Tuesday, November 13, 2018

The demonetization gap

Two years ago, Indian PM Narendra Modi suddenly demonetized all of the nation's 1000 and 500 banknotes. His stated goal was to exact justice on all those holding large amount of dubiously-earned cash. But since these two denominations constituted around 85% of India's currency supply, the demonetization immediately threw the entire nation into chaos.

After suffering from a nine-month note shortage, enough new notes were printed to meet India's demand for cash. But in the interim, what had happened to Indians' demand for cash? Did their experience with demonetization lead them to hold less cash than before, or did they simply revert to their pre-demonetization habits and patterns? I wrote an article in September 2017 dealing with these questions. With another year having passed we now have more data—so I'm going to provide quick update.

My claim is that a demonetization gap continues to exists. This gap is evidence that the cancellation of 1000 and 500s has had an enduring effect on Indians' behaviour surrounding cash.

To measure the gap, we need to compare the evolution of India's cash supply to the path it would have taken in a world in which Modi's demonetization never occurred. We know that the supply of rupee banknotes grew at an average rate of around 12.3% from 2011 to 2016, and on the eve of demonetization there were around 17.5 trillion in banknotes and coins in circulation. Taking these numbers and extrapolating forward, November 2018's cash-in-circulation count would have clocked in at around 21.5 trillion rupees had the demonetization not occurred. But in our actual world, the one where demonetization occurred, India's cash count registers at just 19.5 trillion rupees.

So thanks to demonetization, Indians are holding around 2 trillion fewer paper rupees than they otherwise would have. I've illustrated the demonetization gap below:

What sort of assumptions do we need to make in order for the demonetization gap to be zero? Let's say that demonetization had never occurred. To get from a currency stock of 17.5 trillion in November 2016 to 19.5 trillion by November 2018, we need to assume an absurdly low 6% growth rate for the stock of currency stock. I doubt India has ever experienced such a slow growth rate in cash-in-circulation. So while you can quibble with the size of the gap, I think it's undeniable that some sort of gap exists i.e. demonetization has had long lasting effects on cash holdings.

There are two potential explanations for why a demonetization gap exists. My guess is that both are to some extent true, although as time passes the influence of the second will tend to disappear:

1. Demonetization might have had a permanent effect on Indians' taste for cash. Prior to demonetization, a money launderer might typically have built up an inventory of 100 lakh (US$135,0000) worth of banknotes before laundering them. But the demonetization frightened him, so now he launders his ill-gotten gains whenever he holds just 50 lakh worth of notes.

Or maybe a family that typically conducted most of their day-to-day transactions in cash opened a bank account during demonetization so that they could deposit their notes. And now they've fallen into the habit of paying for half their family expenses with cash, and the other half with debit cards.

These sorts of changes to transactions preferences would get expressed in the nation's overall stock of notes by a reduction relative to trend.

2. Demonetization might have caused large and lasting damage to the informal sector of the Indian economy. With 85% of the nation's money stuck suddenly came unusable, and so many businesses dependent on cash, these businesses may have been forced to go under. With informal production cratering, fewer banknotes would be required for transactions purpose than would otherwise be the case. Over the long term, however, one would expect these sectors to rebuild, any drag they had once placed on cash demand being removed.

If Indian's taste for cash changed, how might these new tastes have manifested themselves? Did Indians swap cash for deposits in the regulated banking system or did they simply swap one form of anonymous "black money" for another (i.e. cash for gold, diamonds, real estate)? In my post from last September I found some interesting charts. I'm going to update them below.

As the chart above illustrates, the number of point-of-sale (POS) terminals installed has experienced a one-time shot to the arm after the November 2016 demonetization announcement.

The second chart shows that the value of POS transactions using debit and credit cards moved to a new and higher plateau.

The last chart shows the growth in value transacted using mobile wallets. Demonetization gave mobile wallet usage a quick shot to the arm, but it is difficult to determine if growth stayed above trend in 2017-18 or below. Given that mobile wallets are recent and started from a very low base, usage had already been growing very fast before demonetization.

In any case, I think the charts provide at least some evidence that the shift out of cash has been captured by digital finance.


The existence of a demonetization gap does not mean that Modi's effort was a success. If someone comes up with a great monetary innovation, expect it to be quickly copied. Polymer banknotes, real-time gross settlement systems, and inflation targeting are all examples of monetary technologies that have spread from their original adopter to the rest of the world. But not a single nation has copied India's demonetization. I would be surprised if any ever do. The costs imposed by the event—lost labour time, foregone transactions, chaos—were all too apparent, and its benefits dubious.

That being said, I sympathize with the intuition that was at the core of Modi's demonetization. Let's face it, plenty of people use cash as a way to avoid taxes. These cheats force everyone else to pay more than their fair share for shared infrastructure. It would be nice if there was a way to rebalance the load by taking back resources lost to the cheats. A classic way to fix this problem is to have the tax department hire more inspectors. An alternative is to design the nation's cash system in a way that corrects for any unfairness. Modi's demonetization is an example of the second approach, a massive cash dragnet designed to flush out cheats.

But it was an incredibly clumsy effort. The tricky thing with cash is that it is simultaneously a vessel for tax evasion for a few cheats and also a vital monetary fluid for the many. Punishing the first group means that the second will also be hurt. If the costs incurred by the second group are too high, then the calculus behind the whole effort fails.

Modi's demonetization tried to lessen the blast radius by targeting users of large denomination 500 and 1000 notes while leaving users of the 10, 20, 50, 100, and 200 untouched. The motivation behind this was that presumably large tax cheats will concentrate their cash holdings in higher denomination notes, while regular note users will tend to concentrate in the lower denominations. But even with this filter in place, the 500 and 1000 were so widely held by Indians that the demonetization of these denominations forced a large proportion of the population to endure massive lineups for weeks. Retail trade ground to a halt.

Over the last two years I suggested a few ideas that would have improved the effectiveness of Modi's demonetization. Instead of targeting all 500s and 1000 notes, demonetizing notes by serial number would have resulted in a much smaller blast radius. Rather than withdrawing 500s and 1000s, it would have been easier to overstamp them with stickers, and then slowly withdraw these overstamped notes over time. This would have helped prevent massive cash shortages.

But even after these modifications, demonetization is too blunt of a tool to effectively solve the complicated inequities arising from cash-based tax evasion. So should Indians (and the rest of us too) simply live with these imperfections, accepting unfairness as an acceptable cost of maintaining cash systems?

Maybe not. In his paper Taxing Cash, Ilan Benshalom has suggested a less invasive technique for reclaiming resources from cash-using tax evaders. Rather than demonetizing all 500 and 1000s, Modi could have introduced a recurring withdrawal fee of 10 for each 500 note and a 20 rupee fee on each 1000 note.

Imagine that Rohit needs to pay Indira 10 lakh. The two can avoid taxes if they agree to transact using cash rather than making a digital payment. So Rohit withdraws one thousand 1000 bills from his bank. He incurs a fee of 20,000 (20 for each note), the bank forwarding the full amount to the government. Rohit might consider avoiding the withdrawal tax by withdrawing 50s and 100s (which don't incur fees), but he is unlikely to go this route since 10 lakh in small denomination notes would be terribly awkward.

The upshot is that Indian tax payers would get 20,000 in compensation for Rohit and Indira's evasion of taxes. Compared to Modi's demonetization, the blast radius of a large denomination withdrawal tax is much smaller. Regular Indians could have easily continued about their business using untaxed 100s or migrating to digital payments. It sounds promising, but after Modi's explosive demonetization, Indians are probably much less open to new monetary experiments, and who can blame them.

There are several other ways to visualize Indian demand for cash over time. For instance, here is the cash-to-GDP ratio:

The idea here is that for each unit of GDP, some amount of cash is used as means of circulating production. (Note: While the cash-in-circulation numbers are current, 2018 Q2 nominal GDP has still not been published. I estimated it by assuming 11.5% nominal growth, the average growth rate over the last few years.)

Vivek Dehejia and Rupa Subramanya measure the ratio of cash-in-circulation to the total money stock, or M3. They find little evidence of a reduced reliance on cash. For the chart below, I've used M4 which includes not only chequing and savings deposits held at banks but also total postal office deposits:

The cash-to-M4 ratio is a bit lower than it was on the eve of demonetization. For each ₹100 held in accounts, Indians now only hold ₹14.9 in banknotes and coins, down from ₹16 in October 2016. Be careful with using the cash-to-M4 ratio as a measure of preferences for banknotes. If Indians want to hold fewer deposits and more government bonds, M4 declines, and so the cash-to-M4 ratio will rise. But preferences for cash haven't changed at all.

Sunday, November 4, 2018

The credit theory of money

Over on the discussion board, Oliver and Antti suggest that I read two essays from Alfred Mitchell-Innes. Here are a few thoughts. 

A British diplomat, Mitchell-Innes was appointed financial advisor to King Chulalongkorn of Siam in the 1890s as well as serving in Cairo. He eventually ended up in the British Embassy in Washington where he penned his two essays on money. The first, What is Money, attracted the attention of John Maynard Keynes, while the second essay, The Credit Theory of Money—which was written in 1914—expounded on his views.

Both are interesting essays and worth your time. One of Mitchell-Innes's main points is that all money is credit. This may have been a controversial stance back in 1914, when people were still very much focused on metallic money, but I don't think anyone would find it terribly controversial today. If we look at the instruments that currently function as money, all of them are forms of credit, that is, they are obligations or "credits on a banker" as Mitchell-Innes puts it.

Having established his credit theory at the outset of his 1914 essay, Mitchell-Innes devotes much ink to patching up its weakest point: coins. Any critic will be quick to point out that the historical circulation of coins contradicts his claim that all money is credit. Coins, especially gold ones, were valued as commodities, not credit.

To protect his credit theory from this criticism, Mitchell-Innes downplays the role played by coins. So in What is Money he claims that for large chunks of history, the "principal instrument of commerce" wasn't the coin, but the medieval tally stick. These ingenious objects look like this:
While I certainly like the idea of tally sticks, to claim that they were the main way of engaging in hand-to-hand trade during medieval times doesn't seem likely. Long and awkwardly shaped, tally stick are not nearly as convenient as coins. It's hard to see why anyone would prefer them. Just like the sleek US$1 bill has driven the bulky $1 coin out of circulation, one would expect coins to push bulky tally sticks out of general usage.
Mitchell-Innes's second, and more radical, line of defence is to claim that coins themselves are a form of credit. "A government coin is a "promise to pay," just like a private bill or note," he says. Elsewhere he writes: "A coin is an instrument of credit or token of indebtedness identical in its nature with a tally or with any other form of money, by whomsoever issued."

This is a strange idea. Why would anyone issue a financial promise encoded on gold? For instance, imagine that I owe you some money. To give physical form to my debt, you ask me to write out an IOU which you will keep in your pocket. But why would I inscribe my promise on something expensive like a gold disc, especially when I could simply record it on a cheap and lightweight piece of paper? Larry White puts it better here:
"This account fails to explain, however, why governments chose bits of gold or silver as the material for these tokens, rather than something cheaper, say bits of iron or copper or paper impressed with sovereign emblems. In the market-evolutionary account, preciousness is advantageous in a medium of exchange by lowering the costs of transporting any given value. In a Cartalist pay-token account, preciousness is disadvantageous — it raises the costs of the fiscal operation — and therefore baffling. Issuing tokens made of something cheaper would accomplish the same end at lower cost to the sovereign."
Mitchell-Innes doesn't make much of an effort to explain why gold might have been selected as a medium for inscribing IOUs. But on the discussion board, Antti has a provocative theory. Credit is often collateralized, an asset being pledged by a borrower to a lender in order to reduce the risk of the loan. If a gold coin is a form of credit, then maybe the gold embodied in the coin is serving as collateral.

Think about it this way. While it would certainly be cheaper for me to record my IOU on paper, if I welch on my promise then the person who holds my IOU is left with nothing but a worthless note. But if I welch on an IOU that is encoded on a gold disc, at least the person who has my IOU is left with some gold (albeit of lower worth than the face value of the original IOU).

According to Antti's theory, a gold coin is therefore a more solid form of credit than a note, since it provides recourse in the form of precious metals collateral. If my credit is bad, the only way I may be able to get a loan is to issue gold IOUs, my paper ones being too risky for people to accept.

It's an interesting theory, but the problem with inscribing my IOUs on gold is that it is a terribly insecure way for me to conduct my business. Gold is highly malleable. Bite a gold coin between your canines and you'll leave a mark on its surface (trust me, I've done it). So if I pay you with my coin IOU, you could clip a tiny bit off the edge and keep it for yourself, passing the rest of the coin off at a store. That store owner could in turn pay it away to a supplier. Unaware that it has been clipped, the supplier returns the coin to me for redemption. I am obligated to accept the clipped coin at full value since it is my IOU. However, I've had a chunk of my collateral stolen somewhere along the transactions chain—and there's nothing I can do about it.

So putting one's gold collateral into circulation is an open invitation for thieves, which is why Antti's collateral theory of coins doesn't seem very realistic to me.


The idea that coins circulated at more than their precious metals content, or intrinsic value, can be found throughout Mitchell-Innes's two essays. He uses the existence of this premium as proof that the metal content of a coin is not relevant to its value, its credit value being the sole remaining explanation.

To some extent, I agree with Mitchell-Innes. Over the course of history coins have often circulated above their intrinsic value, and from time-to-time this premium has been due to their value as credit. The merchants' counterstamps below are great examples. By adding a stamp to a government coin, these merchants have elevated the coin's value from one cent to five or ten cents.

These three coins are straight out of Mitchell-Innes two essays. As I say in the tweet, counterstamped coins effectively functioned as an IOU of the merchant. For instance, take the five cent Cameron House token, on the right. This token was issued by a Pennsylvania-based hotel—Cameron House.  Its intrinsic value was one cent, but Cameron House's owner promised to take the coin back at five cents, presumably in payment for a room. The sole driver of the coin's value was the reliability of Cameron House's promise, the amount of metal in the token having no bearing whatsoever on its purchasing power.

While Cameron House's stamp turned metal into a much more valuable form of credit, not all stamps do this. Last week I wrote about coin regulators who regulated gold coins and shroffs who chopped coins. Both functioned as assayers, weighing a coin and determining its fineness. If the coin was up to standard, the regulator or shroff stamped their brand onto its face and pushed it back into circulation. Below is a chopmarked U.S. trade dollar:

Chopped 1880 U.S. trade dollar (source)

But unlike the Cameron House stamp, the regulation or chopping of coins didn't turn them into a credit of the regulator or shroff. The marks were simply indicators that the coin had been audited and had passed the test, and nothing more.

Both the Cameron House coin and the chopped U.S. trade dollar would have traded at a premium to the intrinsic value of the metal that each contained. But for different reasons. As I wrote above, the Cameron House coin was a form of credit, like a paper IOU, and thus its value derived from Cameron House's credit quality, not the material in the token. But not so the chopped U.S. trade dollar. Precious metals are always more useful in assayed form than as raw bullion. While it is simple to test the weight of a quantity of precious metals, it is much harder to verify its fineness. This is why chopmarks would have been helpful. Anyone coming into possession of the chopmarked coin could be sure that its fineness had been validated by an expert shroff. And thus it was more trustworthy than silver that had no chopmark. People would have been willing to pay a bit extra, a premium, for this guarantee.

Remember that a decline in the amount of metal in a five-cent Cameron House token would not have changed its purchasing power. With a chopmarked trade dollar, however, any reduction in its metal content flowed through directly to its exchange value. This is because a chopmarked dollar was nothing more than verified raw silver. And just as the value of raw gold or silver is determined by how many grams are being exchanged, the same goes for a chopmarked trade dollar.

And so whereas Mitchell-Innes has a single theory of money, we've arrived at two reasons for why coins might trade at a premium to intrinsic value, and why their purchasing power might change over time. The Cameron House theory, which also happens to be Mitchell-Innes's theory, and the chopmarked trade dollar theory, which is completely contrary to Mitchell-Innes's essays.


I've used private coinage for my examples, but these principles apply just as well to government coinage. Our modern government-issued coins are very similar to the Cameron House tokens. They are a type of IOU (as I wrote here). In the same way that trimming away 10% from the edge of a $5 note won't reduce that note's purchasing power one bit, clipping some of the metal off of a toonie (a $2 Canadian coin) won't alter its market value. The metal content of a modern coin is (almost always) irrelevant.

But whereas modern government coins operate on Cameron House principals, medieval government coins operated on the same principals as chopmarked traded dollars. In England, a merchant who wanted coins would bring raw gold or silver to a mint to be converted into coin. But the merchant had to pay the mint master a fee. The amount by which a coin's market value exceeded its intrinsic value depended on the size of the mint's fee.

Say it was possible for a merchant to purchase a certain amount of raw gold with gold coins, pay the fee to have the raw gold minted into coins, and end up with more coins than he started with. This would be a risk-free way to make money. Everyone would replicate this transaction—buying raw gold with coins and converting it back into coins—until the gap between the market price of a coin and the market price of an equivalent amount of gold had narrowed to the size of the fee. 

Premia on coins weren't always directly related to mintage fees. English mints usually operated on the principle of free coinage—anyone could bring their gold or silver to the mint to be turned into coin. But sometimes the mint would close to new business. Due to their usefulness and growing scarcity, gold coins would circulate at an ever larger premium to an equivalent amount of raw gold. Since merchants could no longer bring raw gold to the mint and thereby increase the supply of coins, there was no mechanism for reducing this premium.

So as you can see, whether the mint was open and coinage free, or whether it was closed, the premium had nothing to do with the coin's status as a form of credit. It was due to a combination of the superiority of gold in validated form and the availability of validated supply.

In sum, Mitchell-Innes is certainly right that coins have often been a form of credit. A stamp on a piece of metal often elevates it from being a mere commodity to a token of indebtedness. In which case we get Cameron House money. But as often as not, that stamp is little more than an assay mark, a guarantee of fineness. In which case we have chopmarked trade dollars. Both sorts of stamps put a premium on the coin, but for different reasons.

Coming up with grand theories of money is tempting, as Mitchell-Innes has done, but unfortunately these theories sometimes obscure the finer features of monetary instruments. At times, having twenty or thirty bespoke theories may be a better way to understand monetary phenomena than one grand one.

Tuesday, October 23, 2018

Gold regulators

While our modern monetary system certainly has plenty of detractors, one of its successes is that we no longer need the services of the local gold regulator. In the late 1700s, the job of a gold regulator was to assay gold coins to determine if they were of the appropriate weight and fineness, modifying (ie 'regulating') the coin if necessary. When he was done, the gold regulator stamped the coin with his seal of approval and put it back into circulation.

The job of regulating coins may seem strange to us. But it was an ingenious way to cope with the lack of standardization that bedeviled monetary systems in the 1600 and 1700s, particularly in the colonies. There was no domestic supply of coins in North America back then, so settlers relied on a bewildering array of foreign coins as their media of exchange, each with its own weight and fineness, and most of poor quality. This included not only silver coins such as Spanish dollars, pistareens, and English crowns, but also a gamut of gold coins including Portuguese joes and moidores, English guineas, French Louis d'ors and pistoles, German carolines, and Spanish doubloons.

Each of these foreign gold coins was minted with a unique quantity of the yellow metal. For instance, the popular Portuguese half Johannes, or "half Joe", weighed 221 grains and was 91.7% pure when it left the mint, whereas a full-bodied Spanish doubloon weighed 416 grains (one gram equals ~15 grains). Thanks to constant wear and illegal clipping, these coins would inevitably lose some of their mass as they circulated. For merchants, the task of weighing each gold coin that was presented to them, checking if it was a counterfeit, and calculating its appropriate monetary value would have been fatiguing.

To help merchants determine the rate at which to accept a particular gold coin, the authorities published tables with coin weights and values. These coin standards were issued by various legislative bodies or by the merchants themselves. For instance, here is a table produced by the New York Chamber of Commerce in 1770:

Let's go through an example of how an American merchant might use this table. Say a customer offers a merchant a slightly worn half-Joe. The merchant measures the coin and discovers that it weighs exactly 9 dwt, or 216 grains. A dwt is a pennyweight, an archaic unit for measuring weights that derives from denarius weight. (1 dwt = 24 grains = 1.55 grams). Let's further assume say this exchange occurred in New York.

According to the above table, any half-Joe passed in New York that weighs at least 9 dwt 0 grains is legal tender for £3 and 4 shillings. Look in the columns titled "least weight" and "N. York". So if our merchant happens to be selling horses for £3 4s, then he'd be happy to accept the customer's slightly worn half-Joe as sufficient payment for a horse. But if the half-joe were to weigh less than 216 grains (or 9 dwt), then it would fail to meet the Chamber of Commerce's standards, and therefore the merchant wouldn't accept it—the coin is worth less than the horse's sticker price of £3 and 4 shillings.

Testing a coin's weight is easy, but it is unlikely that many merchants would have had the time or expertise to verify its purity. Whereas a Portuguese half-joe was minted with 91.7% fine, a good counterfeit half-joe might be just 88% pure. If a subsequent trading partner questioned a counterfeit's validity, a merchant who had accidentally accepted it could be out of pocket. To remove any doubt, a suspect coin could be brought to the local regulator to be assayed. After removing a small section of the coin, the regulator would then test the coin's gold content. If it was a good coin, he would plug up the test section and stamp his initials on it. Having been approved by a recognized member of the community, the coin could easily pass in trade.

This watchdog function reminds me of the part played by Chinese shroffs, or money changers, in the Chinese monetary system of the 17th-19th centuries. Like North America, China was inundated with a whole range of foreign coins. Local shroffs would assay a foreign coin to verify its silver content. If the coin passed their purity test, a shroff would stamp it with his own peculiar chopmark, usually a Chinese character or symbol. Over time, foreign coins circulating in China might collect multiple chops. The genius of this system is that a naive Chinese consumer could safely accept a strange coin knowing that it had successfully passed the smell test of professional appraisers—and the more chops the better.

Chopmarked 1807 Spanish silver dollar 

If you are interested in the Chinese practice of chopmarking, I wrote about it here and here.

Testing for fineness is just one theory for the role played by North American gold regulators. There is a second theory. Not only were they watchdogs, but regulators also acted as enhancers or correctors. To see why this might be true, let's delve a bit more into the dynamics in play in North America in the 1700s.   

If you look at the original table above, notice that the New York Chamber of Commerce listed the minimum accepted weight of the half-Joe as 216 grains. However, as I pointed out earlier, a freshly-minted half-Joe actually weighed 221 grains. So the Chamber of Commerce's standard tolerated the circulation of underweight half-Joes. (They did the same with other coins too, including the doubloon, which when freshly minted weighed 418 grains but was accepted by the Chamber of Commerce at 408 grains.) Providing some extra leeway was probably a wise move. The coins used by New Yorkers came from distant realms and inevitably suffered from wear & tear.

But if a half-joe had lost too much of it original heft it would fail to meet the Chamber's standard. This is where a gold regulator might come in handy. Say that a half-Joe was brought to a bank but found to only weigh 207 grains, well below both the Chamber's standard of 216 grains and its original mint weight of 221 grains. The bank would purchase it at discount, say by crediting the customer with just £3 1 shilling instead of the Chamber's standard £3 4s, then send the underweight coin to a gold regulator. The regulator would proceed to cut out a section of the coin and insert a purer (and heavier) gold plug into the hole, bringing its weight back up to the 216 grain standard.

The regulator would then stamp his initials on his modification, upon which the bank would pay the regulated coin out as change. The regulator's marks were proof that the coin lived up to the Chamber's weight standard, and presumably made it easier to pass from hand to hand. Here is what a coin that has been regulated with a plug looks like:
The coin in my tweet is a regulated 1747 Portuguese half-Joe. Notice that the plug is slightly raised on the face side of the coin, or the obverse side. On the reverse side of the coin, the plug is rounded and convex. So it apparent to the eye how the plug might add some heft to an underweight coin.

A coin could also be modified in a way that reduced it to the Chamber's standard. If a fresh half-Joe arrived in New York weighing 221 grains, it made no sense for its owner to spend it as-is. Given the Chamber's standards, a 216 grain half-Joe was sufficient to buy £3 and 4 shillings of goods and services (or settle £3 and 4 shillings debts). A 221 grain half-Joe was overkill. A customer could deposit their full-weighted half-joe at the bank for more then its value (say £3 and 5 shillings). The bank's gold regulator would then shave it down to size and stamp his initials on it. Thus the modified half-joe could circulate legally despite having a small amount clipped from it. It would have looked a bit like this:

Regulated 1774 Portuguese half-joe. Source

Note the flat part at the bottom where the half-Joe has been clipped by a regulator.

So a gold regulator's role, whether it be as a watchdog or an enhancer, or a bit of both, was to bring some much-needed order to the chaos of a multicoin monetary system. By bringing a gold coin up to standard, or reducing it to standard, they would have helped ensure the fungibility of North America's coinage. And by stamping their initials on it, regulators provided a guarantee of purity to the public—removing some of the uncertainty involved in accepting unfamiliar coins.

By the 1800s, there was no longer a need to have a gold regulators. Most of the deficiencies of the old non-standardized monetary system had been fixed. Paper money had largely displaced gold coins, so merchants had fewer occasions where they had to worry about accepting bad coins. As for smaller denomination silver coins, these were eventually replaced by token coins. The issuer's promise to buy them back at a fixed price, and not their metal content, dictates a token's value. While their time may be past, gold regulators remain a testament to monetary ingenuity.

Selected sources:
1. Sedwick, Daniel: The Regulated Gold Coinage of North America and the West Indies in the Late 1700s [link]
2. Michener, Ron: Money in the American Colonies [link]
3. Neufeld, EP: Money and Banking in Canada
4. The Yale University Brasher Doubloon [link]
5. Introductory Note: Report on the Establishment of a Mint, [28 January 1791] [link]
6. Mossman, Phillip: Money of the American Colonies and Confederation [link]

Saturday, October 13, 2018

Bitcoin and the bubble theory of money

A few months ago Vijay Boyapati asked me to "steel-man" the bubble theory of money. The bubble theory of money, which can originally be found in a few old Moldbug posts, has been used by Vijay and others to explain the emergence of bitcoin and make predictions about its future.

So here is my attempt. I am using not only an article by Vijay as my source text, but also one by Koen Swinkels, a regular commenter on this blog. Both are interesting and smart posts, it's worth checking them out if you have the time.

Steel-manning the bubble theory of money and bitcoin

1. Unlike a stock or a bond, which is backed by productive assets, bitcoin cannot be valued using standard discounted cash flow analysis. And since it has no intrinsic uses, it can't be valued for its contribution to various manufacturing processes, nor for its consumption value. Rather, bitcoin is a bubble. Its price is driven by a speculative process whereby people buy bitcoins because they think that other people can be found who will pay an even higher price.

2. There is no reason why bitcoin must pop. At first, bitcoin will be bought by those on the fringe. As more people get in, the price of bitcoin will rise further. It will continue to be incredibly volatile along the way. But once bitcoin is widely held (and very valuable), the flow rate of incoming buyers will fall, and so will its volatility. At this point it has become a stable low-risk store of value. The eventual stabilization of bitcoin's price is a commonly held view among the bitcoin cognoscenti. For instance, bitcoin encyclopedia Andreas Antonopoulos has often said the same thing (i.e. "volatility really is an expression of size").

3. Once its price has stabilized, bitcoin can transition into being a widely used money, since people prefer stable money, not volatile money.

So having steel-manned the bubble theory of money as applied to bitcoin, where do I stand?

I agree with points 1 and 3. My beef is with the middle point.

Will a Keynesian beauty contest ever stabilize?

First off, let me point out that there are elements of the second argument that I agree with. Yes, bitcoin needn't pop, although my reasons for believing so are probably different from Koen and Vijay.  In the past, I used to think that a popping of the bitcoin bubble was inevitable. After all, as a faithful Warren Buffett disciple, I believed that the price of any asset eventually returns to its fundamental value, and bitcoin's is 0.

But the eternal popularity of zero-sum financial games, or gambles, has disabused me of this view. People are lured by the promise of winning big and changing their lives without having to do any work. Heck, even though a Las Vegas slot machine will take on average 8 cents from every $1 wager, people still flock to insert $1 bills into slots. And so they will play bitcoin too, which like a slot machine is also a zero-sum game.

But I digress. The key point I want to push back on is Vijay and Koen's assumption that bitcoin volatility will inevitably decline as it gets more mature. I'm going to accuse them of making a logical leap here.

If bitcoin is fundamentally a bubble, or—as Vijay describes it—if bitcoin's price is determined game-theoretically, then why would its price dynamics change if more people are playing? Almost a century ago, John Maynard Keynes described this sort of game as a beauty contest. Presented with a row of faces, a competitor has to choose the prettiest face as estimated by all other participants in the contest:
"...each competitor has to pick, not those faces which he himself finds prettiest, but those which he thinks likeliest to catch the fancy of the other competitors, all of whom are looking at the problem from the same point of view. It is not a case of choosing those which, to the best of one’s judgment, are really the prettiest, nor even those which average opinion genuinely thinks the prettiest. We have reached the third degree where we devote our intelligences to anticipating what average opinion expects the average opinion to be. And there are some, I believe, who practise the fourth, fifth and higher degrees."
Whether 100 people are participating in Keynes's beauty contest, or 10,000, the nature of the game has not changed—it is still an nth degree mind-game with no single solution. Since the game's underlying nature remains constant as the number of participants grows, its pricing dynamics—in particular its volatility—should not be affected.

The stabilization of Amazon

We can think about this differently by using actual examples. I know of an asset that has become less volatile as it has gotten bigger: Amazon. See a chart below of its share price and volatility over time:

Why has Amazon stabilized, and will bitcoin do the same? When Amazon shares debuted back in 1997, earnings were non-existent. Jeff Bezos had little more than a hazy business plan. Since then the stock price has steadily moved higher while median volatility has declined. Amazon shareholders used to experience day-to-day price changes on the order of 2.5-4.5% in the early 2000s. By the early 2010s, this had fallen to 1-2% or so. Over the past several years, volatility has typically registered between 0.5-1%.

I'd argue that the stabilization of Amazon hasn't been driven by a larger market cap and/or growing trading volumes. Under the hood, something fundamental has changed. The company's business has matured and earnings have become much more stable and predictable. And so has its stock price, which is just a reflection of these fundamentals.

I've just told a reasonable story about why a particular asset has become less volatile over time. But it involves earnings and fundamentals, two things that bitcoin doesn't have. I'm not aware why a Keynesian beauty contest, which lacks these features, necessarily gets less volatile as more people join the guessing game.

Vijay and Koen draw an analogy between gold and bitcoin. Their claim is that if gold once transitioned from being a volatile collectible into a low-risk store of value, then so can bitcoin. But we really don't have a good dataset for the price of the yellow metal, so we really have no idea how its volatility changed over time. Going back to 1969—admittedly far too short a time-frame—gold has certainly increased in size (i.e. the total market value of above-ground gold has increased), but unlike Amazon there is no evidence of a general decline in price volatility:

I'd argue that in gold's case a lack of a correlation between size and volatility makes sense. A large portion of gold's daily price changes can be explained by speculators engaging in a Keynesian beauty contest, not changes in industrial demand or earnings (unlike Amazon shares, gold doesn't generate income). There's no good reason to expect that the volatility generated by gold speculators' beliefs should level off as participation in the "gold game" grows. Any game in which speculators base their bets on what they expect tomorrow's speculator to do, who in turn are guessing about potential bets made by next week's speculators, who in turn form expectations about the choices made by next month's players, is unlikely to converge to a stable answer for very long.

Will Proof of Weak Hands 3D tokens ever become money?

As a third example, let's take Proof of Weak Hands 3D (PoWH3D), an Ethereum dapp that I've blogged about a few times. PoWH3D is a self-proclaimed ponzi game. Basically, a player purchases game tokens, or P3D tokens, with ether. Each player's ether contribution goes into the pot, or the PoWH3D smart contract, less a 10% entrance fee which is distributed pro-rata to all existing P3D token holders. When a player wants to exit the game, their tokens are sold for an appropriate amount of ether held in the pot, less another 10% that is distributed to all remaining players.

So if a new player spends one ether (ETH) on some P3D tokens only to sell those tokens an instant later, they'll end up with just 0.81 ETH, the first 0.1 ETH having been paid to everyone else upon the new player's entrance, the other 0.09 being deducted upon their exit. Why would a new player take such a bad bet? Only if they believe that a sufficient number of latecomers will join the game such that they'll get enough entrance and exit income to compensate them for the 0.19 ETH they have already given up.

PoWH3D is a pure Keynesian beauty contest. A new entrant's expectations are a function of whether they believe latecomers will join, but latecomers' expectations are in turn a function of whether they believe yet another wave of even greater fools will pile in, etc, etc.

Applying Koen and Vijay's assumption that volatility decreases with adoption, then the return on P3D tokens should become less volatile as more people join. It might even transition into a stable investment, say like a blue chip utility stock. Who knows, it could even become a medium of exchange to rival the dollar. But surely Koen and Vijay don't want to walk out on a limb and argue that a pure ponzi game like PoWH3D will ever stabilize. Or that it might become a form of money. I think the most reasonable thing we can say about PoWH3D is that once a ponzi game, always a ponzi game. The volatility of its returns will not decline as the game grows, and that's because the game's fundamentals, its ponzi nature, doesn't vary with size. (If you are interested in PoWH3D, here are some great charts and stats).

At this point, it may be useful to map out a chart of bitcoin's 200-day median volatility. As in the case of Amazon and gold, I use the median rather than the average to screen for outliers:

I haven't updated the chart for two months, but volatility has declined since then. Vijay and Koen will probably say that as of October 2018 bitcoin is less volatile than it was in 2011. That's certainly true. But eyeballing the chart, we certainly don't get the same clean relationship between size and volatility as we do with the Amazon chart.

Here's the biggest oddity. By December 2017 bitcoin had reached a market cap of $300 billion, its highest value to date. If Vijay and Koen are right, peak size should have corresponded with trough volatility. But this wasn't the case. In late 2017, bitcoin volatility was actually quite high. In fact, it exceeded levels set in late 2013, back when bitcoin was still a tiny $3 billion pup! The lesson here is that with bitcoin, bigger is just as likely to correspond with more volatility as it is with less volatility. More broadly, when it comes to Keynesian beauty contests there seems to be no fixed relationship between volatility and size. It's chaos all the way down.

This leads into Koen and Vijay's final point, that once bitcoin's price has stabilized, it can transition into a widely used money. I agree with the underlying premise that only stable instruments will become accepted by the public as media of exchange. But since I don't see any reason for bitcoin to stabilize, I don't see how it will make the leap from a speculative instrument to a popular means of paying people.

Bitcoin isn't on the verge of going mainstream. It's already there.

Vijay's message (Koen's not so much) can be taken as investment advice. Because if he is right, and bitcoin has yet to progress to a popular store of value and finally a medium of exchange, then we are still in the first innings of bitcoin's development. Vijay points to what he thinks are the features that will make bitcoin win out against other popular stable assets, including portability, verifiability, and divisibility.  Given that only the “early majority” has adopted bitcoin (the late majority and laggards still being far behind), Vijay thinks it would be reasonable for the price of bitcoin to hit $20,000 to $50,000 on its next cycle, and hints at an eventual price of $380,000, the same market value of all gold ever mined. So buying bitcoin now at $6,000 could provide incredible returns.

I have different views. Whereas Vijay thinks bitcoin has yet to go mainstream, I think that bitcoin went mainstream a long time ago, probably by late 2013. Bitcoin is often portrayed (wrongly) as a payments system-in-the-waiting, and thus gets unfairly compared to Visa and other successful payments systems. Given this setup, cryptocurrencies seems to be perpetually on the cusp of breaking out as a mainstream payments option. But bitcoin's true role has already emerged. Bitcoin is a successful decentralized gambling machine, an incredibly fun censorship-resistant Keynesian beauty contest.

Viewed this way, bitcoin's main competitors were never the credit card networks, Citigroup, Western Union, or Federal Reserve banknotes, but online gambling sites like Poker Stars, sports betting venues like Betfair, bricks & mortar casinos in Vegas, and lotteries like Powerball. By late 2013, bitcoin was at least as popular as some of the most popular casino games, say baccarat or roulette. It had hit the big leagues.

Whereas Vijay hints at a much higher price, where do I see the price of bitcoin going? I haven't a clue. But if I had to give some advice to readers, I suppose it would be this. Like poker or slots, remember that bitcoin is a zero-sum financial game (For more, see my Breaker article here). You wouldn't bet a large part of your wealth in a slot machine, would you? You probably shouldn't bet too much with bitcoin either. Vijay could be right about bitcoin hitting $380,000. It could hit $3.80 too. But if it does go to the moon, it will do so for the same reason that a slot machine pays off big.

It's worth keeping in mind that when it comes to gambling, the house always wins. Searching around for the lowest gambling fees probably makes sense. As I said earlier, Las Vegas slots will extract as much as 8 cents per dollar. Lotteries are even worse.

In bitcoin's case, the "house" is made up of the collection of miners that maintain the bitcoin system. All bitcoin owners must collectively pay these miners 12.5 bitcoins every 10 minutes to keep things up and running. So if you hold one bitcoin and its market value is $6000, you will be paying around 62 cents per day in fees, or $230 per year. That works out to a yearly management expense ratio of 3.8%. Beware, this number doesn't include the commissions that the exchanges charge you for buying and selling.

So before you start gambling, consider first whether the benefits of decentralization are worth 3.8% per year. If not, find a centralized gambling alternative. If the costs of decentralization are worth it, then buy some bitcoin, and good luck! But play responsibly, please.

Wednesday, September 26, 2018

Are Argentinians paying for Uber rides with bitcoins?

Earlier this month the following tweet elbowed its way onto my Twitter timeline:

The tweet comes from Anthony Pompliano, aka Pomp, who works at Morgan Creek Digital Asset where he runs a cryptocurrency fund.

So, have I been wrong all along about bitcoin? As anyone who has been reading me for a while will know, I've been skeptical of the bitcoin-as-money story. Rather than fulfilling Satoshi Nakamoto's vision as being a next generation medium of exchange, bitcoin has gone mainstream as a new type of gambling technology—an exciting decentralized zero-sum financial game. This is a somewhat useful role, but let's face it, it's not quite as revolutionary as digital cash.

But if Argentinians are indeed hailing rides and paying drivers directly with bitcoins, as Pompliano seems to be saying, then maybe I've been too quick to dismiss the bitcoin-as-money scenario. Paying for stuff is exactly what Satoshi Nakamoto intended bitcoin to do, right? So I dug further into the tweet.

Twitter: Couldn't find anything in the news about this. Anyone got a solid source?
Pomp: https://cointelegraph.com/news/uber-switches-to-bitcoin-in-argentina-after-govt-blocks-uber-credit-cards
Twitter: Pomp that was 2 years ago
Pomp: Does that make it less important?

And that's how Pomp left things. So it looks like I've got some work to do.

Here's the fine print. In 2016, the City of Buenos Aires ordered the major credit card companies to block Uber's App. Stanford Law School's WILmap project has a detailed post on this. So Argentinians suddenly found that while their MasterCard and Visa cards worked for everything else, they could no longer be used to get an Uber ride.

Contra Pompliano, Uber did not respond by allowing users to purchase rides with bitcoin. Rather, the company pointed out that anyone who had a certain type of pre-paid debit card could sneak by the Uber embargo.

To carry out the hack, the first thing that an Argentinian had to do was to apply for a prepaid debit card from any of Entropay, EcoPayz, Payoneer, or ZapZap. These are non-Argentinean payments companies. Entropay, for instance, is based in Europe and issues Visa-branded debit cards in partnership with a Malta-based bank, Bank of Valletta. Once Entropay had approved an Argentinean for an account, either a physical debit card would be sent by mail to the applicant's address in Argentina or a virtual card would be instantly created. An Argentinean could then log on to Entropay's website and use their local credit card, the same one that had had been neutered by the Uber embargo, to top up the Entropay prepaid debit card. With the debit card now funded, it could be used locally to pay for Uber rides.

Under the hood, prepaid cards issued by Entropay are really just regular Visa cards. So when an Uber ride was requested in Buenos Aires, an Entropay card would have used the same Visa rails that a regular Argentinean credit card used. Why would an Entropay Visa card be accepted but a regular Visa card denied?

The nub seems to be this: the ban seems to only have applied to payments instigated by domestically-issued cards. When payments to Uber originated from Entropay or any of the cards listed above, they were earmarked as originating internationally, in Entropay's case probably from Malta-based Bank of Valetta, and so Entropay payments were able to squeak by. Voila, by swapping domestic cards with international ones, Argentinians could avoid the blockade.

A number of bitcoin debit cards also enabled the hack, including Xapo and Satoshi Tango. Maybe this is what Pomp is referring to in his tweet. But it would be wrong to say that these cards allowed Argentinians to "purchase rides with Bitcoin," as he claims.

Prior to paying for an Uber ride, an Argentinian had to load U.S. dollars onto the bitcoin debit card by selling bitcoins for dollars on a bitcoin exchange. Either the card owner did this manually, or the card provider rapidly sold bitcoin in the very same instant that the payment was requested. In either case, bitcoins weren't flowing from the card holder to Uber. A fiat currency had been pre-loaded onto the card, and everything after that was just a  regular transfer over the Visa or MasterCard network.

These bitcoin debit cards are really no different from gold-based debit cards. Nor are they any different from the cheque-writing and debit card privileges provided by some U.S. money market mutual funds. Neither bitcoins nor gold not mutual fund units are being transferred from the card holder to the seller. Rather, each item is being quickly sold and turned into fiat, then processed along the same rails as any other payment.

In theory, all sorts of assets might be debit card-ized in this way. Buy a coffee with your Facebook debit card, for instance, and underneath the transaction's hood your Facebook share(s) are being quickly sold on the stock market for dollars, those dollars being the medium that ultimately settles the payment between you and the cafe. Complicating matters is that Facebook shares, which trade at $165, can't be cut into fractions, unlike bitcoins or fractions of a gold bar, so paying for a $3 coffee might get a bit awkward.

So returning to the tweet, recall that Pomp proclaimed that "more companies will begin using Bitcoin to fight back against corruption." But this wasn't the case with Uber. As you can see from the above, the company fought back by pointing to a neat hack of the existing credit card networks. The reason that Xapo and Satoshi Tango bitcoin cards were able to enable Uber purchases in Argentina wasn't because of their unique bitcoin nature. In Xapo's case, the card was issued by Wave Crest Holdings, a Gibraltar-based company. Like a regular fiat-based Entropay card, the incoming Xapo card payment was classified as an international one, and thus it escaped the domestic blockade.

Most bitcoin debit cards are no longer functional. Wave Crest, the card provider through which most bitcoin firms partnered, was suspended by Visa for non-compliance with Visa's rules in early 2018. If you go to VoyEnUber.com, an independent Argentinean website that reports on Uber, you'll see that it has delisted Xapo and Satoshi Tango from its list of ways to pay for Uber. The non-bitcoin prepaid debit cards are still there. So Pomp's tweet is twice wrong: 1) not only were Argentinians not using bitcoin to pay for Uber rides in 2016, but; 2) by the time of his tweet, they are not even making Xapo card payments, since Visa has cut that option off.

In addition to using foreign cards to pay for Uber rides, it seems that people in Buenos Aires are also paying with cash. According to the article, when riders pay with banknotes, there is no way for Uber to collect its 25% commission, so driver's are increasingly indebting themselves to the company. Or put differently, drivers are accepting cash, then paying Uber with a personal IOU. The irony here is that a combination of old fashioned fiat banknotes and trust-based IOUs—not bitcoin—are being used to "fight back against corruption."

So be careful what you read, folks. This sort of reminds me of the Zimbabwe bitcoin story from last year, which was seen as a crystallization of the long-held dream that bitcoin would help unbanked Africans. I rebutted that particular myth here.

Pompliano seems like a nice guy, so I'll just assume that excitement got the best of him. I normally try to avoid someone is wrong on the internet posts, but since he has over a 100,000 followers on twitter, and this particular meme has been retweeted over 2,000 times, I feel like it's my duty to try undo some of his error. The good thing with twitter is you can untweet retweets, feel free to go ahead and do that right now. :)

Sunday, September 23, 2018

Did Brexit break the banknote?

Nations never experience year-over-year declines in cash in circulation. Sweden (which I wrote about here, here, and here) is one of the rare exceptions. India is another, but this was due to its notorious botched demonetization attempt (which I wrote about here, here, here, and here). But now the UK seems to be joining this small group of outliers.

Why does a nation's cash in circulation generally grow consistently from one year to the next? While economies do experience the odd recession, in general they are always improving. Improving economies coincide with more demand to make transactions, and for this the public needs to have greater amounts of cash on hand. There is a counter-cyclical element to cash holdings. When recessions occur, people often turn to unofficial sectors of the economy to make a living, and this often requires cash. The last explanation for the steady growth in cash outstanding is inflation. Let's assume an inflation rate of 10%. Someone who generally hold $10 worth of purchasing power in their wallet in 2018 will have to hold $11 in 2019 if they want their situation to stay the same. To meet that demand, the central bank has to print more banknotes.

All of this is why the UK's recent flirtation with decashification is so strange. Below is a chart showing the year-to-year change in British paper currency in circulation:

For eight months now, since February 2018, the stock of Bank of England banknotes has been registering below the previous year's count, a phenomenon that Britain has never seen (at least not since the start of the data series I found).

One potential explanation for the recent bout of decashification is increased debit and credit card usage. I am not entirely convinced by this argument. People's transactional habits are notoriously slow to change. When the inevitable card-induced decline in cash does occur, it won't suddenly occur in the space of eight or nine months, but will take place over an extended multi-year period. As in the UK, card usage in Canada is ubiquitous, yet we haven't seen the same sort of effect on the stock of cash. Something unique seems to be occurring in the UK.

The UK has been switching to polymer notes recently, the new £10 being introduced in 2017 and the £5 in 2016. Old paper versions can no longer be spent. The £20 is slated for a switch in 2020. Perhaps this is creating havoc with people's money holding patterns? I suppose it's possible, but here in Canada we went through the whole polymerization process without a hiccup. (See chart here). So I don't see why the UK would experience any sort of discontinuities during its own changeover.

The answer can only have something to do with Brexit. One possibility is that Brexit has reduced immigrant inflows and encouraged outflows, and immigrants are large users of cash. Ipso facto, cash-in-circulation has declined. The problem with this explanation is you'd need really large changes in migrations flows to see that sort of pattern in cash demand, and I am skeptical we're seeing that sort of upheaval.

Another Brexit-based explanation is that Brexit has broken the banknote. British banknotes have suffered a massive credibility shock. All those paper pounds hoarded away under Brits' mattresses, or in criminal vaults, or in foreign pockets, are just not as trustworthy as they were before. So they are being quickly spent or exchanged for other paper, say euros. Eventually these unwanted notes are resurfacing back in the UK where the Bank of England is forced to suck them back up and destroy them.This paints a particularly dour picture. It says that the Bank of England's seigniorage revenues have been permanently damaged, the short-fall having to be made up by the British taxpayer. It makes one worry about potential long-term damages to the Bank's ability to effect an independent monetary policy.

Having had some time to think about this, I think I've got a better story. The changes are indeed Brexit-induced. But the big decline we've seen over the last year isn't a sign of distrust in paper pounds. Rather, it's a reversion to trend. More specifically, the decline in cash-in-circulation so far this year is actually the unwinding of an unusual surge in cash-in-circulation that began in early 2016. Check out the chart below:

Beginning in 2016, as the political competition in the leadup to the Brexit vote intensified, banknotes-in-circulation suddenly started to rise relative to long-term trend line growth (black line). This was the fastest rate at which banknotes in circulation had increased since the 2008 credit crisis. The Bank of England's blog, Bank Underground, commented on the surge in banknote demand back in 2016.

The sudden demand to hold more cash continued through the June 23, 2016 vote into early 2017. I suspect that this was a symptom of an underlying uncertainty shock spreading through the UK economy. Brits were growing increasingly worried about the effects of Brexit. Perhaps they wanted to hold fewer deposits, or have less exposure to assets like stocks and real estate. Cash is a coping mechanism. In uncertain times it one of the few assets that offers the combination of short-term price certainty and the ability to be mobilized in an instant.

This chart from the Bank of England shows that the demand for the the £50 note (pink line) was particularly marked in 2016:

Source: Bank of England

But by mid to late-2017, Brexit-related uncertainty began to subside, and cash began to be redeposited into the banking system. UK cash usage has now returned to the long-term trendline growth rate. Going forward, I'd expect the year-to-year change in cash outstanding to return to its habitual 5%-ish per year. That is, absent more Brexit-induced panics.

For much of this post, I am indebted to this great round of conversation on Twitter:

Friday, September 7, 2018

"The Narrow Bank"

 A strange new bank called TNB, or The Narrow Bank, recently applied to get a clearing account at the Federal Reserve Bank of New York, only to be refused. Funny enough, TNB is run by the New York Fed's former director of research James McAndrews, who left in 2016 in order to get the bank up and running. McAndrews and TNB are now suing the New York Fed.

There's a backstory to all of this kerfuffle. While still employed by the New York Fed, McAndrews coauthored a paper in 2015 entitled Segregated Balance Accounts. The paper proposed a solution to the following problem. Interest rates in wholesale lending markets were refusing to align with each other. Wholesale markets are the sorts of markets which neither you nor I have access to but are reserved for large institutions. For some reason, banks that kept interest-bearing overnight accounts at the Fed were not passing the rate they earned on these accounts to other overnight lending markets in which they were active, say the repo market or the federal funds market. The fed funds rate, for instance, tended to always be 0.2% or 0.3% below the interest rate the Fed paid to depositors.

Why wasn't this gap being arbitraged? After all, if a bank can deposit funds at the Fed and earn 1.95% overnight, then by borrowing in the fed funds market at, say, 1.85%, and putting the proceeds in its Fed account, said bank can earn a risk free return 0.1%. The ensuing competition to profit from this arbitrage should drive the fed funds rate within a hairline of the rate paid by the Fed to depositors. But the massive 0.2-0.3% gap implied that this trade was not being made. 

McAndrews and his co-authors posited that the fed funds market was crippled by a lack of competition. Specifically, there seemed to be a limited number of credible borrowers willing & able to wade into the fed funds market to conduct the trade. This group of borrowers was too small to absorb the funds of all the institutions that were shopping around to lend in the fed funds market. For the most part these lenders did not qualify to get interest from the Fed and were confined to buying fed funds. Thus the small group of borrowers operating in this market exercised a degree of bargaining power over the lenders, allowing them to extract artificially low borrowing rates.

The idea behind the paper was to have the Fed fix these rate distortions by re-introducing competition among borrowers in overnight wholesale lending markets. In short, all those banks that were not considered sound enough to qualify as fed funds borrowers would be able to partner with the Fed to offer risk-free accounts. Specifically, these banks would be able to go to a wary lender and say, "hey, if you lend to us we'll keep your funds hived off from all of our other assets by just depositing them directly at the Fed."

To sanctify this promise, the Fed would create a new type of account, a segregated balance account, or SBA. Once a customer had deposited funds at the the borrowing bank, the bank would transfer these funds into an SBA at the Fed. If the borrowing bank went bust, the swarm of creditors pursuing the bank's assets would not be able to touch the funds locked into its Fed SBAs. The bank itself could not use the funds in an SBA for any other purpose than paying back its customer. By hiving off a wary customers' funds, a risky bank could emulate a Fed account and re-enter wholesale lending markets.

The interest that the bank earned on SBAs would be passed-through to its customer, less a small fee incurred by the bank for providing the service. So if the Fed was paying 1.95% on deposits, the bank might be able to offer 1.90%, thus keeping 0.05% for itself. And since borrowers in the fed funds market were only offering 1.75%, say, then lenders would would avoid them, preferring to invest their funds at banks that offered an SBA solution. To compete, a borrower in the fed funds market would have to offer at least 1.90% themselves. Thus the various wholesale interest rates would be in better alignment.


Maybe upper level Fed officials took McAndrews aside and said, hey James, we're not going to implement this idea. And he thought to himself, but this is a good idea, why don't I run with it by setting up a private bank. I'm not sure, but whatever the case McAndrews quit the Fed and co-founded The Narrow Bank in what seems to be an effort to implement a market-provided version of SBAs.

TNB is a designed as a pure warehousing bank. It does not make loans to businesses or write mortgages. All it is designed to do is accept funds from depositors and pass these funds directly through to the Fed by redepositing them in its Fed master account. The Fed pays interest on these funds, which flow through TNB back to the original depositors, less a fee for TNB. Interestingly, TNB hasn't bothered to get insurance from the Federal Deposit Insurance Corporation (FDIC). The premiums it would have to pay would add extra costs to its lean business model. Any depositor who understands TNB's model wouldn't care much anyways if the deposits are uninsured, since a deposit at the Fed is perfectly safe.

In theory, TNB (and any potential copycat) should fix the competition problem that McAndrews and his coauthors alluded to in their Segregated Balance Accounts paper. Presumably all those lenders in the fed funds market that can't find suitably sound borrowers, and thus submit to being gouged by the only banks that qualify, will turn to TNB. After all, TNB is clean. Unlike regular banks, it doesn't partake in all of the traditional banking activities that make a bank risky, such as lending to consumers or businesses, or trading for their own accounts. TNB does one thing only, it acts as a portal to the Fed. Since TNB collects 1.95% from the Fed and has minimal costs, it should be able to pay interest of around 1.90% to its customers, who might otherwise get a paltry 1.75% from competing borrowers operating in the fed funds market. Thus the presence of TNB should remove, or at least minimize, some of the distortions in wholesale lending markets.   

But all is for nought. The Fed has refused to grant TNB a master account. John Cochrane has recently blogged about this as well as helpfully uploading the lawsuit that TNB has filed against the New York Fed. We don't know why Fed officials are dragging their heels, so all we can do is speculate. Cochrane has a few theories, including potential worries among Fed officials about controlling the size of its balance sheet.


But even if TNB succeeds in its lawsuit, there is a larger threat. The gap the bank is trying to exploit is shrinking. Back in 2016 when McAndrews and his colleagues first embarked on the effort to build a new bank, the fed funds rate was typically 13 to 14 basis points below the rate offered by the Fed. Fourteen basis points was a lot of rope for TNB to work with. But this gap has since shrunk to just 4 basis points (see chart below). Possibly wholesale markets have become more competitive while the bank was being constructed, in which case there may no longer be much of a role for TNB to play. If TNB borrows at 1.91% and invests at 1.95%, that doesn't leave it much wiggle room to pay its fixed costs and salaries.

Even if the gap disappears, could TNB serve as more than just a conduit for engaging in arbitrage? Let's say that in the future rates have normalized. Banks now offer to lend at an overnight rate that is in-line, or even exceeds, the rate that the Fed pays to depositors. TNB no longer has a sweet deal to offer. Even then, large institutions who can't directly bank at the Fed may like the idea of keeping an account at TNB. Although they will earn slightly less then they otherwise would in competing overnight markets, the Fed is a risk-free place to park one's money, unlike say the fed funds market. These institutions could also invest in treasury bills. But even though a treasury bill would provide a higher return than parking funds at the Fed, there is always a risk that it cannot be immediately sold for its face value. Put differently, a treasury bill has duration risk. Funds held at the Fed via TNB have no duration risk. They can be withdrawn in a moment at par.


How big might this demand be? Interestingly, TNB isn't the first of its kind. On Twitter, Karl Storvik informs me that an analog exists in Norway, the Safe Deposit Bank of Norway. SDBN is a self described "conduit" established in 2013 to provide ultra-high net worth individuals, asset managers or corporate treasurers a means to park funds at the Norges Bank, Norway's central bank.

According to the SDBN's website, its license prevents it from holding any other asset than Norges Bank deposits. The interest that the central bank pays on these deposits flow back to SDBN's customers, SDBN taking a fee for itself. This is basically TNB, Norwegian style. But as best I can tell, SDNB's function isn't to arbitrage small differences between the rate of interest that the Norges Bank pays and other overnight rates. It is trying to provide a product that is in and of itself useful to folks like high net worth individuals and corporations.

From a glance at its most recent balance sheet, I'd say that The Safe Deposit Bank of Norway hasn't been terribly successful. Sure, it is still in start-up phase, but as of the end of 2017 it had only NOK 53 million on deposit at the Norges Bank, or a piddling US$6.3 million. Assuming TNB gets Fed approval, one wonders if this wouldn't be its fate as well.


Matt Levine has an interesting take on the whole thing. What if TNB were to allow regular folks like you and me to open an account? The overhead involved in serving a retail customer base would be higher than if TNB served a purely institutional clientele, notes Levine: "you’d need at least a website, a customer service department, ATM cards—but the opportunity is intriguing." But unlike a regular bank it wouldn't need to hire loan evaluators or absorb credit losses. So TNB might be able to provide many of the same payments capabilities as a regular bank (debit card payments, ACH payments, and wire transfers), but pass through a larger share of central bank interest payments to depositors.

If it went this route, TNB wouldn't be the first financial institution to operate as a narrow bank, i.e. to swear off lending in order to focus solely on satisfying the public's payments requirements. This is exactly what mobile money platforms like M-Pesa do. Mobile money providers accept incoming customer funds, park this money in trust at a bank, and issue 100%-backed liquid IOUs to the customer. These IOUs can be used to buy stuff at retailers or exchanged with other users on a person-to-person basis. Unfortunately, liquid deposits held in trust at the commercial bank don't yield much interest, so even if a mobile money operator wanted to flow some interest through to its customers it wouldn't have much to draw on.

The novelty introduced by a retail-facing TNB is that the customer's funds would be parked directly at the central bank instead of an intervening commercial bank. So central bank interest payments could flow straight to the narrow bank rather than being sucked up by an intermediary. And so it would be possible, in theory at least, for TNB to offer retail depositors not only a useful payments option but also a financially meaningful flow of interest.

That seems like a decent financial innovation, no? For instance, the Bank of Canada currently pays 1.25% to banks that have clearing accounts, while I make a meagre 0.15% on my no-fee chequing account. If a Canadian version of TNB could offer me a 1% interest rate on an absolutely-no-frills account with a debit card attached to it, I'd definitely consider it. If James McAndrews and TNB get rebuffed by the Fed, maybe they should come up here and try the Bank of Canada.

P.S. By coincidence, I recently wrote about some of James McAndrews work on financial privacy at the Sound Money Project. And he commented on my Cato Unbound proposal to introduce taxed $500 and $1000 banknotes. Small world.