A Happy New Year for Bitcoin?

The future of Bitcoin and other cryptocurrencies does not depend on whether digital currency is a good idea—digital currency or some equivalent is clearly a good idea. What will matter is whether cryptocurrencies are as good as other approaches to digital currency. From this perspective, cryptocurrencies look a bit ridiculous.

A digital currency is nothing more than a book entry asset with a convenient and widely adopted electronic payments network attached. Perfectly good systems for book entry assets and electronic payments networks have existed for a long time: Bitcoin made no real advances on either front. Where Bitcoin made a brilliant leap is in its governance model.

Bitcoin’s foundation is a governance model so radically decentralized as to be amorphous. This model warrants all the hyperbole in a limited sense: it is nearly miraculous that such ill-defined governance can function at all. Imagine that someone figured out how to harness ants to pull a fully loaded freight train. This would be a miraculous feat, but after marveling at the achievement, one would surely conclude that there are still better ways to pull a train. So too with Bitcoin.

Bitcoin transactions are inefficient, slow, and uncertain relative to natural alternative approaches to digital currency. And these disadvantages are probably not fixable—they are essentially required to support the amorphous governance model. Worse still, amorphous governance probably isn’t a good thing; it is almost certainly not a good enough thing to warrant using such an inefficient, slow, and uncertain payments vehicle.

The fact that Bitcoin must accept gross inefficiency in service of formless governance is not a discovery of mine. The developers emphasize that inefficiency is integral to cryptocurrencies:

If you are OK with 10 or so individuals [or entities] controlling the currency, then you can design a much better system than Bitcoin. [Bitcoin wiki]

Here is the picture: If you are OK with centralized governance, you can do much better than cryptocurrency. In fact, if you can accept governance decentralized across, say, 10 or 20 entities, you can still do much better. If, instead, you are desperate for amorphous governance, you will pay a large efficiency price.

Amorphous governance

Few people I’ve spoken with understand how radically formless Bitcoin governance is. With any book-entry asset, a key issue is how to maintain integrity of the ledger in which ownership, and ownership changes, are recorded. In the case of Bitcoin, there is no official ledger, and, in fact, nobody has any responsibility to keep any ledger at all. Should they wish to do so, anyone can, at any time, start or stop keeping a ledger, and anyone is free to put forward any version of the ledger they like.

Then how do we know who legally owns what? That’s old-fashioned thinking.

What stands in for a conventional notion of ownership arises from the following process: take all the things that are being put forward as the ledger at the current moment and apply a set of consensus rules to determine if a consensus exists. If so, we take this as truth, but only as truth of the moment. This snapshot of reality prevails unless and until some other consensus arises. No rule or principle keeps a very different consensus from emerging at any point in the future. In one consensus, you bought some Bitcoin this morning and sold them this afternoon, but in a subsequent consensus, perhaps you never owned them at all.

(Warning: there is a lot of nonsense on the web about irreversible and instant transactions in Bitcoin. A careful reading makes clear that no transaction is ever final. This will become a clearer below, where we dig deeper into Bitcoin.)

Even without getting into any further details, it should be clear that the rules for determining consensus are critically important to this system. Nothing, however, about the choice of rules is clear. Nobody controls the consensus rules, and anyone can propose new consensus rules at any time. Finally, nobody can enforce the application of any set of rules. The consensus rules prevailing at any time are simply the rules being used by a sufficiently large share[1] of those who happen to be choosing to apply consensus rules.

This is an accurate description of Bitcoin governance, but it is intentionally written to highlight the profoundly formless nature of the governance. Let me give a hint as to the possible magic—that is, a hint as to why the ants might work together to pull the train.

Suppose that the consensus rules create an incentive for a lot of folks to keep what we might call an honest ledger.[2] So long as the incentive of crooks to abuse the system are vastly overwhelmed by the incentives of honest players to keep honest ledgers, then honesty will more or less spontaneously emerge from this amorphous system. So this might work if we get the incentives right between the forces of good and evil.

There are two problems with this approach. First, an essential part of getting the incentives right involves imposing gratuitous inefficiency in the system. Second, even ignoring the inefficiency, nobody has put forth a comprehensive argument that Bitcoin or other cryptocurrencies get the incentives right so that honesty reliably prevails.

Gratuitous inefficiency, the heart of blockchain

Anybody can propose anything as the ledger. Moreover, ledgers are pretty cheaply copied, and anybody can electronically propose as many electronic ledgers as they like. So we could easily find the system swamped by bogus ledgers.

It is difficult to imagine any solution to this problem except the one Satoshi Nakamoto engineered into the blockchain ledger: ledgers must be costly to create. That is, the consensus rules will have to reject out of hand any ledger that doesn’t embed or reflect a very high real cost of production.[3] It should be obvious how this helps: if you make it sufficiently costly to propose any ledger, fraudulent or not, you’ll sharply reduce how many are proposed.

Cryptocurrency ledgers are based on the blockchain idea, and make ledger creation costly in the following way. New transactions do not go straight into anyone’s copy of the ledger. A few proposed transactions pile up and are grouped into a block of transactions, and then the block will be accepted under the consensus rules if, i) all the transactions appear legitimate[4] and ii) the block has attached to it incontrovertible proof that sufficient resources were wasted on behalf of this block.

This wasting of resources is the Bitcoin mining you’ve probably heard about, and the essence of wasteful mining is captured by the following. The block of transactions is formed and then will not be accepted in the ledger[5] until incontrovertible proof that somebody has flipped a coin until they flip 100 heads in a row is attached to the block. In practice, this coin flipping is done by rooms of computers, not rooms of people, and cryptographic math provides the essentially incontrovertible proof that the coins were, in fact, flipped.

Let me repeat for emphasis: the key to honesty prevailing is dissuading folks from creating bogus ledgers by making ledgers costly to create.

The system still must incentivize folks to do coin flipping, and to do so the system offers a reward to the entity that first attaches the incontrovertible proof of the flipping 100 heads in a row to a given block of trades. If the reward is big enough, lots of entities will fill rooms with computers flipping coins, hoping to be the first group to get 100 heads in a row and win the reward.

So how inefficient is all this competitive coin flipping? Nobody knows exactly, but everyone agrees that the waste is immense. For example, a prominent estimate is that when Bitcoin was supporting about 200,000 transactions per day, the waste was equivalent to the energy consumption of all of Bulgaria. Others pooh-pooh this estimate, arguing that it might only be equal to one-quarter of the energy consumption of Bulgaria. Oh, that’s better.

Remember that this waste supports only 200,000 transactions a day, a tiny fraction of the millions handled by conventional payments networks daily. And in conventional networks, the marginal cost of an additional transaction is near zero, but with Bitcoin, the marginal wasted energy for additional transaction must remain high. More transactions must mean proportionally more waste if the system is to deter the creation of fake transactions.

I’m not asking you to trust me on all this: I’m only trying to spell out what the Bitcoin documentation means when it says that, unless you want radically decentralized governance, there are far more efficient systems.

Deeper dive: details and misconceptions

We noted above that some folks claim that Bitcoin transactions are irreversible, but, in fact, no transaction is ever final in this system.[6] Starting from the brief description of the block chain above, this can now be explained a bit more clearly. Focus on a particular transaction. Suppose that the transaction goes into a block that becomes part of the blockchain. Subsequent blocks of transactions are metaphorically stacked upon the one in question, with each block linked to its neighbors in a sort of chain. To undo the transaction in question, someone would have to waste enough resources to redo the block and all subsequent blocks resting on it. Thus, transactions are never final, but the resource cost of undoing a transaction grows as more blocks are piled upon it. The system is designed so that a new block is created about every 10 minutes, and it is becoming a convention that a transaction should be considered pretty secure after about six additional blocks, which should take about an hour.

Regarding the waste of resources, there are a lot of completely confused discussions of how mining could be made more efficient. The whole point is that it must be costly to create a block, so if mining becomes more efficient, the problem will be made more difficult (e.g., flip 200 heads in a row).

There are deep discussions underway about how to maintain integrity with less waste, but almost all of those involve abandoning substantial aspects of the amorphous governance that gave rise to the waste in the first place.

What governance benefits do we buy with inefficiency?

You don’t have to be at a Trump rally to find people who see a few flaws in conventional government. Thus, the idea of currency without government has some appeal. Let’s accept that you can have currency without government; you still cannot have currency without governance. So is amorphous governance better than what we have? Enough better to warrant the cost?

This is the ridiculous part of the Bitcoin model. Defenders essentially claim to have created the holy grail of political economy, a system in which folks with diverse interests reliably reach consensus on a just and fair outcome with no central authority. The key, say the defenders, is game theory:

Possession of Bitcoin is not enforced by business rules and policy, but cryptography and game theory. [Bitcoin Wiki]

Bitcoin discussions are full of statements like this, and read like bad freshman essays. Game theory could help us establish that the decentralized consensus process will reliably pick an honest outcome. The trouble is that nobody has actually done the relevant game theory. Indeed, completing such a theory would be a game theory advance akin to what general relativity was in physics.

Satoshi Nakamoto essentially put forward a conjecture that this would work, but nobody has even attempted to comprehensively analyze the incentives of the multitude of parties involved in order to establish conditions under which honest ledgers will very reliably float to the surface in the consensus process.[7] Because this is a fast moving area, I won’t claim to be on top of it, but there are a couple of deep flaws I see in current analysis.

Most work in this area, including Satoshi Nakamoto’s seminal paper on crytpocurrency, presumes that there is some unambiguous notion of an ‘honest ledger’ and that the goal is to get the consensus spontaneously to pick that ledger. But in this system there is, by definition, no centrally agreed on notion of an honest ledger. While the literature analyzes attacks on the system, there is no distinction between an illegitimate attack vs. a particularly creative effort at consensus building. The consensus, as they say, is what it is.

This is a problem at many levels. An immense share of disagreements in the economy arise not because one party does something everyone (even the offender) agrees is illegitimate, but because various parties disagree on what should be accepted as legitimate. There is no process for orderly resolution of such disputes in Bitcoin. This problem is precisely what is behind the debates that have arisen over hard forks, which have roiled the valuation of Bitcoin and other cryptocurrencies.

But the problem is deeper. Since there is no standard of illegitimate consensus building that we could simply make illegal, the only way to forestall these unsavory types of consensus building is to be sure that the resource cost doing them exceeds the expected benefits. Analysis of these tradeoffs in the cryptocurrency literature, however, seems to be based on the naïve view that the potential benefits from shenanigans are bounded by the value of the Bitcoins that might be stolen.

In a world of derivative securities, the value of the coin stock has little to do with the possible benefit to, say, temporarily deflecting the value of the Bitcoin. This is basically why Warren Buffett called derivatives financial weapons of mass destruction. The value that changes hands due to a deflection in the value of Bitcoin is bounded only by the size of the bets that have been placed, and has no natural bound. (If we didn’t know this before the financial crisis, we learned it from the credit default swap market during the crisis: in many instances the value at stake in the default of a bond was far in excess of the value of the bonds in question due to bets placed using CDS.)

Of course, laws about illegitimate market manipulation might, in principle, come into play to keep this from happening. Put in a more pointed manner, Bitcoin’s amorphous governance might be bailed out by conventional centralized government. But as we found in the financial crisis, this is a pretty hazy area. And the amorphous Bitcoin environment explicitly rejects existence of any centrally accepted distinction between illegitimate manipulation vs. creative consensus building.

More ominously still, if Bitcoin became a centerpiece of the payments system, there would be immense externalities arising from the continued health of the system—Bitcoin would become too big to fail. The political benefits to an evil group or state actor from disrupting the system—perhaps bringing down the economy—might far exceed the value of the coin stock.

In short, everyone agrees that the integrity of Bitcoin rests in a game-theoretic calculation of costs and benefits of shenanigans. But nobody has even figured out the relevant calculation, let alone performed it.

The future of Bitcoin?

On this blog, we’ve made a practice of limiting predictions to stuff we’re pretty confident about; thus, we don’t make a lot of predictions. But here are some pretty surefire predictions about Bitcoin.

Part of the excitement about Bitcoin seems to come from the realization that current prices for electronic transfers are too high and that much cheaper electronic payments are technologically feasible. There is an amazing range of stuff that the internet has made free or nearly so. It is a very safe bet that if mainstream payments systems don’t provide nearly free transactions, some new entity will.

Cryptocurrencies as currently understood must have a far higher marginal cost per transaction than conventional payments networks. Thus, unless folks strongly prefer unproven, slow, and inefficient transactions, other approaches will win out in the market.

At this time, nobody knows how to make a currency system that is reasonably efficient and secure without heavy reliance on some centralized notion of authority. Absent fundamental change—change that pretty much guts the amorphous governance model—cryptocurrencies will not become a central part of the payments system. Put more positively, if some cryptocurrency flourishes it will flourish within a fairly conventional governance structure.

So why the hype? I took a shot at that question in a previous post on crypto-collectibles. The bottom line of that post is a prediction that Bitcoin could continue to have value as a crypto collectible, and—like pet rocks—Bitcoin might even outperform conventional collectibles such as coal, oil, and nickel.


1. Given that ledger keepers can pop into and out of existence at any time, majority is not really a well-defined notion here, so I’ll just stick with the ambiguous term sufficiently large[back]

2. See the discussion below on the problems with the idea of an honest ledger. Much of the literature on Bitcoin presumes that honest ledger is a well-defined notion, but in this decentralized system it is not defined at all. From the standpoint of the post at this point, you should just imagine a case in which some everyday notion of honest has a clear meaning. [back]

3. From the Wiki:

Mining is intentionally designed to be resource-intensive and difficult so that the number of blocks found each day by miners remains steady. Individual blocks must contain a proof of work to be considered valid.… The primary purpose of mining is to allow Bitcoin nodes to reach a secure, tamper-resistant consensus. Bitcoin Wiki[back]

4. for example, the spender currently has claim to the Bitcoins. [back]

5. that is, will not be accepted under the consensus rules. [back]

6. For example, the Bitcoin Wiki tries to have it both ways, beginning a discussion of the topic with “transactions on the blockchain are irreversible” but subsequently stating it correctly: “transactions become more irreversible as the number of confirmations rises.” More irreversible!?! An expression reflective of the sloppy thinking in this area. [back]

7. Instead, folks have provided analyses of a handful of particular ways the system might fail. In short, folks have painstakingly analyzed a few ways the system might go awry, but nobody has claimed any comprehensive analysis of the possible failure modes of the system. [back]