Thursday, October 27, 2016

How anonymous is cash?

Dutch 10 guilder note. Holland and Lebanon are the only countries to have issued banknotes with bar codes.

One of the interesting things that we've all learnt about Bitcoin is that it isn't actually anonymous, it's pseudo-anonymous. While anyone can deal in bitcoins without providing personal information like a phone number or photo ID, all bitcoin transactions are broadcast to the public. By analyzing these transaction patterns, it may be possible to flush a user's true identity out into the open.

Bitcoin is an attempt to digitally replicate many of the features of the old fashioned banknote, but even banknotes are to some degree pseudo-anonymous. Each banknote has a unique serial number on it. By tracking serial numbers, it may be possible to connect a note to a noteholder and thereby destroy their anonymity. The process of unveiling note users occurs most often in kidnapping cases. When their young son was kidnapped in 1932, the Lindbergh family paid a $50,000 ransom in non-sequential banknotes. In an effort to identify the kidnapper, a list of the serial numbers of notes used to pay the ransom was published in the New York Times and circulated in pamphlet form to banks all over the New York area. Anyone who found the note was to immediately alert the authorities, this information being potentially useful in helping to triangulate the guilty party:

Published list of banknotes the Lindbergh's used to pay the ransom

Kidnappers prefer to be payed in non-sequential numbered bills. The Lindbergh kidnapper is no exception, writing in one of several ransom notes: "Don't mark any bills or take them from one serial nomer [sic]." The reason for this is that it's easy for a bank teller to cross reference incoming notes against a list that contains an easy-to-remember range of sequential numbers. When serial numbers are randomized, the list becomes much harder for the human eye to parse; just try to work through the above example. The non-sequential nature of the ransom payment probably explains why only a few of the Lindbergh blacklisted notes were found... least at first. The final pinpointing of the Lindbergh kidnapper really only became possible when Franklin D. Roosevelt decided to temporarily take the U.S. off the gold standard in 1933. Somewhat serendipitously, the authorities who were helping the Lindbergh family had decided to pay the 1932 ransom in gold certificates, a Treasury-issued instrument that was redeemable in a fixed quantity of gold. At the time, gold certificates circulated along with a motley crew of other private and government-issued note types including Federal Reserve notes, U.S. Notes, Federal Reserve Bank Notes, silver certificates, and National Bank Notes (see here).

As part of the process of going off the gold standard, Roosevelt issued Executive Order 6102 requiring all Americans to bring in their gold, gold coins, and gold certificates to be exchanged for Federal Reserve notes. The Lindbergh kidnapper would only tender a few of his gold certificates in 1933, perhaps worrying that bringing in all $50,000 at once would attract attention.

Subsequent to Roosevelt's Executive Order, gold hoarding became an illegal act. So when the kidnapper bought gas with a $10 gold certificate in September 1934, the gas station attendant—probably worried that he might not be able to deposit it—wrote the license plate of the car on the note. Three days later the station managed to deposit the note at its bank where it was successfully cross-referenced against the black list, a much easier process now that the population of gold certificates was so small. Bruno Richard Hauptmann, the kidnapper, had been unveiled.

Using serial numbers to unveil identity requires the cooperation of private banks as well as some luck, in the Lindbergh's case the coincidental alteration of the monetary standard. However, there is no reason that central banks themselves can't be aggressive in monitoring serial numbers. In 1973 the Dutch central bank, the De Nederlandsche Bank (DNB), set up the first real-time database of banknotes in circulation. All banknote serial numbers are registered in the database. As used banknotes are brought into DNB processing points each day, machines read their serial numbers and update the database to indicate that these notes are no longer in circulation. When these same notes are paid out to banks the next day, the system once again updates its database to indicate that they have re-entered circulation. Over time, the system gleans information about the paths taken by each individual note, including how long it stays in circulation and its geographical exit point. It also provides excellent protection against counterfeits. If the DNB detects two banknotes entering its system with identical serial numbers on the same day, then one of them is by definition a fake.

While many central banks were "intrigued" by the Dutch registration system none of them actually implemented the concept (see page 263 of pdf). As of 2012, the DNB  remains the only central bank to register banknotes on a daily basis, a fact which I find kind of shocking. Why have serial numbers if not for tracking? Decoration?

The upshot is that if you had to choose a place to be kidnapped, Holland would probably be it. As long as the serial numbers are recorded by the authorities before the ransom is paid, then the DNB's registration system can be mobilized to catch kidnappers. For instance, the DNB claims it was instrumental in catching the kidnapper of Gerrit Jan Heijn, an heir to the Albert Heijn supermarket empire, in 1987. When the kidnapper spent NLG 250 to buy groceries, the note was soon deposited at the DNB and read into the database, at which point authorities had enough information to trace it back to the commercial bank and then the supermarket.

Interestingly, there are a number of private banknote trackers on the internet, the most well known of which is Where's George. A user logs into the website and registers a U.S. banknote by entering its serial number. When someone else subsequently registers the same banknote, the ‘route’ of the bill is displayed. Where's George tracks around 266 million bills. EuroBillTracker, the equivalent for the euro, tracks around 160 million notes. Below is a map showing the "hits," or connections it has established over the last week:

Hits registered by EuroBillTracker

So cash is somewhat less than anonymous, or anonymous-ish, since behind the curtain an organization like the DNB may be recording serial numbers, and this data might be useful in learning about users' real life identities. By tracking serial numbers more robustly, the anonymity of cash can be further eroded. Imagine a Where's George world where each time a bills is used, the receiver is required to submit the serial number to a government-run central registry. If so, the banknote system would have attained the same level of pseudo-anonymity as bitcoin, where anyone is free to transact using banknotes but transaction chains are fully public.

We could go further and imagine a world where a central bank like the DNB requires that the circulation of high denomination banknotes, say the €200 note, be confined to 'legitimate' channels only. Cash is perpetually being withdrawn from the central bank, used in payments, and then redeposited at the central bank. To confine €200s to legitimate channels, the DNB would simply announce that it intends to limit redeposits to those notes that have fully verified transactions histories. Verification means that when someone receives a €200 note, they must register it by submitting its serial number to the central bank via an app along with some sort of proof of identity.

When someone fails to either register a note or provide adequate identification, that note effectively falls out of the system. After all, because the DNB won't allow a note with an incomplete chain of verified transactions to be redeposited, banks will refuse to accept any note that hasn't been registered by its current owner. And knowing that banks won't accept them, neither will retailers. Bills that have fallen through the cracks will only have value in an alternative black market where they'd likely trade at a large discount to legitimate notes. Incidentally, establishing a verification system for €200s is very similar to Ken Rogoff's idea of abolishing high denomination notes, except instead of withdrawing €200s, they'd be allowed to stay in circulation in 'cleansed' form.

Thanks to a distinctive earmark—their serial number—the anonymity of banknotes is never fully assured. While serial numbers are rarely used these days for tracing, who knows what might happen in the future. Privacy advocates can take some comfort in the fact that, unlike paper money, coins have no distinctive markings and are therefore capable of serving as a purely anonymous exchange medium. Unfortunately coins have a low value to weight ratio so lugging the stuff around is a pain. The Swiss and Japanese stand out here for issuing the highest value coins, the five franc coin and 500 yen coin respectively, each worth around US$5.

As for cryptocoin fans, tomorrow Zcash will be launching. Whereas the entire history of bitcoin transactions is public, Zcash succeeds in hiding everything about the transaction. That's true anonymity.

Monday, October 17, 2016

The strange mania for Swiss National Bank shares

The shares of the Swiss National Bank (SNB), Switzerland's central bank, have almost doubled since July, despite there being no real news. Yep, you read that right, the SNB is listed on the stock market. There are four other central banks with listed shares: Belgium, Japan, Greece, and South Africa. I discussed this odd group back in 2013.

Why are SNB shares catapulting higher? This is a staid central bank, after all, not a penny stock.

Let's look at the fine print. Swiss National Bank shares aren't regular shares. To begin with, the dividend is capped at 6% of the company's share capital. The SNB was originally capitalized back in 1907 with 25 million Swiss francs, an amount that hasn't changed in 109 years. Which means that the dividend is, and always has been, limited in aggregate to a minuscule 1.5 million francs per year (about US$1.5 million). Because this amount must be divvied among the 100,000 shares, each share gets just 15 francs per year.

The SNB has faithfully paid this 15 franc dividend since its founding (apart from 2013, when it was omitted due to massive capital loss on its gold holdings). For instance, here it is paying the dividend throughout the 1980s:

Once the 1.5 million franc dividend is paid, Swiss law dictates that all remaining profits get sent to the Swiss central government and the cantonal governments. It further stipulates that if the SNB is to be liquidated, shareholders will only receive a cash payment equal to the nominal value of their shares. This means that if shares are trading for 1025 Fr, but there is residual firm value (after paying debtors) of 10,000 francs per share, well too bad—shareholders only get 1025 Fr.

Given these peculiar details, an SNB share isn't really equity; it's best thought of as a perpetual government bond with a 6% coupon, sometimes known as a consol. It throws off 15 francs per year for the rest of time. With the SNB as issuer, these securities are pretty much risk-free.

The math behind SNB shares is straightforward. Just use the Finance 101 formula for a perpetual bond, PV=c/r, or present value (PV) equals yearly cash flow (c) divided by yield (r). With annual interest payments of 15 francs and SNB shares trading at 2000 Fr, the yield comes out to 0.75%.

Here's what I think explains the rise in SNB shares. In July, an odd thing happened. The yield on the Swiss 50-year bond, the closest instrument in Switzerland to a perpetual government bond, fell below 0%. SNB shares (i.e. perpetual bonds) were themselves trading at around 1000 francs at that point for a yield of 1.5%. A few large investors probably began to ask themselves: Why the devil are we accepting a negative nominal return on our ultra long-term government bond portfolio if these other government bonds, which happen to be issued by the central bank and masquerade as shares, still have a positive yield? And so they started to buy SNB shares in quantity, driving the price up and the yield down.

This sort of thinking explains why SNB shares have risen, but not necessarily the explosive nature of the move. The shares haven't just risen by a few bucks, after all. They've almost doubled!

Let's take a closer look at the perpetual bond formula. If nominal interest rates fall from 1.5% to 0.5%, a fifty-year bond with a par value of 1000 francs and a 15 franc yearly coupon will rise from 1000 to 1,441 francs. Not bad, but compare that to a perpetual. For the same decline in rates, SNB shares will double in value from 1000 to 2000 francs. As rates continue to fall towards zero percent, the price of a perpetual goes parabolic. At 0.25%, the perpetual will be worth 6000 Fr, at 0.1% it trades at 15,000 Fr, and at 0.01% it sells for 150,000 francs. At 0.001%, they'd be valued at 1,500,000 francs each!

In the table below, I've illustrated the relative dynamic of a 50-year bond and a perpetual as rates fall to zero.
Price of a 50-year bond and perpetual at various interest rates

At the limit of 0%, SNB shares will have an infinite price. If all existing and yet-to-be-issued government fixed term bonds are guaranteed to lose money over the course of their existence, but there exists a government security that is guaranteed to perpetually offer a positive cash flow, then an investor will pay *any* amount of money to own those cash flows. There is no price to which SNB perpetual bond can rise that chokes off their demand.

The behavior of perpetual bonds at low interest rates *might* explain why SNB shares have gone hyperbolic. It also means that if Swiss rates continue to fall, the shares could have another double or two in the tank. So much for staid Swiss central banking.

Hat tip to Leon Oudejans for alerting me to the recent rise in SNB shares.

P.S. Does anyone know how rare perpetuals are, specifically perpetuals like SNB shares that don't provide the issuer with the option to call it? My understanding is that this sort of security just isn't issued anymore, at least not since British consols.

What the difference between an SNB share and a 1000 Swiss banknote? Not much, right? They're both perpetual bonds, one paying 15 francs a year, the other paying zero francs per year. If you were to write the number 1000 on an SNB bearer share, and 500 on a half share, and 100 on a fifth share, etc you'd have the entire series of Swiss banknotes. Cut the dividend on shares to zero, and won't banknotes and shares be exactly the same?

Here's a chart of the Bank of Japan, which is also a perpetual bond underneath the hood.

Wednesday, October 12, 2016

Bitcoin, drowning in a sea of credit card rewards

Satoshi Nakamoto kicked off his famous 2008 white paper with the line: "Commerce on the Internet has come to rely almost exclusively on financial institutions serving as trusted third parties to process electronic payments." He created Bitcoin, a form of decentralized cash, to deal with this problem. But as Meltem Demirors points out, Bitcoin adoption seems to have peaked. Eight years after Nakamoto published his paper, not many people are using the stuff as money.

Here's a way to get more people using bitcoins as money on the internet:

Commerce on the Internet has come to rely on a Visa/MasterCard pricing standard. Although a few online stores like Dell, Expedia, and Microsoft accept bitcoin payments, they still set their prices in terms of Visa/MasterCard dollars. Because the dominance of this pricing standard is preventing innovative money like bitcoin from emerging, it needs to be hacked.

The Visa/MasterCard standard

Credit card issuers aren't mere intermediaries. Along with their core payments offering, they sell a broad range of goods and services. This includes but is not limited to: 1) rewards in the form of points, air miles or cash back; 2) car rental and travel insurance; 3) warranty extensions; 4) coverage of goods purchased against loss/damage; and 5) price protection i.e. should the price of a good fall after you buy it, the card issuer refunds the difference.

Credit card issuers don't give all this stuff for free. Some of the costs are recouped by annual fees and interest on unpaid balances. But by far the biggest line item is something called 'interchange'. An interchange fee is a levy that merchants must pay to the credit card issuer each time a card is used. The better the card reward the larger the interchange fee. In Canada, for instance, the MasterCard World Elite interchange rate for internet transactions is 2.49%. Regular cards are docked interchange of just 1.61%. [source]

Retailers recoup interchange fees by passing them off to customers. The way they do this is to build interchange into sticker prices. In a world without credit cards, Dell accepts nothing less than $1000 for a laptop. But in an economy with credit cards Dell faces an average interchange rate of 2%, or $20 per laptop, reducing revenue-per-laptop to $980.

To recoup its costs, Dell marks laptop prices up to $1020.40. Of this amount, 2%, or $20.40, goes to the credit card issuer to cover the cost of the customer's rewards, insurance, warranty extensions, etc, leaving Dell once again with revenues of $1000/laptop. Dell doesn't actually tell us they are on-charging us for these things. They surreptitiously build this premium into the sticker price.

On the internet, every retail price has been marked up by around 2%. That's what it means to be on a Visa/MasterCard standard.

Bitcoin is being undervalued

The Visa/MasterCard standard has the effect of repelling bitcoin use.

Imagine Alice, a bitcoin user who wants to buy a laptop. In paying $1020.40 worth of bitcoin for a Dell, Alice effectively overpays. She gets the $1000 laptop but does not get the $20.40 in associated rewards, points, insurance, or price protection that are built into the laptop's sticker price. Rather, Dell gets to keep the $20.40 premium for itself, since it doesn't need to pay interchange on Alice's bitcoin payment.

Because retailers like Dell are undervaluing bitcoin, consumers like Alice are always better off using their credit card. The more rewards that credit card issuers add to their cards, the better the get at locking out bitcoin. This isn't just a problem with bitcoin and cryptocoins. The Visa/Mastercard standard has the potential to inhibit the adoption of other new media of exchange like mobile money.

No amount of code can hack the standard

If bitcoin could somehow be altered to pay rewards, cash back, and car rental insurance then it would be competitive with credit cards. This isn't a realistic option. Instead, the best fix is for retailers to offer bitcoin price discounts. If a Dell laptop retails online for $1020.40, bitcoin users should be charged just $1000 so that they aren't paying for points, rewards, car rental insurance and other benefits that they never get to enjoy. This would put bitcoin on an even playing field with credit cards.

How to convince retailers to offer bitcoin discounts? This isn't a problem that can be fixed by the bitcoin brain-trust making alterations to bitcoin source code. It's an interface problem: bitcoin hasn't been properly integrated into the real world, specifically into online shopping carts.

Cash has the same problem. The growing popularity of credit cards has led to the emergence of a Visa/MasterCard standard in the bricks & mortar economy. In a fully competitive economy retailers would compete to reduce their cash prices. However, retailers do not generally offer cash discounts, perhaps because they are worried about causing confusion, distrust, and delays at checkout counters (see discussion here). Alternatively, many retailers seem to believe that offering discounts is in contravention of Visa/MasterCard rules (it isn't.)

There is one big difference between cash and bitcoin. Cash lacks a community of users that can agitate for changes to the Visa/MasterCard standard. Central banks, which issue banknotes, don't really care that cash discounts aren't being offered because they lack a profit motive, and cash-using consumers, which tend to be poor or criminals, lack the means to organize. Lined up against cash is an incredibly powerful credit card lobby that has implemented all sorts of tricks (like prohibiting card surcharging) to prevent cash usage.

Unlike cash, Bitcoin boasts an active community of individuals and businesses with a strong interest in the success of the Bitcoin network. To hack the standard, this community needs to start agitating for discounts. Bitcoin payments providers Coinbase and Bitpay currently offer retailers the technology for setting bitcoin discounts. But Microsoft, Expedia, and Dell haven't taken them up on their offer, opting to keep the Visa/MasterCard standard in place. That's one place for activism to start. If the big three can be convinced to implement changes, this might be enough to kickstart an industry-wide practice of offering a 0.5 to 2% bitcoin discount. If the Bitcoin community succeeds in unbundling the Visa/MasterCard standard, it'll have leveled the playing field not only for itself but all subsequent forms of digital cash, whatever form they take.

P.S. I wrote a similar piece in 2015 on Bitcoin and the VISA/MasterCard standard. That piece focused on the superior stability of credit cards and the fact that credit card users, unlike bitcoin users, needn't incur foreign exchange costs since they spend most of their lives in the dollar universe. This version is more explicit about the reward side of the equation. All facets need to be integrated to determine how large of a bitcoin discount to be applied by retailers.

Thursday, October 6, 2016

Does money enjoy a home advantage?

Do citizens benefit by owning money that is pegged to the nation's own unit of account rather than a foreign unit of account?

Let's start by imagining a money that is not pegged to the national unit of account. Say that U.S. banking giant Wells Fargo establishes branches all over Canada. Instead of issuing chequing accounts that are pegged to the Canadian dollar, it decides that it will steal business from Canadian banks by issuing U.S. dollar-pegged chequing accounts to Canadians. Wells Fargo execs reason that the U.S. dollar is at least as stable as Canadian dollars, if not more so, and this could give their product an edge.

Further imagine that Wells Fargo is able to ensure that its U.S.-denominated money is just as liquid as that of competing Canadian money. Say that Canada's national ATM network is modified to dispense U.S. dollars to Wells Fargo cardholders. Retailers are convinced to accept U.S. dollar electronic payments, and so is Revenue Canada, the national tax authority. The upshot is that within Canadian borders, a U.S. dollar can do anything that a Canadian dollars can. Any Canadian in need of liquidity will be entirely indifferent between regular C$ deposits and Wells Fargo US$ deposits.

Let's also assume that shifting funds between U.S. and Canadian dollars is free, so Canadians who are paid in one unit have no compunctions about exchanging into the other. Apps and other technologies remove all inconveniences of calculating exchange rates. Finally, Wells Fargo's U.S. dollars are FDIC-insured and its Canadian branches have access to the Fed's discount window, making them just as safe as Canadian dollars.

This leaves just one difference between the two types of money; Canadian dollars are the unit of account. This means that retailers set prices in Canadian dollars, not U.S. dollars. Will this feature have any influence on whether Canadian consumers choose to open an account with Wells Fargo or stay with their existing bank?

Consumer prices are peculiar because, unlike asset prices, they stay constant for long spells of time. When examining the frequency of price changes for 350 categories of goods and services, Bils and Klenow found that prices tend to stay fixed for 4.3 months before they are updated.  Coin-operated laundry prices exhibited price spells of 79.9 months, driver's license fees 56.3 months, and newspapers 29.9 months (see below). On the short end of the spectrum, the price spell for gas is just 0.6 months, eggs 1 month, and women's suits 1.6 months.

Source: Bils & Klenow

One reason for this stickiness is that retailers are said to have made an "invisible handshake" with consumers that requires them to avoid raising prices when demand suddenly increases. Retailers bind themselves to this implicit contract because they do not want their loyal customers to view them as being unfair, spitefully bolting to the competition when prices are adjusted. Customers may prefer stable nominal prices because these allow them to carefully match their daily and weekly spending plans with the money currently available in their accounts. Put differently, if you've got $100 in your account, and the set of retailers at which you shop have promised not to budge on pricing for a few weeks, you can determine ahead of time what bundles of goods you can afford. Without the sticky price "handshake," you're in the dark.

So money that is denominated in the unit of account comes with a major ancillary benefit. In Canada's case, the nation's retailers have agreed to offer Canadian dollar owners a convenient planning mechanism, much like a day planner or personal organizer. This mechanism provides anyone who owns Canadian dollars around 4.3 months of  uncertainty alleviation within national borders.

Whereas Canadian sticker prices are characterized by long price spells, the U.S. dollar prices faced by Wells Fargo's Canadian customers will fluctuate by the second. To see why, consider that when a customer wants to pay in U.S. dollars, they will have to indicate their preference at the checkout counter. The retailer inputs the Canadian dollar sticker price of the good, multiplies it by the USD/CAD exchange rate, and arrives at the U.S. dollar price. The foreign exchange market is a 24 hour "flex-priced" market, so the exchange rate--and thus the U.S. dollar sticker price of goods--will always be fluctuating.

Wells Fargo money will therefore not be able to provide the same set of uncertainty-minimizing features that Canadian money provides. Anyone who holds, say, $100 U.S. dollars in their account cannot know ahead of time how much stuff they'll be able to buy over the ensuing three or four days. An equivalent C$100 provides near certainty. Lacking this nice property, Wells Fargo money faces significant headwinds at the outset.

This explains a point I was trying to make in my previous post on bitcoin. Bitcoin's battle to gain general acceptance will be a harder one to win than Wells Fargo's in Canada because the price of bitcoin is so much more volatile than U.S. dollars. Even if bitcoin's price eventually stabilizes so that it is just about as volatile as the U.S. dollar, it must still overcome the same unit-of-account hurdle as Wells Fargo i.e. neither can mimic the 'day planner' properties of the Canadian dollar. I don't think is an easy problem to overcome.

PS: Another complicating factor is product returns. If someone doesn't like the t-shirt they bought, and they paid in bitcoin, refunds are usually issued in U.S. dollars at the bitcoin-to-dollar conversion rates in play at the time of the transaction. So it's possible to make or lose bitcoin on refunds. For bitcoin owners, this adds an extra degree of uncertainty. Pay in dollars and you can be certain you'll get the exact same nominal amount back if the product is unsatisfactory. Pay in bitcoin and who knows what the amount of bitcoin you'll be refunded.