Will the current Bitcoin rules become obsolete?

One conclusion drawn by Kroll and his Princeton colleagues Ian Davey and Ed Felten is that those rules will have to be significantly changed if Bitcoin is to last. Their models predict that interest in “mining” for bitcoins, by downloading and running the Bitcoin software, will drop off as the number in circulation grows toward the cap of 21 million set by Nakamoto. This would be a problem because computers running the mining software also maintain the ledger of transactions, known as the blockchain, that records and guarantees bitcoin transactions (see “What Bitcoin Is and Why It Matters”).

Miners earn newly minted bitcoins for adding new sections to the blockchain. But the amount awarded for adding a section is periodically halved so that the total number of bitcoins in circulation never exceeds 21 million (the reward last halved in 2012 and is set to do so again in 2016). Transaction fees paid to miners for helping verify transfers are supposed to make up for that loss of income. But fees are currently negligible, and the Princeton analysis predicts that under the existing rules these fees won’t become significant enough to make mining worth doing in the absence of freshly minted bitcoins.

The only solution Kroll sees is to rewrite the rules of the currency. “It would need some kind of governance structure that agreed to have a kind of tax on transactions or not to limit the number of bitcoins created,” he says. “We expect both mechanisms to come into play.”

That kind of approach is common in established economies, which tame things like insider trading with laws and regulatory agencies and have central banks to shape economies. But many backers of Bitcoin prize the way it currently operates without centralized control, and would likely rebel at any suggestion of changing the rules.

The full article, which is here, has other points of interest.  I would put the problem this way.  It is easy to trust a non-proprietary network, and that is often cited as an advantage of Bitcoin, or for that matter as an advantage of a common language or a standard.  At the same time, the non-proprietary nature of the network makes the rules much harder to change because there is no authority to mandate such a change.  You will find related points in many papers of Joseph Farrell.

Hat tip goes to Andrea Castillo.


I'm a little confused. Doesn't this assume the price (exchange rate) of Bitcoins remains low. If Bitcoins are not mined while at the same time there is sufficient transaction demand, the price will rise and with that the incentive to mine. What am I missing here?

There will be nothing left to mine.

If the transaction fees aren't enough, then your transaction doesn't get processed, then you give a bigger fee, then the miners are willing to process your transaction.

Bitcoin has problems but this one is really taken care of. There's no need for a "tax" on transactions, the market will find what is necessary to keep your transactions processed.

Bingo. Transaction fees are currently hard coded into 0.9, but core developers are aware of the issue and are planning to move towards dynamic transaction fees in the future. http://www.coindesk.com/bitcoin-transaction-fees-slashed-tenfold/

That this is the first "problem" a researcher would identify says a bit about the depth of their familiarity with the technology.

Just so. Nevertheless, my guess is they wouldn't mind a bit being given the task of governing it, or — even better — the task of writing rules for governing it. More jobs for our boys! I forget who said that this single effect explains so much in political phenomena that other kinds of explanations are almost superfluous.

To give the Princeton guys some credit, they're making a more specific claim: based on game-theoretic intuition (but as far as I can tell from the paper, NOT based on a formal model for how transaction fees will evolve as block rewards continue to drop), they "do not expect transaction fees to play a signi cant long-term role in the economics of the Bitcoin system, under the current rules". To elaborate, they argue that competition will keep transaction fees low (a good thing!) but that this will make mining unprofitable, driving down the supply of hashrate (a bad thing!). Combining this with their "death-spiral" theory -- less hashrate leads to cheaper 51% attack leads to drop in confidence in bitcoin leads to lower prices leads to even less hashrate -- they view this as an unavoidable bad outcome.

To critique this, I'd observe that:
(a) there is no formal model for tx fee prices in their paper, so the basic argument really just boils down to *faith* that tx fees will remain relatively low; however, I think we'd all hope that competition *does* keep tx fees low (else a major benefit of bitcoin goes out the window), so let's stipulate this point.
(b) as in most industries, declining profit margins will spur mining companies to innovate, and successful innovators will indeed have a large incentive to keep mining. so the long-term decline in hashrate supply actually strikes me as quite unlikely. it is very rare for the supply of *any* good to decrease in the very long run, unless it is made obsolete by a superior product (e.g. whale oil -> petroleum). the latter is quite possible, since the superior product could be "litecoin hashrate". \
(c) even assuming an unrealistic fixed set of people who are potential miners using a fixed technology basis, I don't think the paper addresses the key question. Their basic equation for the mining game is (total global cost of mining per sec) = (total mining reward in dollars per sec). Let's stipulate that bitcoin does not replace the dollar as the world reserve currency, and accept that this equation is roughly accurate. But note that this is not an equilibrium that can be "solved for", except perhaps in the very short term. As the paper emphasizes, robust mining network is crucial to the security of the blockchain against 51% attack. The righthand side factors as tx fee * # tx / sec; holding tx fee constant and low, we see that hashrate will go up as demand for bitcoin transactions goes up (an obvious conclusion). But demand for transactions will itself depend on hashrate, as transactions are more desirable on a more secure network. Conclusion? The "equilibrium" after block rewards fall to zero is still a complicated dynamical system, which the paper in question makes no real advance in analyzing.

Link to the paper for those interested: https://www.cs.princeton.edu/~kroll/papers/weis13-bitcoin.pdf

That's not much of a credit, I think :) By the way, the mining premium, currently 25BTC, is set to decline in steps, giving time for a market estimation of transaction costs to develop in a gradual manner.

The fact that core developers are aware of the issue is not the same as saying there is no issue. Perhaps they will devise a dynamic fee system that will solve the problem, perhaps they won't. That's the issue that the paper is raising. As they point out, in the current system, there is no mechanism to enforce a level of transaction fees that are consistent with profitable mining once the rate of Bitcoin growth levels out. It's a classic incentives problem. For the individual miner, once you've mined a block, your incentives are to include any transactions with non-trivial fees, since that is basically free money to you at that point. In other words, how do you require miners to enforce a minimum fee that benefits the group at their own expense? What you need is an OPEC for miners, but how to enforce it?

Note that tx fees always have been (and AFAIK always will be) user-specified. The "dynamic fee system" is merely a proposal to have the *default* tx fee "float" with the exchange rate and perhaps the hashrate. It's a nudge, not a rule change.

And I don't think the paper "points out" what you say it does; it merely asserts it. And I would say that assertion is quite possibly false -- there is such a mechanism, and it is known as the law of supply and demand.

The supply & demand solution may not be a sustainable equilibrium. It could potentially lead to a downward spiral in supply of mining power and demand for Bitcoin.

All I'm observing is that the paper gives neither theoretical nor empirical evidence for that potentiality, whereas I think one's prior should be that it's unlikely, since there are not so many markets in which that has happened. (Even markets with network effects.) Replace "mining power" (resp. "bitcoin [transactions]") with "airports" (resp. "air travel"), or with "telephone service" (resp. "telephone calls"), or -- to give an example that has always been largely devoid of regulation -- "internet service" (resp. "internet service"). Does the argument seem sensible in these cases?

Right the assertion about transaction fees being too low in the long run is not backed by any model or analysis in the paper. Their argument could make sense only if the number of transactions per block is unlimited which is not at all the case. In reality the fees naturally serve as priority so if transaction volume grows, there could be other problems but fees being permanently too low to incentivize mining will not be one of them.

I think an OPEC for miners is self-enforcing and thus stable. It can't be undercut by a small defection, only by a large defection. See my rather lengthy (sorry) essay further below.

One of the problems is actually that the BTC community is not decentralizing hashpower over enough different pools, resulting in the bulk of the power being controlled by about 12 guys (ain't that how many banks are in the Fed?), a lot of them jokers. One of them (luke-jr of Eligius) abused his users' computing power to run a 51% attack on an early altcoin (he seems to not like the altcoin phenom.)

Or Bitcoin gets replaced by something better - anyone still using Hercules graphic cards? PS/2 connections for their mouse or keyboard? 5 1/4 floppies?

And yet, we can still conduct such 'transactions' as viewing graphic and text files.

Bitcoin is an implementation of an idea - as were Hercules graphic cards, PS/2 connectors, and 5 1/4 floppies.

Wrong again.

Looked at from a micro point of view, the transaction fee should roughly equal the risk of not being included in a timely manner + risk of double-spends. Right now, double-spends in Bitcoin are extremely rare - aside from zeroconf exploits (which don't count), you can probably number successful double-spends on a single hand. But this hardly seems optimal! Just as the optimal number of falling houses in an earthquake is not zero, the optimal number of double-spends is not zero. There's a huge overinvestment in mining because of the hardwired block reward.

Care to elaborate on your micro analysis of the tx fee?

'The only solution Kroll sees is to rewrite the rules of the currency. “It would need some kind of governance structure that agreed to have a kind of tax on transactions or not to limit the number of bitcoins created,” he says. “We expect both mechanisms to come into play.”'

Or use another currency without that flaw.

Sort of like how we no longer use 16 bit processors to read ASCII text. And the fact that ASCII text is till readable on more devices than were imagined when it was first standardized in 1963 has little to do with a 'governance structure.' (Standards body, yes - whether that is a governance structure is the sort of discussion that is amusing to think of this web site ever having.)

IPv4 is a lot older than BitCoin. Hell so is IPv6.

Perhaps I am being unfair on you. You make a good point about ASCII.

But then it also just goes to show how long we can stick with a flawed standard. And UTF-8 shows us that even a flawed standard can evolve into an excellent one. As for BitCoin, we don't even understand its flaws yet.

ASCII was a limited standard, not a flawed one. Which is why ASCII works perfectly with all iterations of Unicode.

Of course, one can argue that ASCII's limits were flaws, but that is a bit unfair for a standard which is more than 50 years old.

I am saying that we kept using it long after its flaws were obvious; and pre-Unicode attempts to replace it made the problem worse in many respects. Unicode + UTF-8 are a good solution precisely because they leave ASCII sacrosanct.

It think it is plausible that that BitCoin becomes the ASCII of cryptographic finance. Even when we figure out what its flaws are (which we haven't yet) we might will hang onto it and stay compatible.

The fact that rules are harder to change in a non-proprietary network, is that a bug or a feature?

Tyler sure makes it sound like a bug but I'm not convinced.

When monetary rules become hard to change, many smart people find themselves out of work. That's certainly a bug if your metier is monetary economics.

I think the point is that it's a feature with some clear trade-offs.

Of course, if the trade-offs are sufficiently undesirable, then the value of bitcoin will fall as people use other means to transact business. Which probably makes it a riskier store of value.

I find it interesting that a lot of the publicity around Bitcoin was that it would be a good and solid store of value against government inflation. But I think it's not yet been shown as a good and dependable store of value, given its newness and unclear regulatory hurdles, so the pricing is highly contingent and volatile. However, Bitcoin is incredibly important as a means to transfer wealth without interference.

So many bad papers being published by Acadamia on Bitcoin. It is not really that hard to understand. Tyler, perhaps you should go to a Bitcoin conference and get a better understanding from those who are actually working in the field. There is one in DC in April.

I get the impression that actually understanding subject matter is beneath an economist's dignity. Remember those dummies who refused to look in Galileo's telescope to see the moons of Jupiter, on principle.

How substantial are the costs of verifying transfers? Won't the costs of verification fall given the pace of technological innovation? I don't have a good sense of the computational demands that will be needed for the next decade of Bitcoin transactions, but if the costs of computing power (speaking broadly) continue to fall, then couldn't the load be handled by organizations willing to do it in exchange for the limited compensation?

Bitcoin difficulty is automatically adjusted to increase the costs of verification with technological innovation. Look up a historical Bitcoin network hashing power graph, hashing power has increased many billion times over the years and blocks have been coming at the same 5-10 minutes interval.

But how many transactions are validated in one block and what is there value? Is there an upper limit on this? What economic forces govern these numbers. I tried to figure it out once and didn't come up with anything conclusive.

The miner chooses what transactions they want to include in a block they create. The upper limit is 1 MB of data which at an average 0.5 KB/transaction allows a maximum of roughly 2000 transactions to be confirmed every 10 minutes. The block size limit is one of the easiest parameters to convince the community to change should transaction rates climb high enough to justify it.

Miners themselves will make their own estimates of what the minimum fee/KB of transaction is that makes it worth including in the block. The economic force is that larger blocks take longer to distribute across the network and give a competing miner a small chance of getting the 25 BTC seigniorage for themselves. Thus the rational miner ignores zero and very low fee transactions.

Ok, so there is some cost for larger blocks. But what about the cost of guessing the hash? Are they hashing the whole block for each guess or do they just re-hash a fixed size hash of the block? More generally how much of the cost to miners is per-block and how much is per-megabyte?

Can't reply to your more nested comment, so I'll do it here.

"Are they hashing the whole block for each guess or do they just re-hash a fixed size hash of the block?"

The latter. The block is hashed once to a fixed size, and this + a random guess is then hashed in the hopes it comes out to an answer matching parameters defined by the current difficulty.

Economists should stop trying to dictate what they think will happen with Bitcoin and start watching with open mind. It's a live experiment, and people who were totally unable to predict major events in their own field of expertise need to develop a little humility about this sort of thing.

I agree. Descriptive, not prescriptive. The Krugman types shouting ITS NOT REAL PLEASE STOP IMMEDIATELY are hilarious.

Is there a link to the actual paper somewhere, without knowing the specifics of their claim, it's hard to evaluate whether it's a real concern or not.


Once the mining reward drops to near-zero - which could happen eventually when the bitcoin production rate becomes very small, or could happen tomorrow if the dollar price of bitcoin fell to almost nothing - then the incentive to process transactions will be equal to the total amount of transaction fees.

The amount of computing power dedicated to processing transactions will therefore be equal to the amount that those transaction fees can purchase, minus a slim profit margin. Those fees will pay for hardware (capital costs) and electricity (marginal costs). If transactions are few and cheap, then the network will be small and cheap. If transactions are plentiful and/or expensive, then the network will be large and expensive.

The security of the network depends on it being expensive. An attacker can subvert the network if they can supply more than half of its computing power. If they have less than half but still have a substantial fraction they can still subvert it, just less reliably. But to get that much power they'll have to pay for hardware and electricity. They'll need to invest a substantial fraction of the total amount that all miners have invested, and that amount is (handwave handwave) equal to the NPV of the entire mining revenue stream.

Right now the mining revenue stream is roughly $2,000,000/day just from bitcoin production, not including any transaction fees. There's less than 100,000 transactions per day. If the mining reward went to zero, the transaction fees would have to rise to $20/transaction to sustain the same amount of investment in hardware as we have today (and thus keep the same level of security in the bitcoin network).

To get it down to pennies per transaction, we'd need to see a hundred-fold (or thousand-fold) increase in the number of transactions. This seems unlikely.

Suppose bitcoin use really takes off, and we get up to 1,000,000 transactions per day (more than ten times the current use). Let's say the average transaction fee is ten cents - a lot more than people pay now, but a lot less than people pay for other money transfer services. The network would be worth $100,000 per day. That might pay for a few tens of millions of dollars of equipment. Subverting that network would be beyond the reach of most individuals or groups. It would be expensive but still quite doable for major corporations, and trivial for governments.

What's interesting is that the security of the network is a commons. Individual users have an incentive to pay a large enough fee to get their transactions completed quickly, but no incentive to pay a large enough fee to ensure the network is sufficiently secure. Even if a user were willing to pay a premium to use a secure network, they can't do so - the network will only be secure if enough people are doing so, and one person's decision to do so won't make it more or less secure than it already is at the time.

The pricing of "fast enough confirmations" has an interesting dynamic, too. If almost all miners declared that they would process any transactions with a fee of ten cents or more and ignore all transactions with a smaller fee, that would effectively lock in the transaction price for a very long time. Even if a few miners announced higher or lower prices, users would get almost no benefit from paying a higher fee and would get almost no service whatsoever from paying a lower fee. Unlike a normal market, where whoever announces a lower price ends up taking all the business from the high-priced competition, you have essentially no control over who processes your transaction. It will get processed by whoever happens to win the race, which depends on the computing power they have. If the winner decides to include your transaction, then great. If not, too bad. So if only a few miners accept low fees, then anyone paying a low fee will have to wait a very long time for their transaction to go through. Hours, days, or even years.

This means that once the mining reward goes to near-zero, transaction fees will be whatever the super-majority consensus of miners say they will be, and they'll stay that way until a substantial fraction of them change fees in concert. That decision will have a profound effect on both the adoption of bitcoin and the security of the bitcoin network. Higher fees mean lower adoption, lower fees mean less security. It will be interesting to see if the collective wisdom of this particular crowd manages to strike the right balance.

Note that the phase-out of block rewards is extremely gradual. (https://en.bitcoin.it/wiki/Controlled_supply)

Since we cannot anticipate the exchange rate trajectory, I think it's a bit ludicrous to speculate as to how this will play out. We just don't have enough data. Check back Jan 1, 2021, and then perhaps form an opinion based upon the 2016 and 2020 halvings and the concurrent exchange rate and hashrate dynamics.

My analysis is independent of the exchange rate and hashrate.

I'll have to find time to read the paper... but "what will happen when the block subsidy is replaced by transaction fees" has been discussed quite a bit. See https://en.bitcoin.it/wiki/Funding_network_security for a summary and links to past discussions.

One link I like:

This is no different from any other "profitability"/"marginal provider" problem.

Mining not profitable enough -> marginal miners drop out -> profitability restored

Of course, it's more complex than that because miners dropping out affects the difficulty factor, which introduces a whole new dynamic, but same idea.

But if miners drop out in means you have a better chance of winning the tournament and receive any remaining reward + TX fees. That is, the feedback works partly through the TX price and partly through the feedback loop legislated by the protocol. As far as I can see on displaces the other, and the "dynamic" is more or less unaffected.

I think the thing about Bitcoin that disturbs most academics (economists specifically) is that fact that someone? just came up with an idea and just did it. While the real world tries new ideas, economists can't even agree on the nature of money let alone create a new medium of exchange. Its embarrassing how ineffectual they are.

Based on the reaction in reddit.com/r/economics , the hate for it is basically, "No! Wrong! This can't work, because it's deflationary! It doesn't solve the currency area problem! It doesn't allow for intelligent monetary policy to manage demand! It has to fail. STOP HAVING FUN!"

The only solution Kroll sees is to rewrite the rules of the currency. “It would need some kind of governance structure that agreed to have a kind of tax on transactions or not to limit the number of bitcoins created,” he says. “We expect both mechanisms to come into play.”

I may be misunderstanding, but I believe the 21 million coin cap is a function of the math that drives Bitcoin. Agreeing to not to limit the number of bitcoins created is akin to "agreeing" that henceforth 2 + 2 = 5.

Actually, the cap was a chosen design parameter, essentially part of the "constitution" of Bitcoin. Altcoins forked from the Bitcoin codebase chose different parameters, and thus have different maximum (or unbounded) amounts of generated coins.

It could be changed, but any node(s) deviating from the parameters would be obvious (by referring to bits of divergent history that doesn't match to founding parameters) and have to form their own network, separate from the 'orthodox' network.

Personally I expect several such schisms in the future, when the network faces hard challenges about which significant self-sustaining constituencies cannot reach consensus. The neat thing, though, is that the not just the code but the blockchain (shared history and account balances) can be forked too. It'd be like a corporate spin-out: everyone with 1 share in the common parent could continue to have 1 share in each of the child systems, separately spendable, with free-floating relative values.

One potential trigger for a big schism would be a credible alternate mechanism for achieving similar or better double-spend protection, without the same work-burning function. Miners with sunk costs would hate this, and prefer the classic approach, while bitcoin transactors paying escalating transaction fees to cover mining power bills would love it, if it really worked. So as fees rise, you might see one or more attempts to fork the entire existing endowment into a cheaper system... and the failure of any one (with the 'safe old mining' prevailing) wouldn't necessarily mean the next couldn't succeed.

It's not mathematically enforced, it's enforced by network majority. Changing it just means rolling out a new version of the bitcoin application and getting a majority of nodes to run it. Several altcoins have made changes both to the maximum number of coins and other parameters like transactions fees.

Dogecoin, for example, decided there would be no maximum number of coins. Litecoin changed the algorithm for determining minimum transaction fees. Bitcoin's had several point releases containing minor revisions over the years.

Yes, but this is kind of my point. Those other currencies were able to change the parameters, but only by starting a new currency. The cap isn't fundamental to digital currency, but it is fundamental to Bitcoin. Once Kroll's idea to change the rules are followed, you are no longer producing Bitcoins.

I'm still puzzled by how in the long run it can be cost effective to have a transaction verification mechanism that depends on multiple people competing to verify the same transaction using an algorithm designed to 'make work' for proof of work? In the long run, it would become more economical for various centralized authorities (e.g. banks or bank like entities) to maintain the records and validate transactions using the technology that balances cost and security.

On average, it's far from cost effective. Very few miners turn a profit unless they're among the lucky few that got a new class of mining hardware before everybody else.

Centralizing the ledger, however, would completely defeat the purpose of a decentralized currency, and it is likely that there will continue to be miners operating at a loss for a long time. Much like there will always be drug cartels willing to pay a fee (often to legitimate banks) to launder money.

In other words, bitcoin is the latest Ponzi scheme … but a Ponzi scheme nonetheless

Comments for this post are closed