Thursday, 28 March 2013
Tuesday, 26 March 2013
This Monday, I was a guest at the weekly University of Texas Mises Circle. Originally, this was supposed to be attended by Konrad S. Graf and Patrik Korda, who have opposing views regarding the Mises' Regression Theorem and Bitcoin. However, Graf couldn't attend in the end and Korda had connection problems, so instead of being a passive observer as I originally intended, I had to kind of supplement for both of them. I apologise for my subpar performance, I wasn't expecting to talk so I didn't do any preparations, I screwed up a couple of times, such as with the "use value" or the confusion about credit vs. claim. But it looks like it's mostly understandable, so feel free to enjoy. Thanks to all the attendees, the grinning Jeffrey Tucker attending from a flying airplane, and Mattheus von Guttenberg for his contributions to the debate.
PS. I'm writing this blog post from my mobile phone to test how well it works, if you see any problems let me know.
Monday, 25 March 2013
Update #1: The interview aired, I've been getting positive responses and I am allowed to put up a translation later on my blog. The recording isn't available yet but it probably won't take longer than a day or two. I didn't hear it myself, as I don't have a radio and don't live in Slovakia.
Update #2: The recording is online: http://rozhlas.sk/radio-slovensko/k-veci/K-veci:-v-Pitsburghu-o-U.-S.-Steel...?l=2&i=61204&p=1, the part about Bitcoin begins at around 7:12.
This post is a third reaction to Patrik Korda's article about Bitcoin. There have been three main topics in our debate:
- regression theorem
- competition under network effect
- money substitutes
Definition of money substitutes
The fact that is peculiar to money alone is not that mature and secure claims to money are as highly valued in commerce as the sums of money to which they refer, but rather that such claims are complete substitutes for money, and, as such, are able to fulfill all the functions of money in those markets in which their essential characteristics of maturity and security are recognized. It is this circumstance that makes it possible to issue more of this sort of substitute than the issuer is always in a position to convert. And so the fiduciary medium comes into being in addition to the money certificate.Korda interprets this in such a way that Mises was trying to show that money substitutes compete with money in the narrower sense. But this is an incomplete interpretation. If we view it from the perspective of the pricing mechanism, we realise that Mises was also trying to explain that the price of money substitutes is not determined by the same mechanism than that of money in the narrower sense. The price of money substitutes is derived from the price of the price of money in the narrower sense. That is the essence of the quote.
Now, Mises, and the Austrians that follow him, have a more precise definition of money substitutes, they argue that they are "mature and secure claims to money [in the narrower sense, ed]". In my thesis, I argued that this is imprecise from both sides: first, not all mature and secure claims to money are money substitutes (it's unlikely if the claim does not decrease transaction costs), and second, causal relationships other than claims also may result in money substitutes (so-called "complementary currencies" for example). But I didn't analyse the question of price in that depth. I assumed that the price is the same (now, other Austrians do that too, as I thoroughly quoted from in my previous article, so don't blame me, I didn't start it).
Korda argues that the price of money substitute doesn't have to be the same as the money in the narrower sense. I countered that the price of the money substitute can be (in the extreme) lower, but it can't be higher. Korda argued that Mises himself showed examples of money substitutes that trade at a premium. I was kind of rushing my argument, so I didn't really think it through and I must admit he has a point. I didn't find a Mises quote to the same, but I found mises.org wiki article on the Bank of Amsterdam which confirms it.
This caused me to think more deeply about the issue of price. Austrians in general assume that money substitutes have the same price as money in the narrower sense. There are few exceptions. Rothbard, I believe in America's Great Depression, argued that some claims must only be considered from the point of view of money supply only at a discounted rate. JP Koning, on the other hand, argues all the time that different forms of money have different liquidity levels and also different prices. So, the argument that the price needs to be exactly the same certainly is flaky. Heck, if Korda is right in attributing the report of premium on bank notes to Mises, Mises contradicted himself. That's a problem, and it needs to be fixed. I am pretty strict about having consistent definitions.
So how do we fix it? We define money substitutes from the same point of view that caused Mises to create the classification: the mechanism of establishing of price. But how to do that to allow for fluctuating prices? We borrow the concept of derivation from mathematics. In other words, if a price of money in the narrower sense changes, ceteris paribus, the price of the money substitute will change proportionately (i.e. there's a linear dependence). This allows for the possibility of a price change for another reason. If there is a perception of risk in the claim, the substitute will trade at a discount. If the money substitute is perfectly riskless, then it may appreciate to reflect the saving of transaction costs. It can't get out of these boundaries, because that would create an arbitrage opportunity. The funny thing is that the arbitrage opportunity also arises when complementary currencies (which I also count under money substitutes) attempt to trade at a fluctuating exchange rate, an argument which I made some time ago in another debate. Now, my point isn't that other factors can't influence the price, but that the price of the underlying money in the narrower sense must. I don't think that there are factors other than perception of risk and transaction costs, but I'm not building my argument to rely on that. For the purposes of this post, I will call this derivative relationship "coupling". The price of a money substitute is coupled to the price of the underlying money in the narrower sense.
In order for me to conclude my definition, it helps to investigate what happens when the price decouples, i.e. that changes in the price of the underlying money in the narrower sense won't have a proportionate effect on the money substitute anymore. Assuming that the former money substitute survives, it evolves into the new monetary base. One possibility that Mises himself mentioned was credit money, which happened when the country went off a metallic standard and people were expecting this to be temporary. Another possibility, one that many of us know from personal experience, is fiat money. In my thesis, I demonstrate this on the example of the Euro: at the end of 1998, the exchange ratios of 14 (if I calculate correctly) currencies were fixed with respect to Euro, and thus the Euro became a money substitute. After the old currencies were decommissioned, the Euro became fiat money (a separate money in the narrower sense). Yet another possibility, arguably, is commodity money: Luther and Symes can be interpreted in a way that this is what happened with the old Somali Schilling: after the central bank and legal tender laws broke down, this increased the (now competitive) production of paper notes, until the production price of those notes was very close to the market price of these paper notes. Technically, the old Somali Schilling had been fiat money, not money substitute, but I argue in my thesis that all fiat money must begin as a money substitute, and according to Symes' historical analysis, originally, Somalia did have multiple competing commodity monies, and several monetary reforms after that, which introduced paper money substitutes.
Which subtype of money in the narrower sense will be the result of a decoupling depends on the circumstances. But the point is that the coupling is a necessary feature of the money substitute. Without it, it either collapses or changes its nature, i.e. its price will be determined by a different mechanism.
What determines the price of Bitcoin?
Now, let's look back at Bitcoin. Korda argues that Bitcoin is a money substitute of all of the fiat currencies it trades against. Let's take then the dollar for example. What happens to the price of Bitcoin if the price of dollar decreases by 1%? We don't know. It may change or it may not. We certainly can't be sure that it will decrease by 1%. We can't even be sure it will change in the same direction, it could perfectly well cause people to prefer Bitcoin to USD and increase its price. The issue becomes even more apparent if we investigate what happens if the USD decreases by 1%, and the EUR increases by 1%. The price of Bitcoin is not coupled to any other money, or any other good for that matter. So it's not a money substitute.
I believe that, motivated by Korda's resolute defense, I managed to both explain and fix the gaps in Mises' arguments. I exposed the essence of Mises' arguments, and the essence of a money substitute. I also showed that Bitcoin does not have this essence, and never had.
Saturday, 16 March 2013
The interview concentrates on how Bitcoin affects both theory of practice of fractional reserve banking, and the related questions of the money supply, credit markets and the interest rate.
Tuesday, 12 March 2013
Patrik looks like a smart chap, he makes plenty of reasonably arguments, and it turns out that we actually agree on a very large proportion of arguments about economics and Bitcoin in specific. He can make on topic responses, which is not that common in internet debates. Nevertheless, in some important issues we disagree and I think these merit a closer assessment.
Classification of Bitcoin
However, should it develop into money, it could present a problem. Bitcoin is not and never was a money substitute, never had a special legal status, nor was it a claim against anybody, nor was a commercial commodity. Therefore, it would not fit into any of the subcategories of the classification.Now, merely because we cannot classify something, that does not really mean that there is an actual empirical problem. In Antifragile, Nassim Nicholas Taleb argues that the inability to formulate knowledge verbally is not a valid reason for ignoring this knowledge. I remember that during my studies I read something similar. Tacit knowledge still has an important role in our activities, and ignoring it decreases our ability to make sound decisions. It's funny that I agree with Taleb, because I'm usually pretty anal about getting the definitions right and I go mental when people cannot formulate coherently.
The issue here, however, is not some abstract classification for its own sake, so that I know what shelf to put books about Bitcoin into in a library. The purpose of the classification system provided by Mises is to assist in the economic analysis of trade, money supply, price building, liquidity and so on. From this perspective, if we insist that we must keep the number of categories the same that Mises used, the economically closest category of Bitcoin would be commodity money.
There are several reasons for this. First of all, Bitcoin never was a claim against anyone (credit) (or, if we use the second definition of money in the broader sense, it never was a nearly perfect substitute to any other money). Patrik claims that if Bitcoin trades against fiat money, it means it's a money substitute, and the issue of a exchange rate is irrelevant. However, the Austrian economists disagree:
- Rothbard in Austrian Definitions of the Supply of Money (referring to Mises):
Because, as he pointed out, bank demand deposits were not other goods and services, other assets exchangeable for cash; they were, instead, redeemable for cash at par on demand. Since they were so redeemable, they functioned, not as a good or service exchanging for cash, but rather as a warehouse receipt for cash, redeemable on demand at par as in the case of any other warehouse. Demand deposits were therefore "money-substitutes" and functioned as equivalent to money in the market. Instead of exchanging cash for a good, the owner of a demand deposit and the seller of the good would both treat the deposit as if it were cash, a surrogate for money.
Bluechip stocks, for example, can be easily sold for money, yet no one would include such stocks as part of the money supply. The operative difference, then, is not whether an asset is liquid or not (since stocks are no more part of the money supply than, say, real estate) but whether the asset is redeemable at a fixed rate, at par, in money. Credit instruments, similarly to the case of shares of stock, are sold for money on the market at fluctuating rates. The current tendency of some economists to include assets as money purely because of their liquidity must be rejected; after all, in some cases, inventories of retail goods might be as liquid as stocks or bonds, and yet surely no one would list these inventories as part of the money supply. They are other goods sold for money on the market.[emphasis added]
To paraphrase the Austrian masters, money substitutes are perfectly secure and IMMEDIATELY convertible, PAR VALUE claims to standard money which, by virtue of this immediate convertibility substitute FULLY for standard money in individual’s cash balances, and as such, are used by individuals as a surrogate for cash – namely, a thing that all other goods and services are traded for, the final payment for such goods and services on the market.[emphasis added]
- and finally, the urquelle of all, Mises in The Theory of Money and Credit:
The special suitability for facilitating indirect exchanges possessed by absolutely secure and immediately payable claims to money, which we may briefly refer to as money substitutes, is further increased by their standing in law and commerce.[emphasis added]
Admittedly, he does not use the term "par value" here, but later, he points out to the difference in price when differentiating between credit money and money substitutes:
A third category may be called credit money, this being that sort of money which constitutes a claim against any physical or legal person. But these claims must not be both payable on demand and absolutely secure [i.e. the criteria that Mises uses to define money substitutes, ed]; if they were, there could be no difference between their value and that of the sum of money to which they referred, and they could not be subjected to an independent process of valuation on the part of those who dealt with them.[emphasis added]
Bitcoin is not a claim, nor is it treated at par with anything else. It is, to use Mises' terms, subjected to an independent process of valuation. It therefore cannot be a money substitute according to Mises. The issue is exacerbated because Patrik refuses to answer the question who is Bitcoin a claim against, i.e. who is obligated to redeem it (I'm sticking to Misesian terminology here, I disagree with the Austrians that money substitute must be a claim, but that just means I put more relaxed restrictions on Patrik than Mises would).
If we insist on the Misesian definition, then we're stuck with the controversial question of whether Bitcoin is, or was, a commodity. I don't want to get into that. The alternatives that I provide might or might not mean it's a commodity, but I think it gets the economic fundament of both gold and Bitcoin more accurately. But we should at least be able to say that Bitcoin is a quasi-commodity, or that it's a commoditised service, if we insist on using "commodity".
Bitcoin is not the only medium of exchange that shows the Misesian definitions inadequate. For example, the old Somali Shillings still work as money. They do not have a special legal status anymore, they just used to have one in the past (that is how they obtained their initial value). So they used to be fiat money, but they evolved into something which a strict Misesian can't classify (quasi-commodity money, as Selgin suggests). The old Somali Shillings have been counterfeited so much that their production price is close to their market value. During the inflation the supply was elastic but now that the market price has dropped it's not elastic anymore. We see on this example that the removal of the legal status causes them to eventually gain features similar to the "real" commodity money. Another example of insufficient classification system are various complementary currencies, mutual credit, LETS, or privately issued scrip like Ithaca Hours or BerkShares. Now, these are technically not money yet, merely media of exchange (same situation as Bitcoin), but if they ever developed into money, the question of classification would arise (I use a bit more relaxed definition than Mises, so to me, they would then count as money substitutes).
I think that rather than insisting on a particular umbrella classification system, the classification should be done from the perspective which is most relevant for a particular context. In most economic debates, I tend to agree with Schlichter and Selgin that the elasticity of supply is the most relevant criterion. But I presented several other options too, they might be more suitable for different situations.
The future of Bitcoin
BTW I would be caution of not calling the increased price a bubble. The relevant factor for monetisation is liquidity and I haven't seen an improvement of liquidity (measured as bid/ask elasticity of Mt. Gox as laid out in my thesis), however I didn't analyse data after beginning of January (I don't have the full data set yet), so I am not making a final statement.This post is not publicly available but I'm sure people who have access to it can confirm that I'm not making this up.
However, Patrik makes a second argument too. He claims that this is relevant for the future of Bitcoin. And here is where we disagree. In the previous article, I argued that due to the network effect, I expect a small number of dominant cryptocurrencies to emerge, even if their relative market share and composition changes. Patrik however seems to have a more fundamental argument, in that the market shares will change quickly and unpredictably. This I also do not hold realistic, and will attempt to address it in this section.
This section will be a bit more loose than normally. I'm not going to make a "hard" argument, merely point out to factors Patrik missed and that might be relevant. First I will formulate a "rule of thumb" for how I think the network effect behaves under competition, and then I'll explain it on the examples presented by Patrik, and some of my own.
- If a protocol is technologically better suited to fulfill the requirements of the users, it will take the leading role
- Friction could prevent a change in the market share if the advantage is not sufficient
- If protocols are technologically too similar to provide advantages vis-a-vis each other, a lagging one won't overtake the leading one (path dependence)
- State interference can count as technological advantage
- Multihoming (using of more than one standard simultaneously) is only possible to a certain extent, and can't affect the fundamental issues long term
- 1999-2001: Napster. First mover advantage.
- 2001-2004: Kazzaa (FastTrack protocol). Won over Napster because Napster shut down.
- 2004-2007: eDonkey. I think it won because it was less centralised and Kazzaa was hostile against third party clients.
- 2007+ BitTorrent. I think it won because it is less centralised, and open source. BitTorrent is the only one that actually gained wider industry acceptance (e.g. Blizzard downloader uses it).
Wednesday, 6 March 2013
I get agitated when I disagree with others, so I wrote a rebuttal.
Competition under the network effectKorda's first argument is that because it is easy to create a new cryptocurency, they will compete each other out of the market, kinda "race to the bottom", ending up with a hyperinflation and a collapse. However, he does not seem to understand the network effect, one of the most important aspects of money. The network effect both allows that money actually exists in the first place, as well as creates switching costs. As JP Koning said, "liquidity is sticky".
The problem with Korda's argument becomes more apparent because he himself shows a counterexample. He quotes Mises in explaining that silver has been replaced by gold and this demonetised silver. This explains how competition works under the presence of a strong network effect: the expected long term state is where a small number (maybe even one) media of exchange are the dominant ones, and other market players are far less liquid. For the same reason, a situation where there is a large number of competing cryptocurrencies without clear dominant players is not a stable state, rather a small number of dominant players will emerge. The network effect is recognisable in particular with immaterial goods: there are a small dominant number of general purpose operating systems (Windows, iOS, Linux), a small dominant number of languages (Mandarin and English dominate, then Spanish and Hindi follow after a gap, and those four account for about half of the world population (I'm approximating, as people can speak more than one language)), there is only one dominant general purpose communication protocol (what we commonly call "the internet"), and so on. Surely, the composition of the dominant players can change, but it is a relatively slow process that does not magically happen overnight. Surely, there are a myriad of minor players, but there always tend to be a low number of dominant players.
If I said that everyone can create their own language, therefore without barriers to entry, everyone would end up with their own language and the ability to communicate would collapse, you'd surely think that I'm a moron.
Another important factor related to the network effect is path dependence. This means that the order in which choices are made influences the end result. This can mean, for example, the first mover advantage. Bitcoin is the first practically usable cryptocurrency, so it has a head start against others. And surely, JP Koning has a useful infographic showing that the market capitalisation of Bitcoin, the first mover, far outstrips the market capitalisation of others (by a factor of 100, at the time of making the graphic). This is also consistent with my claim from above that there typically is a small number of dominant players. Even though there is government interference in the choice of media of exchange (what we call fiat monies), international trade is affected significantly less than national, and we still have a small amount of major players (the USD and Euro).
I'm not arguing here that Bitcoin can't be replaced by something else, but that the scenario described by Korda makes no sense.
Mises' Regression theorem
Can Bitcoin become money?
Classification according to ToMCAs I wrote in my thesis, as Bitcoin is not money (yet), merely a medium of exchange, it is impossible to use the classification system of Mises to classify Bitcoin. If it becomes money, then we would have a classification problem. How to solve it I leave open in this post, as I consider it merely a theoretical question with no practical relevance. In my thesis I present options for solving it.
Now, Korda claims that Bitcoin is token money. However, going upwards from the bottom of the graphic in Mises' ToMC Appendix B, Bitcoins are not token money, because they are not fiduciary media, because they are not money substitutes. The Austrians use two definitions of money substitutes (I explain the difference between the two in my thesis):
- absolutely secure and immediately payable claims to money (in the narrower sense)
- things that act as full substitutes to money (in the narrower sense) from economic point of view
- Due to network effect, the market structure will move towards a small number of dominant cryptocurrencies, so there's no hyperinflation
- Whether Bitcoin can become money is not important, as using it is already advantageous now, as a medium of exchange. If it ever becomes money, that would really rock though.
- Non-economists do not understand the regression theorem and invent their own versions of it which are nonsensical
- Bitcoin is not token money as it never was a money substitute
- Bitcoin is pseudonymous, and has a comparative advantage against competitors from this point of view
- The price of Bitcoin is irrelevant