Bitcoin's security pricing problem
I can't resist this.
That is an appreciating currency. Each unit of BTC now buys more USD than it did a year ago - rather a lot more, actually. Which means that each unit of BTC buys more goods and services than it did a year ago. Were the prices of goods and services quoted in bitcoins, we would see price falls, not rises. This is deflation, not inflation.
So if Bitcoin is deflating, not inflating, what on earth does Buterin mean when he says that Bitcoin is paying for security through inflation?
He does not mean economic (price) inflation. He means MONETARY inflation - sustained expansion in the supply of money. Bitcoin has a hard limit on the number of bitcoins that can ever be mined, but it is nowhere near that limit. At the present time, the supply of bitcoins is expanding at the rate of 12.5 every 10 minutes.
What Buterin means by "inflation" paying for security is that Bitcoin miners are rewarded for their validation of transactions by being given additional bitcoins, created "ex nihilo" as a result of their work ("proof of work") - which in practice means their expenditure of energy in validating transactions back through the chain.
Now, we can argue about whether monetary inflation matters, and if so to what extent. In economic circles, the jury is out on this, but Bitcoin is ideologically driven by hard-money Austrian principles, so we might expect that its adherents are very concerned to limit the supply of money. But my objection here is to simple sloppiness in terminology. When people say "inflation", they generally mean price inflation. So if you mean monetary inflation, Vitalik, say so.
Ok, so we've cleared that up. Now on to my main objection to this piece.
Buterin is essentially discussing the relative merits of two different ways of paying miners for the validation of transactions: additional bitcoins, or transaction fees. There is a lot of geekery here, so I'm going to do my best to translate.
When Bitcoin reaches its hard limit on the supply of bitcoins, all future transaction validation will be rewarded with transaction fees, not additional bitcoins. Personally I am of the opinion that the switch to transaction fees might occur before then, since the cost of mining additional bitcoins will rise exponentially as the hard limit is approached, making it prohibitively expensive to mine them - this is a fine example of the law of "diminishing returns". But I could be wrong. Anyway, the point is that unless the Bitcoin community agree to raise the limit - which is the subject of fierce dispute at present - at some point Bitcoin will switch to a transaction-fees-only reward model.
Buterin's concern is that in a transaction-fees-only model, the fees need to be set high enough to discourage malicious operators from taking over the network by making miners offers they can't refuse. How much would it cost to take over the entire network? Buterin has some ideas:
Of course, Ethereum's proof-of-stake (investment) is MUCH better, natch:
This is the uncomfortable truth that Buterin is trying to avoid. The security he thinks his system needs is expensive. Users might not be willing to pay for it.
In a transaction-fee-only system, fees that are high enough to discourage miners from selling out to a higher bidder may be too high to discourage users from transferring to another system. Losing your users is every bit as bad as losing your miners. So having discarded the hard cap on the coin supply, Buterin thinks he has a solution to the problem of expensive transaction fees (or poor security). He wants to identify the optimum mix of new coins and transaction fees in miners' rewards. And he tries to use Ramsey pricing to determine it.
In a typical Ramsey pricing model, two different products are transported on the same railway network. The railway network is a monopoly with high fixed costs: if the price charged to transport the products were set at the marginal cost of their production, the railway would lose money. So a percentage markup is added which is sufficient for the railway to break even. We assume this forces the producers of the products to raise their end prices. However, the two products have different price elasticities of demand: one has a much higher value than the other, and hence a higher price elasticity of demand (which means that people are more likely to refuse to buy it if the price goes up). Clearly, if the railway charges the same percentage markup on both products, sales of the higher-value product will slump relative to those of the lower-value. To prevent this - since it is discriminatory - Ramsey pricing puts a lower markup on the product with the higher price elasticity of demand, and a higher one on the product with the lower price elasticity of demand.
Sadly, Buterin's example misconstrues the context of Ramsey pricing. The differential markup is not because of social harm to purchasers of the products: there is no captive audience. No, the differential markup is to protect the producers of the higher-value good, since the higher price elasticity of demand for their product means that the railway's markup would cause them greater loss of sales than the equivalent markup on a product with lower price elasticity of demand. It's the equivalent of a variable sales tax - indeed Ramsey pricing is sometimes used in tax policy.
Clearly, we do not have a similar situation here. What we have is a single good (coins) which are transported at a fixed fee (transaction fee). The problem is simply one of supply and demand. If the price demanded by miners for supplying security is too high, users will vote with their feet. If the amount that users are prepared to pay for security is too low, miners will sell out to attackers. There should be some equilibrium point at which most users will pay an amount that most miners are happy with: there will remain a tail risk of takeover, just as there will remain a tail risk that the crypto equivalent of Uber entices away the users by massively undercutting the transaction fees. Equilibrium is NOT a risk-free position.
Buterin thinks that it is not possible to set transaction fees high enough to reduce the risk of attack to acceptable levels. I might say that in a decentralised democratic system, if users don't want to pay for security, they shouldn't have to. If the consensus is that a raised risk of attack is an acceptable price to pay for lower transaction fees, that is how it should be. Buterin has no right to start dictating to people what level of attack risk they should be allowed to take.
And he is quite wrong to suggest that mining rewards (new coins) can be used to subsidise transaction fees. Coin rewards for miners are in fact an additional cost for the users. This brings me back to where I started in this article. He misused the term, but he was right about inflation - though he didn't follow it through.
Classically, expanding the money supply is associated with price inflation. In a whole economy, the relationship is very complex. But in this case, it is much less so. BTCUSD has been rising constantly over the last year, which shows us that the rate of expansion is less than the rate of increase in demand for the coins. Nonetheless, the price would have risen more if no new coins were produced. Even though Bitcoin is deflating, the difference between the price (in USD) of 1 BTC when the money supply is expanding and the price (in USD) of 1 BTC when the money supply is fixed is a real cost to the users of the system, since it means their bitcoins don't buy so much. It is equivalent to a higher transaction fee.
So Ramsey pricing just doesn't apply. What we are really dealing with here is human preferences. Do miners prefer to receive transaction fees or new coins? Do users of the system suffer from money illusion - so prefer to have lower transaction fees and a higher rate of price inflation (or lower rate of deflation)? It should be possible to determine some mix of coin rewards and transaction fees that keeps most people happy. And it is entirely conceivable that Bitcoin might choose to operate with transparently higher transaction fees as the price of retaining its hard cap, while Ethereum chooses to run with lower fees and some price inflation. That is for the respective communities to decide.
In fact, what Buterin is trying to solve here is exactly the same conundrum as central banks face (and have never solved). What is the optimum balance of interest rates and inflation? Should we fix interest rates (transaction fee) at some level, and allow inflation (coin supply) to flex? Or should we fix inflation (coin supply) and allow interest rates (transaction fee) to flex? Or some combination of flexible coin supply AND transaction fee?
Whatever approach is adopted, in a simple niche system like Bitcoin or Ethereum the equilibrium would be pretty much the same. The system would have to be far more complex for the choice of target to make much difference. The mix of coin rewards and transaction fees is largely irrelevant: it is the total cost of security that matters, not how it is paid for. In his concern to protect his system from attack, Buterin is chasing a chimera.
Blithely ignoring the utter mess he and his developers have managed to make of the cryptocurrency Ethereum, Vitalik Buterin has written a post on inflation and monetary policy. Wow. Is there no end to his talents?
In this case, there is most definitely an end. Economics 101 is the end, pretty much. I don't claim to be the world's greatest economic expert - not by a LONG way - but the errors in this piece leapt out at me.
Firstly, inflation. Here is Buterin on inflation:
The primary expense that must be paid by a blockchain is that of security. The blockchain must pay miners or validators to economically participate in its consensus protocol, whether proof of work or proof of stake, and this inevitably incurs some cost. There are two ways to pay for this cost: inflation and transaction fees. Currently, Bitcoin and Ethereum, the two leading proof-of-work blockchains, both use high levels of inflation to pay for security; the Bitcoin community presently intends to decrease the inflation over time and eventually switch to a transaction-fee-only model.
Inflation? Really? The cryptocurrency whose adherents promote it as the world's greatest ANTI-inflationary currency is suffering from inflation? Surely not!
Definitely not. Let's remind ourselves of the economic definition of inflation. Since Buterin seems to like referencing Wikipedia, this is Wikipedia's definition:
Definitely not. Let's remind ourselves of the economic definition of inflation. Since Buterin seems to like referencing Wikipedia, this is Wikipedia's definition:
In economics, inflation is a sustained increase in the general price level of goods and services in an economy over a period of time. When the price level rises, each unit of currency buys fewer goods and services. Consequently, inflation reflects a reduction in the purchasing power per unit of money - a loss of real value in the medium of exchange and unit of account within the economy.....The opposite of inflation is deflation.So, if Bitcoin is inflating, it should be buying less of whatever is purchased with it. Bitcoin can be used to purchase some goods and services, but as it is not yet widely accepted as a medium of exchange (MOU), the best way of establishing its purchasing power is to look at its exchange rate to a widely accepted MOU, such as the US dollar. This is BTCUSD over the last year:
That is an appreciating currency. Each unit of BTC now buys more USD than it did a year ago - rather a lot more, actually. Which means that each unit of BTC buys more goods and services than it did a year ago. Were the prices of goods and services quoted in bitcoins, we would see price falls, not rises. This is deflation, not inflation.
So if Bitcoin is deflating, not inflating, what on earth does Buterin mean when he says that Bitcoin is paying for security through inflation?
He does not mean economic (price) inflation. He means MONETARY inflation - sustained expansion in the supply of money. Bitcoin has a hard limit on the number of bitcoins that can ever be mined, but it is nowhere near that limit. At the present time, the supply of bitcoins is expanding at the rate of 12.5 every 10 minutes.
What Buterin means by "inflation" paying for security is that Bitcoin miners are rewarded for their validation of transactions by being given additional bitcoins, created "ex nihilo" as a result of their work ("proof of work") - which in practice means their expenditure of energy in validating transactions back through the chain.
Now, we can argue about whether monetary inflation matters, and if so to what extent. In economic circles, the jury is out on this, but Bitcoin is ideologically driven by hard-money Austrian principles, so we might expect that its adherents are very concerned to limit the supply of money. But my objection here is to simple sloppiness in terminology. When people say "inflation", they generally mean price inflation. So if you mean monetary inflation, Vitalik, say so.
Ok, so we've cleared that up. Now on to my main objection to this piece.
Buterin is essentially discussing the relative merits of two different ways of paying miners for the validation of transactions: additional bitcoins, or transaction fees. There is a lot of geekery here, so I'm going to do my best to translate.
When Bitcoin reaches its hard limit on the supply of bitcoins, all future transaction validation will be rewarded with transaction fees, not additional bitcoins. Personally I am of the opinion that the switch to transaction fees might occur before then, since the cost of mining additional bitcoins will rise exponentially as the hard limit is approached, making it prohibitively expensive to mine them - this is a fine example of the law of "diminishing returns". But I could be wrong. Anyway, the point is that unless the Bitcoin community agree to raise the limit - which is the subject of fierce dispute at present - at some point Bitcoin will switch to a transaction-fees-only reward model.
Buterin's concern is that in a transaction-fees-only model, the fees need to be set high enough to discourage malicious operators from taking over the network by making miners offers they can't refuse. How much would it cost to take over the entire network? Buterin has some ideas:
Hence, we have $1.2-4m as an approximate estimate for a “Maginot line attack” against a fee-only network. Cheaper attacks (eg. “renting” hardware) may cost 10-100 times less. If the bitcoin ecosystem increases in size, then this value will of course increase, but then the size of transactions conducted over the network will also increase and so the incentive to attack will also increase. Is this level of security enough in order to secure the blockchain against attacks? It is hard to tell;it is my own opinion that the risk is very high that this is insufficient and so it is dangerous for a blockchain protocol to commit itself to this level of security with no way of increasing it.Translation: in a proof-of-work system, imposing a hard limit on the number of coins in existence is a VERY silly idea. (Bitcoin enthusiasts will no doubt disagree.)
Of course, Ethereum's proof-of-stake (investment) is MUCH better, natch:
In a proof of stake context, security is likely to be substantially higher. To see why, note that the ratio between the computed cost of taking over the bitcoin network, and the annual mining revenue ($932 million at current BTC price levels), is extremely low: the capital costs are only worth about two months of revenue. In a proof of stake context, the cost of deposits should be equal to the infinite future discounted sum of the returns; that is, assuming a risk-adjusted discount rate of, say, 5%, the capital costs are worth 20 years of revenue. Note that if ASIC miners consumed no electricity and lasted forever, the equilibrium in proof of work would be the same (with the exception that proof of work would still be more “wasteful” than proof of stake in an economic sense, and recovery from successful attacks would be harder); however, because electricity and especially hardware depreciation do make up the great bulk of the costs of ASIC mining, the large discrepancy exists. Hence, with proof of stake, we may see an attack cost of $20-100 million for a network the size of Bitcoin; hence it is more likely that the level of security will be enough, but still not certain.Well, OK, a perpetuity pricing model does make sense, though the 5% discount rate is huge for a system that supposedly disperses risk and is free from price inflation. But if the price of security in a transaction-fee system is that far below the price of security in a sensibly discounted return-on-investment model, the transaction fees aren't anywhere near high enough.
This is the uncomfortable truth that Buterin is trying to avoid. The security he thinks his system needs is expensive. Users might not be willing to pay for it.
In a transaction-fee-only system, fees that are high enough to discourage miners from selling out to a higher bidder may be too high to discourage users from transferring to another system. Losing your users is every bit as bad as losing your miners. So having discarded the hard cap on the coin supply, Buterin thinks he has a solution to the problem of expensive transaction fees (or poor security). He wants to identify the optimum mix of new coins and transaction fees in miners' rewards. And he tries to use Ramsey pricing to determine it.
Sadly, Buterin's example misconstrues the context of Ramsey pricing. The differential markup is not because of social harm to purchasers of the products: there is no captive audience. No, the differential markup is to protect the producers of the higher-value good, since the higher price elasticity of demand for their product means that the railway's markup would cause them greater loss of sales than the equivalent markup on a product with lower price elasticity of demand. It's the equivalent of a variable sales tax - indeed Ramsey pricing is sometimes used in tax policy.
Clearly, we do not have a similar situation here. What we have is a single good (coins) which are transported at a fixed fee (transaction fee). The problem is simply one of supply and demand. If the price demanded by miners for supplying security is too high, users will vote with their feet. If the amount that users are prepared to pay for security is too low, miners will sell out to attackers. There should be some equilibrium point at which most users will pay an amount that most miners are happy with: there will remain a tail risk of takeover, just as there will remain a tail risk that the crypto equivalent of Uber entices away the users by massively undercutting the transaction fees. Equilibrium is NOT a risk-free position.
Buterin thinks that it is not possible to set transaction fees high enough to reduce the risk of attack to acceptable levels. I might say that in a decentralised democratic system, if users don't want to pay for security, they shouldn't have to. If the consensus is that a raised risk of attack is an acceptable price to pay for lower transaction fees, that is how it should be. Buterin has no right to start dictating to people what level of attack risk they should be allowed to take.
And he is quite wrong to suggest that mining rewards (new coins) can be used to subsidise transaction fees. Coin rewards for miners are in fact an additional cost for the users. This brings me back to where I started in this article. He misused the term, but he was right about inflation - though he didn't follow it through.
Classically, expanding the money supply is associated with price inflation. In a whole economy, the relationship is very complex. But in this case, it is much less so. BTCUSD has been rising constantly over the last year, which shows us that the rate of expansion is less than the rate of increase in demand for the coins. Nonetheless, the price would have risen more if no new coins were produced. Even though Bitcoin is deflating, the difference between the price (in USD) of 1 BTC when the money supply is expanding and the price (in USD) of 1 BTC when the money supply is fixed is a real cost to the users of the system, since it means their bitcoins don't buy so much. It is equivalent to a higher transaction fee.
So Ramsey pricing just doesn't apply. What we are really dealing with here is human preferences. Do miners prefer to receive transaction fees or new coins? Do users of the system suffer from money illusion - so prefer to have lower transaction fees and a higher rate of price inflation (or lower rate of deflation)? It should be possible to determine some mix of coin rewards and transaction fees that keeps most people happy. And it is entirely conceivable that Bitcoin might choose to operate with transparently higher transaction fees as the price of retaining its hard cap, while Ethereum chooses to run with lower fees and some price inflation. That is for the respective communities to decide.
In fact, what Buterin is trying to solve here is exactly the same conundrum as central banks face (and have never solved). What is the optimum balance of interest rates and inflation? Should we fix interest rates (transaction fee) at some level, and allow inflation (coin supply) to flex? Or should we fix inflation (coin supply) and allow interest rates (transaction fee) to flex? Or some combination of flexible coin supply AND transaction fee?
Whatever approach is adopted, in a simple niche system like Bitcoin or Ethereum the equilibrium would be pretty much the same. The system would have to be far more complex for the choice of target to make much difference. The mix of coin rewards and transaction fees is largely irrelevant: it is the total cost of security that matters, not how it is paid for. In his concern to protect his system from attack, Buterin is chasing a chimera.
Well done for debunking Vitalik on economics!
ReplyDeleteSome minor technicalities:
"The longer the chain, the more energy it takes to validate transactions and the more difficult (and expensive) it becomes to mine additional bitcoins."
This is incorrect. The mining process has 2 steps conceptually:
1) The actual validation process which is computationally cheap (one cheap laptop will do) and essentially constant, because it's incremental. It does not depend on blockchain length.
2) A (decentralised) lottery to choose whose result of the above gets in the blockchain and gets the reward. The proof of work produces tickets for this lottery. Your chance to win is your work / everybody's work. The protocol dynamically adapts to make someone "winning" that lottery take on average 10 minutes, that is if the last few transactions take less because more people are playing it makes it harder, and if they take longer it makes it easier.
Step 2 is not proportional at all to blockchain length, but simply to the number of players (and the speed of their ticket making machines) at any given time. If mining goes out of fashion, it will get easier.
"Personally I am of the opinion that the switch to transaction fees might occur before then"
This makes it sound the "switch" is a binary operation. It is not. The mining reward is already equal to the sum of a set amount of new coins and the transaction fees in the block. So the "switch" is expressed in the relative value of the fixed reward (new issuance) vs. transaction fees. Today the transaction fee is 3-4% of the 25 BTC reward. The theory is as the new coins component diminishes in nominal terms through regular halving, that proportion will progressively increase so that at some point the transaction fees dominate the new coins component of the mining reward.
It could switch back and forth, because the primary motivation for individual users to pay fees is congestion. If there are free slots in blocks, you would be a fool to pay any significant transaction fee (however small any non zero fee is worth collecting for miners, as the marginal cost of including a transaction in a block with spare room is very close to zero, because all the cost of mining is in making lottery tickets, which does not depend on what is inside a block). So if the demand for transactions is lower than the space in blocks, transaction fees should tend towards zero until congestion happens again.
Yeah, I admit accuracy is a bit lacking on the technical side. I wrote this at 2 am and my brain was focusing on the economics not the tech. I will correct the blockchain length thing.
DeleteOn the "diminishing returns" matter - yes, I agree, how close we get to all the bitcoins being mined depends on transaction pricing. Indeed that is Vitalik's point, really.
Hi Frances,
ReplyDeleteGreat post, especially your discussion of Ramsey pricing which I found particularly interesting.
Just wanted to quickly comment on the idea that BTCUSD heading higher shows that BTC is suffering from inflation.
BTCUSD is an exchange rate and shows the relationship of the price of BTC relative to USD, this means that BTCUSD heading higher shows BTC is now more valuable relative to USD (excuse the simplicity of this, but just wanted to ensure that all my definitions are articulated). Inflation/deflation can be seen as a change in the price of a currency relative to real goods such as food, housing, etc.
Let us say that real goods are now given the currency code REAL. Inflation, then, would be a move lower in the BTCREAL exchange rate, showing that Bitcoin now buys fewer real goods than it did previously, similarly if BTCREAL headed higher it would show deflation as Bitcoin can now buy more real goods than it did previously. In order then to assess whether Bitcoin is going through a period of inflation or deflation we need to know what the BTCREAL exchange rate is and how it has changed over time. As you have mentioned BTC isn't really that widely accepted as a store of value that can readily be exchange for real goods, therefore observing this BTCREAL exchange rate is very difficult.
In order to illustrate the difference between the exchange rate and inflation levels it is probably useful now to look at a normal fiat currency and see the difference. Suppose for example that the EURUSD exchange rate falls, as it has been, showing that EUR are now less valuable relative to USD, would this then imply that the EUR is experiencing deflation as would be expected based on your comments regarding the BTCUSD exchange rate? No it would not, we would need to observe the EURREAL exchange rate to see whether the EUR was experiencing inflation or deflation. There will be second-round effects of the EURUSD fall that may affect the EURREAL exchange rate (i.e. USD input prices will increase, causing EUR inflation to increase, i.e. EURREAL will fall), however the fall in EURUSD in and of itself does not imply that EURREAL will fall ceteris paribus.
Again, as BTC is not widely accepted as a medium of exchange it is very difficult to observe the BTCREAL exchange rate, and whilst using BTCUSD may be a rough-and-ready approximation or give some insight, it is certainly not the same.
Best Regards,
Richard
Hi Richard,
Deleteyes indeed, in the absence of a REER some kind of trade-weighted index would be much more informative, but unfortunately we haven't got one of those either. I'm making the best I can of the information available.
This comment has been removed by the author.
DeleteIt took Ms Coppola about 1,000 words to explain that Buterin confused inflation and currency appreciation.
DeleteAt the end of that long-winded explanation, she says, "So if Bitcoin is deflating, not inflating".
Currencies don't really inflate or deflate, prices do. She means, Bitcoin is appreciating, not inflating.
She then gets confused about monetary inflation There is no need to ever use that phrase which introduces misunderstanding. Just say monetary growth.
Sorry, I couldn't bear to read beyond that.
I took a deep breath, read another sentence.
Delete"At the present time, the supply of bitcoins is expanding at the rate of one every 10 minutes."
Wrong. It's 12.5 Bitcoin per block i.e. every 10 minutes after the recent halving.
Corrected.
DeleteI am not confused. Far from it.
DeleteI described Bitcoin as "appreciating" immediately below the chart:
"That is an appreciating currency."
I then went on to explain how that translates through into deflation:
"Each unit of BTC now buys more USD than it did a year ago - rather a lot more, actually. Which means that each unit of BTC buys more goods and services than it did a year ago. Were the prices of goods and services quoted in bitcoins, we would see price falls, not rises. This is deflation, not inflation."
It is Buterin, not me, who has used "inflation" to mean monetary growth.
You wrote:
Delete'So if Bitcoin is deflating, not inflating,'
That is simply meaningless an gets a red line through it in any economics essay. US prices measured in Bitcoin are deflating.
Bitcoin is inflating in monetary terms, i.e. undergoing monetary growth, which is the only correct term when it comes to discussions of currencies.
I think it is abundantly clear what I mean. But if you want to nit pick, don't let me stop you.
DeleteThis comment has been removed by a blog administrator.
ReplyDeleteYou say 'the cost of mining additional bitcoins will rise exponentially as the hard limit is approached, making it prohibitively expensive to mine them - this is a fine example of the law of "diminishing returns".'
ReplyDeleteThat is not right. Miners are free to increase use of all factors of production to produce PoW to obtain coins: mining computers, air-conditioning, electricity, plant, labour. Thereis not a fixed factor of production.
But I get your drift that loosely speaking there is a diminishing return to mining, after each halving of the miners' reward. Butthen again the Bitcoin price has always risen to more than offset this so mining will continue until 21 million are issued.
Law of Dim. MR underlies the positive slope of the Supple curve and is at the very heart of economics. I'm sorry but if you fail properly to understand that I'm not sure why you are opining in this way.
DeleteOf course I understand it. I would not mention it if I did not.
DeleteFor Law of Dim. MR to apply all factors of production must be fixed except for a single variable factor which will eventually be subject to dim MR. With Bitcoin mining which factors are fixed?
DeleteI will answer that for you. In the Long Run none are fixed. Therefore it's not anything to do with economics to say that the Bitcoin miners are subject to Law of Dim MR as successive halvings take place (every 4 years approximately). It's just throwing economic terms around as in a GCSE economics essay.
DeleteThis is so obviously wrong I'm not going to bother to answer. Think about it.
DeleteAre you in fact aware of the distinction between Short Run and Long Run in micro-economics and their definitions? Law of Dim MR is a SR phenomenon. Decrease in Miner Rewards at halvings is a LR phenomenon.
DeleteAnd that decrease in rewards is governed by the Law of Satoshi written in the Bitcoin code.
DeleteMaths isn't your strong point, is it? Just think about what you are saying.
Delete> Well, OK, a perpetuity pricing model does make sense, though the 5% discount rate is huge for a system that supposedly disperses risk and is free from price inflation.
ReplyDeleteOk, do you think the discount rate will be 1%? That would only make the case for proof of stake stronger, and I suppose when the ecosystem becomes mature that may well happen.
On Ramsey pricing, quoting from Wikipedia:
https://en.wikipedia.org/wiki/Ramsey_problem
"The Ramsey problem, or Ramsey–Boiteux pricing, is a policy rule concerning what price a monopolist should set, in order to maximize social welfare, subject to a constraint on profit."
The "social welfare" is then defined by the equation here: https://wikimedia.org/api/rest_v1/media/math/render/svg/4eac2176a0bd1bd028da58e36bb79e9ca0647736. You might notice that it takes into account both producer and consumer welfare - and that is what I am interested in maximizing. I totally get that charging high prices for goods with inelastic demand hurts consumers; however, it also benefits producers, and I'm equally interested in all participants in the ecosystem.
> Clearly, we do not have a similar situation here. What we have is a single good (coins) which are transported at a fixed fee (transaction fee)
I disagree with your characterization of the economic problem. Rather, I see the problem as follows: the blockchain is a railway, which transports two products. The first is securing the holdings of cryptocurrency holders (ie. similar to a vault for gold), and the second is processing transactions. These are two products, and in fact some users consume the first much more than the second, and others consume the second much more than the first. The marginal cost of production for transaction fees is T (as the blockchain loses security and decentralization the more transactions per second it processes), and the *marginal* cost of production for the security service is zero.
Regarding the lack of a captive audience, I would disagree; in any model containing imperfect competition (which certainly describes blockchains), every consumer is captive at least in a marginal sense.
I'll be happy to replace inflation with issuance in the blog post.
Hi Vitalik, good to hear from you!
DeleteA few things:
1) In what way is Bitcoin or Ethereum a monopoly? Users have - and in my view always will have - alternatives. They can vote with their feet. Miners have fewer alternatives in this asset class, of course, but if they are prepared to step outside it they too can vote with their feet.
2) On social welfare. I get your point, but you are missing the fact that price elasticity of demand tends to rise with price if substitution is possible. (You missed this in your example, by the way.)
To give an example, if oranges are low priced - say $5 per kilo - and the price rises 10%, a few people will balk but most people will still buy them. But if oranges are ALREADY expensive - say $10 per kilo - a 10% price rise would make people much more likely to switch to a cheaper fruit. Only if there is no substitute and "nothing" is not a realistic option can we say that there is really a welfare impact. So charging high prices for oranges only has a welfare impact if no other fruit are available.
Interestingly, I remember seeing a queue a quarter-mile long for oranges in Moscow in 1982....I didn't see any other fruit the whole of the time I was there. I have no doubt they were shockingly expensive.
3) On your definition of products: I think we need to have a conversation about stocks and flows. You see transactions as stocks. I do not. My view is that the RECORD of transactions is a stock, but transactions themselves are flows. I'd be interested to know why you think this is not the case.
4) As long as substitution is possible, there can be no captive audience. There can be a reduction in utility as a consequence of substitution, and at the margin some people will accept a higher price rather than substitute an alternative they like less, but that is their choice. They cannot in any sense be regarded as held "captive", except by their own preferences.
Thanks for conceding the point on inflation!
Oh, and I agree with you about the effect of a lower discount rate. I just wanted to make the point that 5% is too high for a (supposedly) risk-free investment when there is zero inflation. It's Piketty's very long-run rate of return, but that is not risk-free.
DeleteThanks for replying!
Delete> In what way is Bitcoin or Ethereum a monopoly?
It's a limited monopoly; I am assuming an imperfect competition model. Particularly, blockchains do have network effects; even though my estimates of these effects are smaller than those of Bitcoin supporters (and IMO if bitcoin supporters were right, bitcoin could have never gotten off the ground in the face of the almighty USD), the fact is that Ethereum proper is the most well-supported network for running many kinds of decentralized applications, and would remain so even with, say, a 2x fee increase. If someone is priced out of using Ethereum and must instead use Expanse (a tiny Ethereum clone with a much weaker mining network and weaker infrastructure support), they do still suffer a welfare loss. In a world where blockchains are more competitive, the margins for pricing above cost would certainly be far lower.
> On your definition of products: I think we need to have a conversation about stocks and flows. You see transactions as stocks. I do not. My view is that the RECORD of transactions is a stock, but transactions themselves are flows. I'd be interested to know why you think this is not the case.
Hmm, I think we might be having a misunderstanding here. First of all, in an Ethereum context particularly, transactions often have nothing to do with moving ether around; they often are just used to interact with some application on the Ethereum blockchain - you can think of them as kind of like highly secured SMS messages, where the cost of production *actually is* a few cents per message because the specific kind of security that blockchains provide requires (with present tech) many thousands of nodes to verify the message. These messages can carry ether in them but do not have to. I view the *secure transmission* of these messages as being a service, and one which can be priced above cost. The second service in my model is that of protecting the integrity of the record of ether balances.
> They cannot in any sense be regarded as held "captive", except by their own preferences.
I agree! Though I generally believe that using words describing necessity ("must", "captive", etc) is a reasonable shorthand for words describing extreme preference, as the behaviors that result from such situations are similar. A sentence like "John had no choice but to work at a bad job at company X" is similarly technically false as John _could_ voluntarily quit and seek alternative employment, but because of the often large gap in utility between John's current situation and the alternative, the language of necessity is often used as shorthand to describe John's realistic choice set, and imo this is perfectly okay in some contexts. Even in your Ramsey railway, the customer _could_ always try to buy products that have been shipped by horse. So this really boils down to an empirical question of just how large the utility gap is between blockchain X and its nearest competitor (note that the transaction fee increases I was mentioning in my blog post were on the order of $0.1, so the utility gap need not be that large to be substantive.
Hi Vitalik,
Delete1) I don't think there is any such thing as perfect competition. And the existence of perfect (natural) monopoly is disputed. So let's rule out both of those.
Ethereum is neither a natural monopoly nor a regulated monopoly such as we find among infrastructure providers. There is no "government" (or state regulator) that can set rules for pricing as, for example, the UK state regulators for telecoms and electricity providers do. Putting it brutally, users do not have to use Ethereum. I understand your concern for welfare, but in a private unregulated system, pricing is determined by supply and demand - and users can vote with their feet.
You seem to want to control prices in your system, because of your concern about security. That's understandable, but it is interference with the free market and the principle of "caveat emptor". And it is far from new. Charging a levy to make deposits safe....now where have we come across that before?
2) Ethereum transactions (messages).
What you describe is pretty much the way ALL payments systems work. A payment is in fact a stock and flow transaction. You seem only to look at the stock and ignore the flow.
For example, we can break down a wire transfer into four elements:
- an instruction from me to my bank to make a payment
- a posting to my bank account reducing the balance by x amount
- an instruction from my bank to the payee's bank to credit the payee with x amount
- a posting to the payee's bank account increasing the balance by x amount.
In the existing financial system the postings are of course double entries, so there are parallel reserve movements for the banks. But as yours is a single entry system, I'll follow your accounting.
How do I instruct my bank to make the payment? There are a number of ways, but it basically boils down to some kind of message, whether a face-to-face conversation with a bank teller, a telephone call, a letter or some kind of electronic communication.
On receipt of the message, my bank issues ANOTHER message - a highly secure electronic communication. Ultimately this goes to the payee's bank (though it would go via a payments network such as CHIPS and be settled through a central bank RTGS such as Fedwire, which adds complexity and time.)
So my "payment" ultimately consists of two messages and two ledger updates. The messages are flows - transient communications to achieve some end. The record of the messages and the associated ledger entries is a stock.
What Bitcoin, Ethereum and the like do is eliminate the intermediary layers in the present system. But they don't really change the way payments work.
All messages can be priced above or below cost. A domestic wire transfer between free-while-in-credit bank accounts is priced below cost. But an international wire transfer using the SWIFT messaging service is (probably) priced above cost. SWIFT is not cheap.
It is interesting that you say some of the messages simply interact with applications on the blockchain. The same is true of existing messaging services. Not all messages move money around.
I agree that your verification procedure carries a cost. But so do verification procedures in existing payments networks. They are a large part of the cost and delay involved in making payments. If all bank accounts were held at the central bank, payments would be instantaneous, wholly secure, and virtually cost-free. It is the need to transmit via intermediaries and verify at each step that adds time, expense and risk.
Interestingly, if all transaction accounts were held at the central bank we would have a regulated monopoly for domestic payments, so Ramsey pricing would apply.
I... actually don't think I disagree with any of that. Is there some specific claim that you disagree with me on?
DeleteI would recommend we taboo the words "imperfect competition", "perfect competition", "Ramsey pricing" and "monopoly". I'm not really interested in "controlling prices"; I'm interested in setting up the economic parameters of a blockchain system (namely, issuance rate, block size, and the mechanism by which fees are distributed) in a way that minimizes risk and deadweight loss, balancing between different risks and losses appropriately where Pareto improvements are not possible. I do not share many bitcoin users' phobia of what they might call "technocratic dial adjustment" and their attraction to sacred limits like 1 MB and 21 million, and in light of this fact I am interested in exploring the wider policy space that becomes available when such issues are not part of your constraint set. I'd like to understand, is there a specific positive claim that I am making that you disagree with? Would be very happy to dig further.
Ah, if we can agree on those taboos then we have a way forward, I think!
DeleteFinding the optimal balance between issuance (seigniorage) and transaction fees (interest rates) is the central banker's Holy Grail. For Ethereum the problem is a little simpler, since you don't have to worry about such complications as unemployment and economic growth. But that doesn't mean it is easy.
In theory, if both issuance and transaction fees are flexible, then market forces should find the optimal balance (stable equilibrium) between the two. That's what I alluded to in my first comment. Users and miners between them will price the risks.
The obvious problem is that when risks are under-priced, the welfare costs can be very high, and they can fall on the innocent. So the question is, is there an overriding case for forcing users and/or miners to pay for security, "for the greater good"? This seems to be what you want, and I can understand why. It's a compositional problem. Individual choices tend to under-price the risks to the system as a whole, because they ignore correlations and fat tails. Decentralization mitigates this, but not entirely.
But I think you've also identified, and tried to solve, a deeper problem. I suspect (though I confess I haven't done the math) that this is a multiple-equilibrium model. There appear to be two stable equilibria - one with fixed issuance and flexible transaction fees, the other with flexible issuance and fixed fees - plus a sliding scale of unstable equilibria in between.
If I am right, then if both issuance and fees are flexible, the system will be unstable. One or the other must be fixed (or managed) to maintain stability. Bitcoin chooses to fix issuance, which is akin to a gold standard. You seem to be heading down the fixed-fees route, which is the equivalent of what central banks do in most fiat currency systems. They set the benchmark interest rate (fee) which governs all other interest rates in the system. They might set it as range limits (the Fed does this), but it's still a managed rate. Does this make sense?
I don't apologize for translating this into existing-system terms. Although the tech is new, the underlying economics is as old as the hills.
I prefer the word dilution. I think it is much more clear and has less spin. It describes what is happening rather than possible complex effects. But, I bet the word dilution is not seen as helping the selling of the policy.
DeleteI actually think, one of the problem of economics, is some of the definitions seem much less clear or accurate than accounting or possibly finance.
"dilution |diˈloō sh ən; dī-|
noun
the action of making a liquid more dilute.
• the action of making something weaker in force, content, or value ...
• a liquid that has been diluted.
• the degree to which a solution has been diluted ...
• a reduction in the value of a shareholding due to the issue of additional shares in a company without an increase in assets.
DERIVATIVES
dilutive |-ˈloōtiv| adjective ( chiefly Finance)."
Information on share dilution:
Deletehttp://www.investopedia.com/terms/d/dilution.asp
“In fact, what Buterin is trying to solve here is exactly the same conundrum as central banks face (and have never solved). What is the optimum balance of interest rates and inflation?”
ReplyDeleteI suggest that conundrum was solved by Milton Friedman and Warren Mosler who argued that the optimum rate of interest is zero, or at least that that should be the target, with occasional deviations perhaps allowable in emergencies. For Mosler, see:
http://www.cfeps.org/pubs/wp-pdf/WP37-MoslerForstater.pdf
My one sentence summary of the argument for a permanent zero rate (and my summary more or less accords with Friedman’s ideas – I’m not sure about Mosler’s) is thus. Government or “the state” should not issue so much money that it then needs to borrow some of back in order to rein the the inflationary effect of having issued too much money.
That of course is not to say that the rate for mortgages etc should be zero: they’ll always be a bit above the central bank rate.
This comment has been removed by a blog administrator.
ReplyDeleteDynamic_Unreality, your comment has been removed because it did not conform to the comments policy on this blog. You can find the comments policy on the About This Blog page.
ReplyDeleteI didn't see any calculations related to increased velocity, number of transactions and people haven't even modelled the security benefits of side chains which can have much lower transaction fees, re-enabling microtransaction based business models while also assisting in making attacks on the Bitcoin network more difficult.
ReplyDeleteInstead Vitalik likes to imagine all cryptocurrency could be re-modeled within the Ethereum structure even though with relatively tiny adoption for economic use Ethereum is a rather expensive way to create smart-contract distributed applications that are compelling to the consumer vs just ordinary computing solutions.
This transaction growth and increased income for miners will tend to favour the biggest foothold and adoption that Bitcoin is likely to see. While Vitalik searches for interesting dynamics he is really chasing a kind of perpetual motion machine - where he gets security without the thermodynamic proof of work (energy investment). Yet there are so many new factors as i indicated above that will change the profitability of mining, strength of security and diminish the ability to profit from an attack on the bitcoin network because rolling back or faking fraudulent transactions will be very limited in scope.
Sidechains have 2 major problems
Delete1. Fail KISS principle
2. Introduce hierarcies.
I think that the cost of securing a blockchain is way lower than the costs of actual security. If Cryptos keep paying the costs of actual security with SorryForYourLosses. Blockchain security is moot.
ReplyDeleteTo the point, even if bitcoin did not halve the block reward and had low fees, monetary inflation and money velocity (assuming fees are friction) would lead to price inflation and that would undermine miner's profitability. No? So fees are unavoidable imo. But fees would always be higher in a decentralized system than a central system yes? So in the end why would anyone use a *blockchain based* crypto to transact? For some imaginary liberty from central banks? I would choose 100 times a central bank over a tech cabal.
So unless cryptos get a clear manifesto, process management, and transparency, Satoshi/Buterin et al may as well change their names to Rothschild.
And an Idea if you want find the equilibrium best to let someone else decide especially someone closer to the problem. Whenever a block is mined a miner can either take the reward or take the fees (not both). So if many transactions are in queue (can we assume high money velocity) they compete to enter the block by increasing fees making miner choose fees over reward (no monetary inflation). Recursively if transaction count low fees would be low and miner would claim reward. Blocks should be always full