Thursday, December 26, 2019

The Watergate banknotes

Cash isn't quite anonymous, it's anonymous-ish.

To illustrate this, a few years ago I wrote about the 1932 Lindbergh kidnapping case. The ransom was paid in gold certificates, not Federal Reserve notes. By coincidence, the U.S. went off the gold standard the next year, and all gold certificates were called in. So when the kidnapper spent some of his gold certificates in 1934 to buy gas, his purchase was odd enough to out him to the authorities.

I recently stumbled on a more recent example of cash de-anonymization. Most people know the gist of the Watergate scandal, but to recap five burglars were caught breaking into Democratic headquarters at the Watergate building on June 17, 1972.

Who were they and what were they doing there? At first, no one had a clue. But the police did find around $3,600 in cash on them, much of it in sequentially-numbered $100 banknotes. See the serial numbers below:

Testimony of Paul Leeper, May 1973, Hearings Before the United States Congress, House Committee on the Judiciary [source]

A series of sequential banknotes meant that the cash had come fresh from the U.S. Bureau of Engraving and Printing, the agency that produces banknotes on behalf of the Federal Reserve. The new notes would have been sent to a bank which in turn distributed the notes in their original sequential order to customers.

The serial numbers of the Watergate notes also gave a geographical sense of where they had been issued. There are 12 regional Federal Reserve Banks. Notes beginning with F are issued into circulation by the Federal Reserve Bank of Atlanta, those beginning with C by the Federal Reserve Bank of Philadelphia.

Two days after the break-in, FBI agents contacted these two district banks for more information. It turns out, Federal Reserve Banks do keep track of banknote serial numbers. The Philadelphia Fed informed agents that the notes in question had been shipped to a private bank, the Girard Bank & Trust in Philadelphia, while the Atlanta Fed had shipped theirs to the Republic National Bank in Miami. 

Unfortunately, the two commercial banks did not record the serial numbers of the bills that they had distributed to the public. However, one of the burglars--Bernard Barker--happened to have an account at the Republic National Bank in Miami. (The trail to the Girard Bank & Trust turned cold).

Scanning through Barker's bank information, investigators discovered that several months before five large deposits had been made into his account. This included four checks totaling $80,000 drawn on a Mexico City bank and one for $25,000 from a Miami bank. These funds were eventually traced back to the the Committee to Re-Elect the President, an organization created to help raise funds for Nixon's upcoming election campaign.

And there was a smoking gun. A money trail from the pockets of the Watergate burglars to the President's administration. The burglars, it was further discovered, had been hired by Committee to Re-Elect the President to wire-tap Democrat party phones and photograph documents. So the President was spying on his political enemies.


There is an interesting sidebar to this story. Two days after the break-in, Senator William Proxmire, chairman of the Financial Affairs Subcommittee, contacted the Fed for information about the banknotes. In a book published in 2008, economist Robert D. Auerbach accused Arthur Burns, Nixon's appointee to lead the Fed, of refusing to cooperate with Proxmire's requests. Presumably Burns wanted to protect his boss.

In his book, Auerbach cited the following internal document, a timeline of the Federal Reserve Board's actions after the Watergate break-in. It is available in the Arthur Burns papers in the Gerald R. Ford Presidential Library:

Congressman Ron Paul aired Auerbach's allegations in front of Congress in 2010. An investigation was soon initiated by the Office of the Inspector General, an independent body that conducts oversight and audits of the Federal Reserve. In 2012, the OIG exonerated Arthur Burns and the Fed. The report noted that Burns was complying with FBI requests to avoid sharing the information lest this interfere with the investigation.


So banknote users are never entirely anonymous--the serial numbers can be used to unveil who they are. In Watergate's case it was a fairly blunt tool. The notes could only be traced to the burglar's Miami bank, not to his account.  But if tellers at the Republic National Bank had been dutifully recording the serial numbers of notes they gave to their customers, the tool would have been much more accurate, pinpointing Barker as the direct source of the Watergate banknotes.

No bank teller want to tediously record numbers by hand. But in today's world, it is technically possibly for ATMs to record banknotes using built-in serial number readers. Is this actually happening? I am pretty sure that serial numbers are not being collected by North American banks. People would be furious about potential invasions of privacy.

But not so in China. Since 2013, Chinese ATMs and tellers are required by law to record the serial numbers of all banknotes. (I am not sure if the same law applies to Hong Kong):

I snipped the above screenshot from a marketing brochure from Glory Ltd, a Japanese company that specializes in cash handling technology. Chinese laws surrounding banknote serial number collection have ostensibly been put into place to prevent counterfeit yuan from entering into circulation. But one could imagine this technology being used by the authorities to track licit money flows.

Say that a Chinese human rights activist deposits ten ¥100 banknotes into their bank account. The police might be curious about who is financing this activist. In theory, they could ask the People's Bank of China, the nation's central bank, to search the various serial number databases for the name of the owner of the bank account from which the ten ¥100 notes originally came. In this way an anonymous cash donation could be de-anonymized.

Chinese citizens who use cash in potentially risky transactions have probably already devised a solution. They can evade serial number sniffing by going through an extra step of "mixing" their cash prior to spending it. This might involve breaking up the notes in a few shops prior to passing them on to the intended recipient. Of course, this trick will only work as long as retailers are not required to install their own note-reading hardware.

I often write about the contradictions of anonymous payments. It would have been great to catch Nixon's thugs with technology that completely de-anonymizes banknote movements. But it is abhorrent to know that human rights activists might be prosecuted using this same technology. Striking a balance is difficult.

Thursday, December 19, 2019

Buying coffee with Tesla shares

It's fascinating to see how brokerages these days are offering no-commission trades, fractional share ownership, and debit card-linked accounts. With this combination of features, maybe we're getting closer to the day when we can buy a $2.50 coffee with 0.007 Tesla shares.

Right now, a debit card purchase can only proceed if there are uninvested cash balances in the linked-to account. But what if the securities held in your brokerage account could also be debit-cardized?

Imagine going to Tim Hortons, ordering a double double, and paying with your RobinHood MasterCard debit card. Behind the scenes RobinHood, an online brokerage, checks your account. All you own is a few shares of Tesla. RobinHood won't actually transfer the shares to Tim Hortons. Instead, it quickly sells a small fraction of these—0.007 shares—for $2.50 cash.

Since RobinHood doesn't charge commissions, selling the shares costs you nothing. MasterCard signs off on the transaction and presto, you've got your coffee. You own 0.007 fewer Tesla shares while Tim Horton's will soon get $2.50 in cash from RobinHood.

Stock markets aren't open on the weekend. So what happens if you want to buy groceries on Sunday? Maybe you've got 2.1 shares of Tesla in your RobinHood account. They were worth around $825 at Friday close. Something catastrophic could occur over the what remains of Sunday, but RobinHood is pretty sure that come Monday morning, those shares probably won't be worth less than $500. And so it will allow you up to $500 in weekend debit card payments. When the market opens on Monday it sells whatever Tesla shares are necessary to settle up your grocery purchase.      

If the option of paying with volatile assets like Tesla were to be widely adopted, you'd expect traditional banks to get into the game. Right now banks offer deposits denominated in fiat units like dollars or yen or pounds. But there's no reason they couldn't provide Tesla-denominated checking deposits. The fact that banks don't do this is a good tip-off that there isn't a very big demand to make transactions using volatile instruments.

Why do people prefer to pay for things using stable instruments rather than volatile ones like Tesla shares? My guess is that it has something to do with FOMO.

Given a choice between paying with their regular bank debit card or a RobinHood card, most people will choose their regular card. Spending away Tesla shares could mean that they miss out on a potentially big jump in price. But spending away fiat-denominated deposits doesn't produce any negative emotions, since deposits can always be replaced at the exact same price come next week's paycheck.

(As I suggested last year, this is a weird example of Gresham's law, where lottery-type instruments like Tesla don't get recruited as money because the market puts less value on them than a hopeful individual does.)

If no one wants to use Tesla shares to buy coffee, they might prefer to set up their RobinHood debit card to sell lower-risk securities. For example, a RobinHood customer could have their card draw down on a bond ETF like the iShares Short Treasury Bond ETF (SHV), which primarily invests in US Treasury bills.

Since SHV's price hardly fluctuates, anyone who uses SHV units to buy coffee needn't fear missing out on a big payday.

At the same time, they'd earn far more than they would on checking account. SHV currently pays around 1.68%, which after a 0.15% management fee comes out to around 1.53%. That's about the rate you could get on a high-interest saving account, which aren't usually designed for everyday spending.

Will this sort of debit-cardization of stocks & ETFs ever happen? I don't know. There could be regulations that prevent the practice, or maybe some sort of hidden cost that makes it too expensive. On the other hand, cryptocurrencies and gold have already been debit-cardized, the gold and bitcoins being sold the moment that a card purchase is initiated. I don't see why it wouldn't be technically possible to do the same with other exchange-traded liquid assets like stocks.

Crypto-linked cards haven't been very successful, probably due to the FOMO problem I mentioned earlier. (Last year, Coinbase shut down its Shift crypto card). Tesla shares would probably suffer from the same. But a low-risk ETF held in a Robinhood account wouldn't be quite so hobbled. 

Thursday, December 12, 2019

Bitcoin and sanctions

I recently watched a video with Alex Gladstein on the importance of financial privacy. In general I agree. We should be working on expanding the scope for transacting privately, although I am conscious of the tradeoffs. Anonymity helps good people evade bad rules, but we need to be wary of how it abets bad people evading good rules. (See for instance my recent post on the good & bad of using prepaid debit cards to donate anonymously).

In the above list, Gladstein intimates that bitcoin has a positive role to play in evading U.S. sanctions. I have two quick points to make.

I mean, there are U.S. sanctions that I agree with and those that I don't agree with, and I'm sure the same goes for Gladstein. I hope that the sanctions that I agree with are in fact the morally justified ones, and the ones I don't agree with are the immoral ones. By my reckon sanctions on the apartheid regime in South Africa were justified, and same with the ones on North Korea and Zimbabwe government officials. Those on Cuba and the recent ones on Iran are not.

One (admittedly-blunt) sorting mechanism for determining the morality of U.S. sanctions is how much international consensus there is on levying them. If plenty of nation's support sanctioning a regime then the odds that the target is a genuinely bad actor are higher than if just the U.S. thinks so.

Trump's recent round of Iran sanctions has almost no international buy-in. America's European allies are furious that the U.S. left the Iran nuclear deal, a carefully negotiated agreement to control Iran's access to nuclear technology. The Chinese and Russian are upset too. On the other hand, Obama's earlier round of Iran sanctions had broad support. Even Russia and China were on board.

So if consensus is a reasonable hurdle for judging sanctions, then Trump's Iran sanction don't pass muster, but Obama's passed the smell test.

All of which is to say that if bitcoin is indeed an effective tool for evading the current round of Iranian sanctions, then it had a negative role to play as spoiler to the previous round of "good" sanctions. Bitcoin might have delayed (or prevented) the 2015 Iran nuclear deal that did eventually emerge. Which would have been unfortunate.

If we do care about the morality of sanctions, bitcoin doesn't really solve anything. We need to get the sanctions correct at the outset. Don't like Trump's Iran sanctions? Try to convince your neighbours about it. Tell your American friends. Go to a protest. Dial up your government representative. Yep, it's an incredibly blunt tool. But it's the best we got.

The second point I want to make concerns how useful bitcoin actually is as a sanctions buster. In his presentation Gladstein mentions Ziya Sadr, an Iranian who can't use Visa or PayPal but can use bitcoin.

Sure, Bitcoin may give some tech-savvy Iranian freelancers a means to connect to external buyers. But it hasn't helped where it really counts. It hasn't allowed Asian refiners to keep buying Iranian oil or European manufacturers to keep their factories running. Pretty much every foreign company has stopped dealing with Iran. As a result, the nation's oil exports have cratered and its economy has gone into a tailspin. This has had a tremendously negative effect on regular Iranians.

Let's not just single out bitcoin for being ineffectual. The euro, the world's second largest currency, has also been a useless sanctions buster.

Last month I wrote a post about why U.S. sanctions are so effective. Let me give a brief recap. A foreign company--say a European refiner--currently has to choose between continuing to buy crude oil from Iran or no longer accessing U.S. markets. This means not only being shut off from U.S. energy exports, but also doing without U.S refining technology, expertise, and financial access. Any refining executive who ignores the sanctions could be blacklisted from entering the U.S. for travel, or sending their kids to U.S. schools, or getting U.S. medical care.

Source: BBC

And so given an explicit choice between the US and Iran, most companies have chosen the US, since it has much more to offer. In our globally interconnected world, the population of willing-to-be-sanctioned companies is pretty much non-existent.

Having some sort of alternative money like euros or bitcoin doesn't provide much of a work-around.

Say that our European refiner decides to do a bit of business with Iran under the table in euros (or bitcoin) rather than dollars. All sorts of people will touch this transaction, not just the payment side but also the movement of crude. A banker, a bitcoin exchange, or a freighter captain could rat the refiner out to the US Justice department. And so the refiner would be fined or even worse blacklisted, which would means losing all access to US resources. The risk is simply too high.

In sum, the case for bitcoin as a sanctions buster is not clear-cut. Genuine evil leaders probably should be sanctioned, the less ways to short-circuit the blockade the better. And given the way that U.S. sanctions are structured, bitcoin may not provide much help anyways.

Saturday, December 7, 2019

A way to make anonymous online donations

Paying for things online usually means giving up plenty of privacy. But this needn't always be the case. Last night I donated to a local charity via their website and didn't have to give up any of my personal information.

The trick for achieving a degree of online payments anonymity? Not bitcoin, Zcash, or Monero. I used a product created by old fashioned bankers: a non-reloadable prepaid debit card. (I wrote about these cards here and here).

Had I used a credit card or PayPal, all sorts of parties would have gotten access to my personal information including the site owner, the payments processor, my bank, the site owner's bank, the credit card networks, my partner, and many more. To get a good feel for how many different parties touch an online payment, check out this graphic by Rebecka Ricks, which shows how PayPal shares your information.

I bought my prepaid card--a Vanilla card--with $25 cash at a pharmacy. For it to be usable online I had to register it at Vanilla's website. That meant inputting my postal code. But that's all the information that Vanilla asks for. In my case I used my actual postal code, but I doubt that the system would have protested if a privacy-conscious user were to submit the wrong one.*

So at this point I've got a fully-loaded online-enabled card that has not been directly fed any information about my identity. (Note that this is how the process work in Canada. It may be different in the U.S. and elsewhere).

Next step, choose a charity. At the charity's website I entered my Vanilla debit card number, the CVV, and $10 as my amount to donate. The site also asked me to enter the name on the card. Because a prepaid card only says "For You" on it and not your name, just enter that or John Doe. Voila. Payment made:

Why on earth would anyone want to make an anonymous online donation? For my part, I was simply experimenting with my prepaid card. But I can think of several licit reasons for why people might want to donate anonymously with prepaid debit cards:
  1. Many people share bank accounts. They might not want their partner to know that they are donating to a cause that their partner might not support.
  2. A donor may not want the donee to know their identity lest the donee use it in a way that hurts the donor. For instance, if in public life I am a well-known conservative Evangelical, but I donate to a cause (say abortion education) that I privately support, I might prefer avoiding any chance that the donee leak my information in an attempt to 'out' me.
  3. I like the charity, but don't trust it or its chosen payment processors to protect my information from hackers.
  4. I don't want the charity to have my information so it can't inundate me with spam.
If non-reloadable prepaid cards can meet people's legitimate privacy needs, there is also a nefarious side to them. Anonymity allows people to evade rules about donations. For instance in Canada, there is a certain type of donation that is highly regulated: political donations. Below I've listed a few keys regulations:
  • No cash donations above $20
  • No anonymous donations above $20
  • The identities of contributors that have given $200 or more must be reported to Elections Canada, which will publish them
  • No single individual can contribute more than $1600 in a year.
Canadians have good reasons for supporting these limits. We don't want wealthy people to have an outsized influence on politicians. And we want donations to be transparent so we can see how politicians might be influenced by certain donors.

I can imagine plenty of scenarios in which motivated donors may want to break these rules. Say that a set of business owners in the restaurant industry stand to profit if the Liberal candidate wins because she supports removing regulations that increase restaurant operating costs.

After legally donating $1600 to the Liberals, some less savoury restaurant owners might want to illegally funnel more funds into party coffers. Cheques, credit cards, and other banking routes would be too risky. They establish a clear connection between the owner's identity and the donated funds.

A motivated restaurant owner can instead use $10,000 in cash to buy prepaid debit cards. They then go to the Liberal's website and donate $199.99 fifty times (for a total of $9999.50) using bogus names like John Doe, Jane Doe, etc. Since each transaction is under $200, they won't trigger the rule that requires such donations be reported to Elections Canada. And the Liberal Party probably doesn't have the capacity to cross-check each of the fifty payees to verify that they are associated with real identities.

This rinse-and-repeat strategy highlights one of the ambiguities of prepaid regulations. To reduce the potential for fraud and money laundering, regulators in Canada and the U.S. disallow non-reloadable prepaid cards with a face-value in excess of $500 (I believe that's the number). But since these cards are relatively anonymous, there's nothing preventing a would-be fraudster from using multiple cards to get around the cap.**

In any case, I'm not saying that this sort of donation fraud is occurring. But it's plausible. There's a reason that gift card and prepaid card fraud is rampant in North America. The relative anonymity that cards offer makes them a tempting tool for criminals.

As always, there is a yin/yang nature to anonymous payments. Anonymity is great when it protects well-meaning people from harm, but not so great when it protects bad people from good rules. Striking a balance is tricky.

* To access Vanilla's website, I had to disable my tracking blocker. Which means that the website probably has all sorts of processes going on in the background while a Vanilla user enters their postal code. These processes could link the user to their identity by cross-referencing the data gleaned by Vanilla's trackers against other data that has been collected elsewhere. This is probably an issue for these who want all-out privacy, and steps would have to be taken to mitigate information leakage. As they say, there is probably no such thing as pure anonymity, only degrees of anonymity. But for anyone who simply wants to enable a prepaid card in order to prevent their partner or the donee from seeing their transactions, then it's probably not a big deal.

**The way to nip donation fraud in the bud would be to require payment processors to avoid processing any prepaid debit card payment for political parties, or to limit cards to some inconvenient amount like $5 so that a rinse-and-repeat strategy is too costly to perform. It is possible that this tactic has already been adopted by payment processors.

Wednesday, December 4, 2019

Mooning over daylight overdrafts

Every few days for the last month or so I've been refreshing a Federal Reserve page that shows data about daylight overdrafts. For some reason the Fed only updates it every few months. I had been getting quite curious to see what happened during the great September interest rate spike. Well, finally the Fed has uploaded the data.

If you don't know about the September rate spike, I'd suggest reading Nathan Tankus's tweet, listening to David Beckworth's podcast with Bill Nelson, or picking through this blog post from Stephen Cecchetti and Kermit Schoenholtz. In short, there was a sudden increase in the demand for Fed balances (also known as reserves), and the Fed was slow to react by increasing the supply of balances. And so the rate at which banks were willing to borrow balances spiked to desperation levels.

Why did the demand for balances increase?

That's where daylight overdrafts can inform us. By way of background, the Fed has had a long-standing policy of providing banks with daylight overdrafts. These are loans that last a few minutes or hours and are always paid back by the end of the day.

Why overdrafts? Banks often have to make large and unexpected payments to other banks on behalf of their customers. To facilitate this, banks keep balances on hand in payments accounts held at the Fed. For instance, say that Microsoft wants its bank to wire $200 million to Google. Microsoft's bank tells the Fed to debit its account and credit the account of Google's bank by $200 million.

But say Microsoft's bank only has $50 million in its Fed account. How can it make good on Microsoft's request? No problem. It gets a $150 million overdraft from the Fed. Combined with the $50 million it already had, it can now make the $200 million payment to Google's bank. Microsoft and Google are square, but Microsoft's bank still owes the Fed $150 million. Over the course of the day, Microsoft's bank may get enough incoming payments from other banks to settle up. (Or it may have to sell some assets to pay back its overdraft, or borrow from another bank to make good with the Fed).

Banks have to pay a fee for each minute that they borrow from the Fed. They can avoid this fee by providing collateral to secure the loan. Collateralization protects the Fed from fallout should a bank that is in overdraft go bankrupt. Either way, overdrafts are costly to banks. Fees must be coughed up or collateral sacrificed.

In the 2000s, the Fed kept a tight lid on the supply of balances it issued. Banks couldn't keep very much money in their Fed accounts--there just wasn't that much of it. So there was a big demand to get daylight overdrafts from the Fed to facilitate payments like the one between Microsoft and Google.

But during the credit crisis, the Fed created a massive amount of balances. Suddenly, banks could (and did) keep a lot of money at the Fed. And so when Microsoft wanted to pay Google $200 million, Microsoft's bank just didn't have to rely on overdrafts anymore. Odds were that it already had the money in its account. Overdraft usage cratered, as the chart below illustrates. Whereas peak Fed daylight overdrafts typically came in near $250 billion in 2008, by 2013 peak borrowing by banks typically registered below $10 billion.

(Note, to bring out the data I have used a log scale for this chart, which may not entirely convey the degree to which overdraft usage collapsed.)

Which leads back to the September jump in demand for Fed balances. One potential explanation is that banks are concerned about holding sufficient liquidity in anticipation of customer payment requests. Perhaps banks are running into more situations in which they can't immediately satisfy Microsoft's request to pay $200 million to Google, and so they want to build up the amount of balances they keep in their accounts at the Fed.

But if this was the case, we'd also expect to see a big jump in overdrafts. After all, overdrafts are an alternative way for Microsoft's bank to satisfy Microsoft's payment requests. The chart above shows that peak daylight overdrafts clocked in at $19 billion during the two-week period ending September 25, the highest level since 2013. But that's still a tiny amount compared to $250 billion in 2008.

So the big jump in the demand for Fed balances probably arises from something other than a desire on the part of banks to facilitate routine intraday payments flow. Banks still aren't making much use of daylight overdrafts, a good indication that they still have plenty of balances on hand for transactional purposes. Other motivations must be behind the demand for Fed balances, presumably various regulations that might not have existed in 2008 or increase supervisory expectations.