Monday, September 23, 2019

A fifty-year history of Facebook's Libra

Last week, we finally got some information about what Libra's currency basket would look like.
If you haven't heard, Libra is a proposed global blockchain-based payments network. It is being spearheaded by Facebook along with a coalition of other companies including Uber, MasterCard, PayPal, and Visa.

The hook is that rather than going the conventional route and expressing monetary values using existing units-of-account like the dollar, yen, pound, or euro, the Libra network will rely on its own bespoke Libra unit-of-account as its "base language." Libra originally revealed in its whitepaper that the Libra unit would be defined as a basket, or cocktail, of other currencies. Now we know what that mix will likely look like.

Interestingly, the Libra isn't the world's first private unit-of-account. Back in the 1960s and 1970s, several financial institutions came up with their own bespoke units. I learnt about this strange and fascinating episode courtesy of a very readable paper by two economists, Joseph Aschheim and Y.S. Park.

As I gathered from the paper, the first private artificial currency unit was Luxembourg-based Kredietbank's European Accounting Unit (EUA). Originally devised in 1961 as 0.88867 grams of fine gold, the EUA was soon used to denominate a bond issue by SACOR, a Portuguese oil company. Over the next two decades, Aschheim & Park claim that around sixty or so bond issues would rely on Kredietbank's EUA as their accounting unit.

Between 1968 and 1971, the U.S. Treasury ceased to redeem dollars with gold. When the Smithsonian Agreement—a band-aid attempt to re-cement all currencies to the U.S. dollar—collapsed in 1973, the post WWII system of fixed currencies came to its final end. To help people cope with the sudden babble of floating currencies, several new private units-of-account joined Kreietbank's EUA.

N.M. Rothschild & Sons kicked things off in 1973 with its European Composite Unit, or Eurco. The Eurco was made up of nine currencies issued by members of the European Community, including Deutsche marks, French francs, and Danish kronor. According to Aschheim & Park, Rothshild developed the Eurco "to elicit investors' confidence" in long-term bonds, but as of 1976 only three bond issues had been denominated in Eurcos.

In 1974 Hambros Bank introduced the Arab Currency-Related Unit, or Arcru. The Arcru was comprised of twelve Arab currencies and designed to appeal to Arab investors flush with oil profits. The next year Credit Lyonnais created a bouquet of the ten currencies, both European and non-European, and dubbed it the International Financial Unit, or IFU. This was a far more broad-based unit than the Arcru or Eurco, the relative weights of the IFU's component currencies being based on each country's share of international trade.

Barclays Bank also got into the game in 1974 with the Barclays Unit, or the B-Unit. The B-Unit was made up of five currencies: the U.S. dollar, the British pound, the German mark, the French franc, and the Swiss franc. Aschheim & Park note that whereas the Arcru, IFU, and Eurco were primarily intended for denominating bonds, the B-Unit was designed to be used for making international payments.

--------

Which makes the B-Unit a direct predecessor of the Libra unit.

Look around you today, however, and not one of these private units-of-account listed below exists. Anyone want to pay me in B-Units? I didn't think so. I think this says something quite fundamental about the market's demand for artificial currency units. Businesses and consumers don't really like to use them.

Table from Aschheim & Park

If private artificial currency units have been failures, what about government-provided ones?

Take the International Monetary Fund's Special Drawing Right (SDR) basket, which has been in existence since 1970, almost fifty years. If there was a demand to make international payments using public artificial units of account, surely commercial banks would eventually have met that demand by implementing SDR-denominated payments systems. Indeed, Aschheim & Park speculate on the possibility in their 1976 paper. It's worth reading this section in full:
"International banks may soon be willing to accept deposits denominated in SDRs because a potential demand for SDR funds already exists, as manifested by recent SDR bond issues by the Swiss Aluminum Company, the Swedish Investment Bank, and Electricite de France. The process, indeed, is already under way. In July 1975 the Bank Keyser Ullmann in Geneva (a subsidiary of Keyser Ullmann of London) announced that it would henceforth accept demand and time deposits denominated in SDRs. These SDR deposits are to be convertible at any time into any currency at the SDR exchange rate applicable on that day. Similarly, in August 1975 the Chase Manhattan Bank in New York instituted a range of banking facilities in SDRs, including loans, deposits, and futures trading. As this process spreads and as more international transactions are denominated in SDRs, banks may begin to allow direct transfers between SDR accounts, internally and then between banks. In consequence, the SDR may be transformed from mere numeraire (international quasi-money) into an outright means of payment (full-fledged international money)."
Again, look around you today. How many banks let you open an SDR-denominated bank account and make SDR payments? None that I'm aware of. Maybe the IMF's SDR was never well designed, or maybe Barclays was too small to drive B-Unit adoption. Or more likely SDRs, B-Units, and the other artificial currency units mentioned in Aschheim & Parks paper are all monetary dead-ends. In pursuing the same path, Libra could be making a big mistake.

--------------

What is it about artificial currency baskets that makes them non-starters? My first post about Libra delved into this question. Let me repeat my argument below to spare you the effort of clicking through.

In an alternative reality, let's imagine that Facebook only allows users to join and converse on its platform after having learnt Facebook's artificial language, Facebookish. English, French, Chinese, and all other languages are banned.

In this alternate reality, all Facebook users understand each other because each one is fluent in Facebookish. Comprehension is a great thing. But hardly any of us would be on Facebook to begin with. Who wants to go through the effort of learning a new language? Not me.

In the real world, Facebook has long since decided against the Facebookish approach. Instead, it supports a multitude of local languages—Arabic, Chinese, English, Hindi, and more. Sure, the drawback is that we can't always understand what other Facebook users are saying. But at least users don't have to go through the hurdle of learning new grammar and syntax. And Facebook has thrived as a result of this simple and obvious design choice.

The adoption of a Libra unit of account is the monetary equivalent of forcing users to learn Facebookish. Sure, at least with Libras we'll all be using the same currency units. But this ignores the costs we'd all have to incur as we learn a new monetary patois. From a very young age we all figure out how to "speak money". We speak in our local unit-of-account. As a Canadian, the Canadian dollar has always been the means by which I describe prices to people around me, and remember values, and engage in cost-benefit calculations. Facebook wants to force us all to learn a new monetary language, a Libra-based one. But in doing so it's setting a huge hurdle to adoption.

So I'll just repeat. No matter how skillfully it goes about designing Facebookish (or Libras), artificial languages and artificial units are dead-ends. They're utopian, and definitely not user-friendly. (Ok, I may have described it all better in my original post, so just head on over.)

----------

That being said, over the last few months I've been slowly warming up to Libra. Out of the millions of crypto projects that have come out over the last decade, it comes close to the Fedcoin vision I originally outlined on my blog back in 2014, and twice now for R3.

To begin with, a Libra token would be stable (unlike bitcoin) thanks to credible and strong issuers. Since it would be decentralized, the network would be resilient. And since a Libra is a token, and not an account, it should be relatively open for everyone to use. At the same time, David Marcus, the architect behind Libra, is making the right noises about financial privacy. (Whether his intentions are genuine or not, it's tough to say.)

From the Libra whitepaper

I think (and I could be wrong here) that there is a growing desire on the part of consumers for more financial privacy. Unfortunately, governments hew to a post-9/11 mindset that regards privacy as a pervasive threat. Facebook may be one of the only organizations with the financial heft to articulate consumers' desires for more privacy in a way that regulators can't ignore.

Having Facebook as financial privacy advocate is a fragile win, no? It would be too bad if Libra (and whatever level of financial privacy it promises to bring to mainstream consumers) never attains widespread usage because of a basic design flaw, one that obligates us all to adopt the monetary-equivalent of Facebookish

-------

If not an artificial currency basket, what should Facebook do? I think that most consumers who engage in cross-border transactions want to keep swimming in their domestic currencies up until the last minute. Only at the 'buy now' or 'send now' moment—i.e. when a purchase it to be consummated or funds transferred to a friend—do we want to leave the bubble of our home currency. Pre-accumulating some strange alien token, whether those be SDRs, B-Units, or Libra, just isn't on the table.

If it wants to stay customer friendly, Libra needs to design its network to allow for the flow of tokens denominated in state currencies (U.S. dollars, Chinese yuan, British pounds, Indonesian rupee). And then it needs to design a cheap, transparent, and easy way for these tokens to move from person to person. This is what PayPal does. It's also worked for Transferwise. Visa and MasterCard too. None  of these platforms have created their own curious units, PayPalios or TransferWise-units or Visa-oos. They've allowed customers to remain safely ensconced in their domestic currency bubbles until the final 'send now' moment.

-------

Putting aside my criticisms of Libra's decision to use an artificial currency unit, what do I think about its choice of basket?

I am left wondering what sort of process David Marcus and the folks at Facebook have used to generate the basket components. Potential Libra users will want to know ahead of time how they can expect a basket's components to be updated as time passes. After all, if their wealth is to be held on the platform, customers will wonder what is to prevent a sudden rewriting of the basket in a way that favors the network at their expense?

One important rule that everyone will want to know is what economic thresholds are being used to filter out or include various currencies. For instance, if the Korean won starts to become a popular international currency, at what point will Libra decide to include it in the basket? If so, would it boot out another currency to make way for the won, or keep it?

The current Libra components are definitely odd, and give no indication of what process the architects are using to populate the Libra basket. For instance, I'm not aware of any selection process or rule that would lead to the Singaporean dollar comprising 7% of what is supposed to be a "global currency." Don't get me wrong. I like Singapore. It punches above its weight. But Singapore doesn't account for 7% of world trade, or 7% of the world's population, or 7% of global anything.

Or why does the euro account for just 18% of the Libra basket while the U.S. makes up a mammoth-sized 50%? The European Union has twice the population of the U.S. and accounts for a far larger share of exports. And where is the Chinese yuan? Exiled for political reasons?

One wonders if the euro's small share has to do with the effect that Europe's negative interest rates might have on network profits. For each Libra it has issued, the consortium will have to keep a Libra's worth of assets in reserve. Far larger profits can be earned it it reduces the euro portion of the basket and increases the U.S. dollar portion. After all, that would mean more exposure to high-yielding U.S. dollar assets and less to negative-yielding European ones. But that's a terribly ad hoc way to construct a currency basket.

My last thought is this. If Libra has its heart set on choosing an artificial currency unit as the basis for its global currency, it should have probably just go with the IMF's SDR basket rather than brewing its own strange currency concoction.

The IMF's SDR basket (source)

Consider how exchange-traded funds which track an index outsource all of the decisions about index methodology and components to third-parties like Standard & Poors, MSCI, and FTSE. This makes the exchange-traded fund more credible. Using SDRs would pre-commit Libra to avoiding conflicts of interest and thorny politics, the IMF becoming the theater for determining the basket. One could find worse third-parties than the IMF.

Wednesday, September 18, 2019

The life and death of an internet monetary meme


Over the last few years I've increasingly crossed paths with the following claim on the internet: "The average life expectancy for a fiat currency is 27 years." Is this claim true? What definitions are being used? I mean, are we talking about inconvertible paper money here, or currency that was convertible into gold, too?

I finally got curious enough that I decided to chase down the source of this meme. After all, without knowing what data it is based on, it's hard to evaluate the claim's truthfulness. Below I give a description of my trek through internet history.

The average-life-of-fiat meme has become particularly popular among cryptocurrency types. For instance, here is Jimmy Song, a popular bitcoin educator/developer, confidently invoking the slogan back in 2017:
"When a society lacks prudence, what happens is that the society collapses or goes into chaos. It’s not a coincidence that the average lifespan of a fiat currency is only 27 years."
A long list of cryptocurrency luminaries have dutifully mentioned the meme including Dan Held (2018), Taylor Pearson (2019), Barry Silbert (2019), Tuur Demeester (2015), Francis Pouliot (2018), and Adam Back (2019). Grayscale Investments, a firm that provides cryptocurrency-based investment products, even includes it in their marketing material:

As is often the case these days, crypto bugs have cribbed their ideas from their older cousins, the gold bugs. Nathan Lewis, author of Gold: The Once and Future Money, mentioned the idea in written testimony to Congress in 2012. Ralph Benko, a gold standard advocate, invoked the meme in a 2011 article. And Max Keiser, a long-time gold bug turned cryptocurrency advocate, began mentioning it as early as 2013. Where did Keiser, Benko, and Lewis get the meme from?

One of the meme's earliest and most cited appearances comes from Washington's Blog, a platform for a group of anonymous financial writers. We know little about this group except for the fact that George, "website owner and lead writer – is a busy professional, a former adjunct professor, an American and a family man." 

In August 2011, Washington's Blog published an article with the brutally long title The Average Life Expectancy of a Fiat Currency is 27 Years... Every 30 to 40 Years the Reigning Monetary System Fails And Has To Be Retooled. The author failed to explain how the 27 years claim was derived. Instead, he/she relied on another article by a writer named Chris Mack for backup. In a disclaimer the author noted that "I don't know Chris Mack," and thus couldn't vouch for the figures. "However," he/she went on to say, "the general concept is correct."

That's an odd way to do analysis, no? I've to this number for you, 27. I don't know how the number was generated, or who came up with it. But it's good enough. So go ahead and use it.

After a bit of hunting, I found Chris Mack's article here (the link at Washington's blog is dead). It is dated January 2011, pushing back the meme's genesis by another few months. At the time, Mack was President of Trade Placer, a "real-time marketplace where you can buy or sell items such as gold, silver, platinum, wine and other collectibles." Now he is the CEO at Levidge, a platform for trading cryptocurrencies. Note again the well-trodden corridor between gold buggery and crypto buggery.

Anyways, in 2011 Mack wrote:
"According to a study of 775 fiat currencies by DollarDaze.org, there is no historical precedence for a fiat currency that has succeeded in holding its value. 20 percent failed through hyperinflation, 21 percent were destroyed by war, 12 percent destroyed by independence, 24 percent were monetarily reformed, and 23 percent are still in circulation approaching one of the other outcomes.

The average life expectancy for a fiat currency is 27 years, with the shortest life span being one month. "
Mack mentions a study by DollarDaze, but doesn't provide a link to the article. Aha, the missing data! I hopped over to DollarDaze's website, a blog dedicated to talking about the failings of the U.S. dollar. But there are no blog posts older than 2018. No study, folks.

This puts all the meme users – Jimmy Song, Grayscale, Max Keiser, and the rest – in an absurd situation. There are some words on some websites about a study, but ask our meme users where the study is and none of them can actually find it. Did it ever exist? Why are they so confidently transmitting data when there is no data? DollarDaze has got to be right, no? How could you doubt it, JP?

Since no one was able to help me, I turned to the Wayback Machine to see if I could pull up older versions of the DollarDaze website. It took a while, but I finally found pay dirt. Back in 2009 the editor of DollarDaze, Mike Hewitt, wrote an article entitled The Fate of Paper Money. I tried to track Hewitt down, but he seems to have disappeared from the internet.

No matter. Finally, some data to evaluate! In his article, Hewitt claims to have counted 176 currencies in circulation and 599 currencies that are not in circulation. Of the 599 in his discontinued list, Hewitt comments that the "median age for these currencies is only fifteen years!" He provides links to both lists (1 and 2).

I downloaded Hewitt's list of 599 defunct currencies and put it into an Excel spreadsheet. The median age is indeed 15 years, as Hewitt claims, and the average is 27 years, as Mack claims in his subsequent 2011 article. So voila, we finally have the basis for the modern internet meme that the average age of a fiat currency is just 27 years. It's based on Hewitt's list of 599 dead currencies, with Mack taking the average duration. (They conveniently don't include the list of 176 existing currencies in their calculation, which would have increased the number). 

Now for my criticisms.

The 27-years meme has been used for many years now as a prop for making gold and cryptocurrencies look good. "Ha ha, suckers! Only 27 years until your paper is worthless!" But many of the 599 defunct currencies in Hewitt's list weren't failures. Rather, they were replaced for political, economic, and cultural reasons.
For instance, the list contains all of the pre-euro currencies (Dutch gulder, French franc, Italian lira, etc). These currencies had good reasons for disappearing: they were swapped for a new monetary unit. Existing currency holders weren't robbed. They were fairly compensated for this switch.

Another example of monetary reorganization occurred in East Africa. From 1919 to the 1960s, Britain's former east African colonies relied on the East African shilling, produced by the East African Currency Board. When these countries gained their independence, the currency board was dismantled. In its place Kenya began issuing its own shillings at par with the old ones, as did  Tanzania and Uganda.

In each case, existing owners of East African shillings could convert their holdings into new currency. No wealth was being destroyed during any of these switches. But people who throw around the phrase the average life expectancy for a fiat currency is 27 years as a criticism of the very institution of currency are using the data in a way that implicitly assumes that the East African experience – and others like it – were negative. They weren't.

So the idea that Hewitt's list somehow measures the length of time between a fiat currency's birth and its impending worthlessness is just wrong. I'd go even go so far as to say that the plasticity of the listed currencies is one of their strengths. As national borders change and political circumstances shift, the writing on the bills should be updated too. 

Hewitt's list contains many data errors. He makes the odd claim that the Japanese gold oban and silver momme were created in 1904 and met their end in a hyperinflation in 1905. But these were both historic Japanese coins that had existed for centuries. Hewitt also lists the U.S. greenback ("US Paper Dollar) as lasting from 1862 to 1878. But this isn't correct. Greenbacks were repegged to gold in 1878, but they continued to be issued for many decades after.

Or take Hewitt's categorization of British Military Authority (BMA) lira as dying in hyperinflation. This is an odd claim to make. BMA Lira were issued in Libya by occupying British forces both during and after World War II to provide the nation with a circulating medium. These notes were basically a military version of the British pound, with 480 lira equal to a pound sterling. In 1951 BMA Lira notes were converted into Libyan pounds, issued by Libya's new currency board, at a rate of 480-to-1. No hyperinflation here.


Finally, take Hewitt's claim that the Hawaiian dollar was "destroyed" by WWII. Not so. I've written about the Hawaiian overprints before. To protect against a potential Japanese invasion of Hawaii (and a confiscation of dollars by Japanese soldiers) all dollars on Hawaii were overprinted. Once the threat of Japanese invasion had disappeared they were swapped for regular U.S. dollars and withdrawn . But not a single Hawaiian lost anything during the entire process.

I don't want to nitpick too much, but given that it only took me a few minutes to find these four mistakes, one can only conclude that the rest of Hewitt's list is riddled with errors.

Finally, there are some semantic issues. Fiat currency is generally considered to be inconvertible money. It can't be redeemed for gold or silver. The world really only shifted onto a fiat standard between 1968-71 as the dollar ceased to be redeemed in gold. But Hewitt's list is replete with many metallic currencies (i.e. the riksdaler riksmynt). Are people using his data to make a claim about currencies in general, or just fiat ones? The meme isn't clear on this.

So having examined the life of an internet monetary meme, I'd like to kill it. It's time for us to retire the idea that says that "the average life expectancy for a fiat currency is 27 years." God knows there's plenty of problems with currencies. But good criticism requires diligence and accurate data. The meme in question is an example of sloppy work and bad data.

I know that the crypts and the bugs and the fiats are engaged in constant meme warfare, the bugs and the fiats for many decades now, the crypts only joining the battle a few years ago. Messages must be crafted for best efficiency, whether this be to pump bitcoin to the moon, or to push gold into the stratosphere, or to lock the fiat system in place. But most of the serious people involved in these debates, no matter which side, still keep at least one foot in the truth. Let's flush the 27-years meme down the toilet, folks.