve management may not have been punished this time around (indeed, they seem to have been rewarded), if issuers internalize these lessons, then the next stablecoin crisis will only come sooner, and at a much larger scale.

Wednesday, April 19, 2023

In which I catch Canadian banks paying more interest to customers than US banks do

TD Bank operates on both sides of the border, yet pays more in interest on one side than the other. Source: TD

Long-time readers will know that I like to muse on the competitiveness of U.S. and Canadian banking. In my last post on the topic, I was surprised to see how Bank of Montreal's net interest margins a measure of how much a bank is squeezing out of its customers were far lower in Canada than the U.S. The fact that Bank of Montreal is squeezing more out of Americans than Canadians suggests that competition is stiffer north of the border than south of it.

The idea that Canadian banking is more competitive than U.S. banking goes against what I'll call the "standard view." In short, this view is that while Canadian banks are safer and better-regulated than U.S. banks, this comes at a steep price. Up here in Canada, we've ended up with a few big oligopolistic institutions capable of charging exorbitant fees and paying unnaturally low interest rates, which hurts the consumer. Meanwhile, there are many more banks in the free-wheeling U.S., and while this makes for more bank failures and runs, the public benefits from lower fees and receives higher interest rates.

Anyways, I recently stumbled on another anecdotal piece of evidence that contradicts the standard view. TD Bank operates on two sides of the border, as the screenshot at the top of this blog post shows. Yet it pays depositors a different rate, depending on what country they live in:


As the chart shows, TD Bank consistently pays higher interest rates to its Canadian depositors. Why a persistent interest rate differential? You can't blame it on central bank policy rates being higher in Canada, since over much of this time frame policy rates were higher in the U.S. Is it possible that TD has to be more competitive in Canada in order to attract deposits?

An alternative explanation for the differential is that TD's deposits are of a different character in Canada. A big chunk of TD's Canadian deposits are fixed-term deposits that mature in 12-months or more, whereas its U.S. base of fixed-term deposits is typically shorter term, usually 3-12 months. Because it costs more in interest to convince depositors to stick around for longer periods of time, could it be that it is the difference in term and not competition that explains why TD's interest costs are so much higher in Canada?

Not quite. Even if we compare U.S. and Canadian interest rates for the same fixed term, Canadian banks still pay more interest to depositors. In Canada, the term of art for a fixed-term deposit is a guaranteed investment certificate, or GIC. In the U.S., it's a certificate of deposit, or CD. In the chart below, I've charted out the interest rate that Canadian banks pay for 1-year GICs compared to what U.S. banks pay for a 1-year CDs.


(For more on the data, see footnote*). 

You can see that over the last few years, the big-6 Canadian banks have consistently paid more to 1-year fixed-term depositors than U.S. banks have. Again, you can't pin this on central bank policy rates being higher in Canada. 

Canadian banks not only consistently pay more interest, they are also far more responsive to central bank policy rate increases. Both nations' central banks, the Fed and the Bank of Canada, began to hike rates in lockstep with each other beginning in March 2022, starting from close to 0% and rising to 4.75% and 4.5% respectively as of today. Yet Canadian banks began to pass-off these increases months before U.S. banks did. They have also done so far more completely; the 3.0% on a 1-year GIC is far closer to central bank policy rates of 4.5%-4.75% than the 1.5% on an equivalent CD. (And by the way, I wrote about sticky U.S. deposit rates last year.)

Could it be (gasp) that Canadian banks are more competitive?

Trust me, I don't like this conclusion. I quite enjoy thrashing Canadian banks for being noncompetitive. So if you have some good counter-evidence, please send it my way.

By the way, the above data confirms an anecdote from last year. I caught Canada's largest bank, Royal Bank, paying much more to Canadian depositors than U.S.'s largest bank, Chase, pays its American depositors:

It's very possible that the standard view is wrong, and the tug of war between Canada's 6 nationwide heavyweights result in a more competitive price than in the U.S. where although there are more than 4,000 banks they are often small and regional and lack the heft to engage in high calibre competition.

If so, it appears you can have your cake and eat it too. Not only can a country have a safe and robust banking system, but that needn't come at the price of less competition.



*A note on my data sources for this chart. Canadian data comes from the Bank of Canada, which is compiled from posted interest rates offered by the six major chartered banks in Canada. The number is the statistical mode of the rates posted, which may explain the series' choppiness. U.S. data comes from FDIC, which compiles the data from S&P Capital IQ Pro and SNL Financial Data. Certificate of deposit rates represent an average of the $10,000 and $100,000 product tiers. Averaging across dozens or hundreds of banks would explain the smoothness of the U.S. series.

There are a number of complexities that are not explained by either data source. For instance, is the Bank of Canada's GIC data made up of non-redeemable GICs only, or do they include redeemable GICs, too? As for the U.S., banks like Chase offer a "relationship rate" to customers that far exceeds the non-relationship rate. Which rate is the FDIC collecting? I've suggested in my blog post that the difference in U.S. and Canadian 1-year fixed term deposits rates could be explained by competition, but it could also come down to data artifacts like these.

Thursday, April 13, 2023

Payments stablecoins vs trading stablecoins

Circle's Gordon Liao recently sketched out a new distinction between 'payment stablecoins' and 'trading stablecoins,' and then places Circle's stablecoin, USD Coin, in the former category, while confining competitors Tether, Dai, and Binance USD to the trading bucket.

I don't know about you, but I'm not convinced. 

According to the article, a payments stablecoin is defined as a stablecoin that has a low ratio of daily trading volume to circulation and has little correlation to the price of Ethereum. The suggestion is that any stablecoin that sports these statistics isn't being used for cryptocurrency speculation, and so by default it must be getting used for payments. By contrast, any stablecoin that has a high trading-volume-to-circulation ratio and is more closely correlated to Ethereum's price is defined as a trading stablecoin; its primary function is speculation, not payments.

You can sorta see where this is going. Speculation is unsavory whereas payments (the article cites cross border remittances in particular, but you can also put retail point-of-sale payments in that category) are wholesome & useful. As a stablecoin issuer, it's probably better to be slotted in the payments bucket, since that's the bin that regulators, politicians, critics, and economists will take more kindly to. Who knows, it might even merit a more beneficial regulatory touch.

And so no surprise that the article finds that USD Coin, or USDC, qualifies as a payments stablecoin, one that has "minimal speculative exposure."

Having watched stablecoins markets for a while now, my internal library of anecdotal evidence tells me that USDC isn't used much in payments, but mostly for trading. It's just that one of the big roles USDC plays in trading is a relatively sedentary one, as a form of collateral in decentralized finance (or DeFi), and so its turnover is relatively low. By contrast, stablecoins like Tether and Binance USD are less popular as DeFi collateral and more popular as trading chips for centralized exchanges (like Binance), and that's why they have such high turnovers.

My anecdotal database also tells me that the most payments-ish of the stablecoins is probably Tether. Don't get me wrong, Tether is still mostly used for trading, specifically as a proto-dollar on exchanges like Binance and Bitfinex to support crypto gamblers. But when you do hear stories about stablecoins being used for cross-border payments, it tends to be Tether that's involved, not USDC.

For the time being, I think that a payments stablecoin is a fable. Everyone wants to be in that category, but an actual payments stablecoin, one who's main use-case is remittances and POS payments, doesn't exist. Stablecoins remain primarily used for trading, gambling, and speculation.

Saturday, April 1, 2023

FedNow is not a tool for freezing money

I've been seeing a lot of misunderstanding about FedNow, a new retail payment system built by the Federal Reserve (the FED), but the one I want to deal with right now is the allegation that FedNow has the ability to freeze accounts.

Balaji Srinivasan, for instance, suggests that FedNow could "facilitate Cyprus-style deposit seizure at the speed of light":


That's not quite right, and let me show you why. 

Let's set the stage. Say you do something illegal. Maybe some light treason, or better yet, investment fraud. You've got all of your victims' funds in your Wells Fargo account, and law enforcement wants to quickly stop you from making a getaway.

The authorities won't bother sending a court order to the FED to freeze your money. FedNow is a communications system between member bank, allowing them to quickly and accurately update their databases. It has no control over Wells Fargo's database, which is where the stolen funds reside.

To freeze the funds, the authorities have always had a much more powerful tool at their disposal. They will send a court order directly to Wells Fargo. And then Wells Fargo will lock your account. At that point you'll no longer be able to ask your bank to send FedNow transfers (or any other sort of transfer for that matter.) The FedNow network is closed to you, my friend.

But it wasn't the FED that carried out the freeze, it was your bank.

Next, let's broach the allegation that authorities might be able to order the FED to block any FedNow message mentioning your name, in essence marooning your funds at Wells Fargo. In his Twitter exchange with me Balaji seems to think this is possible, although he provides no real proof of his claim.

Even if this was possible (and I doubt it is), that's a fairly ineffective tool. You could get around a FedNow blockade by having Wells Fargo transfer the stolen funds via any non-FedNow network. There are many of them, including one of the ACH networks, Zelle, The Clearinghouse's RTP system, or one of the Visa/MasterCard debit networks. Alternatively, wire it overseas via SWIFT. Or send the funds internally to a confederate's Wells Fargo account and have them evade the blockade. Or, if you're feeling old fashioned, you could withdraw cash, maybe even write out a paper check. How about using your debit card to buy gold or crypto or a yacht if that's your thing?

As you can see, you've got so many ways to exit stolen funds from your Wells Fargo that blocking just one of the holes, FedNow, provides the authorities with no extra ability to freeze funds. FedNow or not, the authorities have the same very powerful tool they've always had: block funds at the bank level by deputizing bankers to do it.

(And that's not a bad thing. Freezing funds before they can get dissipated is one of the best tools for protecting fraud victims.)

As further proof that I'm right about FedNow's alleged freezing ability being a nothing-burger, here's a reality check: When 280 convoy-linked individuals were frozen out of the Canadian payments system by the Federal government last year, did the authorities ask the Bank of Canada to configure its various payments systems to carry out this order? 

No, they didn't. They delegated freezing responsibility to Canada's banks and credit unions. And guess what? It worked like a charm.

In summary, let's not get too worked up about FedNow's alleged freezing ability. It's not the tool for the job.