In a radical market

Author: Vitalik Buterin

Original Title: 《On Radical Markets》

Publication Date: April 20, 2018

Recently, I had the privilege of receiving an advance copy of Eric Posner and Glen Weyl's new book, Radical Markets, best described as an interesting new way to look at the topic sometimes referred to as "political economy" - addressing the significant issues of how markets intersect with politics and society. The overall philosophy of the book, as I explain, can be expressed as follows: markets are great, and the price mechanism is an excellent way to guide resource use by integrating the goals and information of many participants into a coherent whole. However, markets are socially constructed, as they rely on socially constructed property rights, and both markets and property rights can be constructed in many different ways, some of which are unexplored and may be much better than what we have today. Anti-dogmatic liberals view freedom as a high-dimensional design space.

There are many reasons why this book interests me. First, while I spend most of my time in the blockchain/crypto space working on the Ethereum project and providing various support for projects in that field, I also have broader interests, where the use of economics and mechanisms aims to establish more open, free, equal, and efficient systems for human cooperation, including improving or replacing today's corporations and governments, is an important endeavor. The intersection of interests between the Ethereum community and the work of Posner and Weyl is multifaceted and rich; Radical Markets devotes an entire chapter to the concept of "personal data markets," redefining our economic relationship with services like Facebook, well, look at what the Ethereum community is doing: individuals, data, markets.

Second, blockchain is likely to be used as the technological backbone for some of the solutions described in this book, and Ethereum-style smart contracts are well-suited for the various complex property systems explored in the book. Third, the economic ideas and challenges proposed in the book are also ideas that the blockchain community has already explored in depth for its own purposes and will continue to explore. Posner and Weyl's ideas often have the characteristic that they allow economic incentives to adjust in place of subjective, ad-hoc bureaucracies (for example, the Harberger tax can essentially replace the right of eminent domain), and given that blockchains cannot access trusted human-controlled courts, these solutions may prove to be more ideal for blockchain-based markets than for "real life."

I would caution that readers cannot be guaranteed to accept the proposals in this book. At least the first three have sparked significant controversy, and they do indeed violate many people's moral preconceptions about how property should operate and what money and markets can and cannot do. The authors are no strangers to controversy. Posner has even shown a willingness to oppose concepts like human rights law on previous occasions. That said, the book does spend a considerable amount of time explaining why each proposal could improve efficiency and provides multiple versions of each proposal, hoping that at least one (even if partially) of each idea can be realized for a given reader.

What Are Posner and Weyl Talking About?

The book is divided into five main sections, each advocating for a specific reform: self-assessed property tax, quadratic voting, a new immigration plan, breaking up the large financial conglomerates that currently make banks and other industries behave like monopolies, even if they seem competitive at first glance, and a market for selling personal data. Summarizing all five sections correctly and doing justice to them would take too long, so I will focus on an in-depth summary of one specific section dealing with a new property tax to give readers a sense of what the various ideas in the book are about.

Harberger Tax

See also: "Property is Just Another Name for Monopoly," Posner and Weyl

Markets and private property are two concepts that are often considered together, and it is hard to imagine one without (or even less of) the other in modern discourse. However, in the 19th century, many economists in Europe were both liberals and egalitarians, and it was common to appreciate markets while being skeptical of excessive private property. A rather interesting example is the Bastiat-Proudhon debate from 1849 to 1850, in which the two debated the legitimacy of charging interest on loans, with one side focusing on the common good of voluntary contracts and the other on their potential for unearned income for those with capital, leading to imbalances in capital accumulation.

It turns out that it is absolutely possible to have a system that includes markets but does not include property rights: at the end of each year, collect every piece of property and let the government auction each one off at the beginning of the next year to the highest bidder. This system is intuitively very impractical and unrealistic, but its benefit is that it achieves perfect allocative efficiency: every year, every item goes to the person who can derive the most value from it (i.e., the highest bidder). It also provides the government with a large amount of revenue that could fully replace income and sales taxes or fund a basic income.

Now you might ask: doesn't the existing property rights system also achieve allocative efficiency? After all, if I have an apple that I value at $2, and you value it at $3, then you can offer me $2.50, and I will accept. However, this does not take into account imperfect information: how do you know my value is $2 and not $2.70? You could offer to buy it for $2.99, ensuring that if you really value the apple more, you will get it, but you will not actually gain anything from the transaction. If I set the price, how do I know your value is $3 and not $2.30? If I set the price at $2.01, then I will gain almost nothing from the transaction. Unfortunately, there is a result known as the Myerson-Satterthwaite theorem that means no solution is efficient; that is, in this case, any bargaining algorithm will at least sometimes lead to mutually beneficial trades failing, resulting in inefficiency.

If you have to negotiate with many buyers, things become even more difficult. If a developer (in the real estate sense) is trying to do a large project that requires purchasing 100 existing properties and has already gotten 99 to agree, then the remaining one has a strong incentive to charge a very high price, far above their actual personal property valuation, hoping the developer has no choice but to pay.

Well, not necessarily no choice. But it is a very inconvenient choice that wastes both private and social resources.

Re-auctioning everything every year completely solves the allocative efficiency problem, but the cost of investment efficiency is very high: if it is going to be taken from you and re-auctioned six months later, then it makes no sense to build it. All property taxes have this problem; if it costs you $90 to build a house that brings you $100 in revenue, but if you build the house, you have to pay an additional $15 in property tax, then you will not build the house, and that $10 in profit will be a loss to society.

A more interesting idea from 19th-century economists, particularly Henry George, is a property tax that does not have this problem: the land value tax. The idea is to tax the value of the land rather than the improvements made to the land; if you own a piece of land worth $100,000, you must pay $5,000 in tax every year, regardless of whether you build an apartment on it or just use it as a place to walk your dog.

Weyl and Posner do not believe that a Georgian land tax is feasible in practice:

For example, consider the Empire State Building. What is the pure value of the land beneath it? People can infer its value by comparing it to the value of adjacent land. But the building itself defines the community around it; demolishing the building would almost certainly change the value of the surrounding land. Land and buildings, even communities, are tightly interconnected, making it difficult to determine a separate value for each.

One could argue that this does not preclude another Georgian-style land tax: a tax based on the average value of property in a sufficiently large area. This would preserve such properties in that improving a piece of land would not (greatly) improperly increase what they have to pay in taxes without having to find a way to distinguish between land and improvements in the absolute sense. But in any case, Posner and Weyl continue to propose their main suggestion: self-assessed property tax.

Consider a system where owners self-designate the value of their property and pay a tax rate of 2% of that value each year. But here is the twist: whatever value they designate for their property, they must be willing to sell it to anyone at that price.

If the tax rate equals the opportunity to sell the property each year, then this would achieve optimal allocative efficiency: if you raise your self-assessed property value by $1, your tax payment will increase by $0.02, but this also means there is a 2% chance someone will buy that property and pay $1 more, so there is no incentive to cheat in either direction. It does harm investment efficiency, but far less than re-auctioning all properties every year.

Posner and Weyl subsequently point out that if higher investment efficiency is needed, a mixed solution with lower property taxes could be adopted:

When gradually reducing taxes to improve investment efficiency, the loss of allocative efficiency is less than the gain in investment efficiency. The reason is that the most valuable sales are those where the price the buyer is willing to pay is far above the price the seller is willing to accept. These trades are the first to be realized through price reductions, as even a small price decrease will avoid blocking these most valuable trades. In fact, it can be seen that the scale of social loss caused by monopolistic power grows quadratically within this power range. Therefore, reducing the tax by one-third eliminates nearly five-ninths of the allocative harm brought by private ownership.

This concept of quadratic deadweight loss is a truly important insight in economics and can be said to be a deep reason why "moderation in all things" is such an appealing principle: the first step you take away from extremes is often the most valuable.

Then, the book continues to give a series of ancillary benefits that this tax could bring, as well as some drawbacks. One interesting ancillary benefit is that it eliminates the information asymmetry flaw present in today's real estate sales; even in potentially misleading ways, owners have an incentive to make their properties look good. With a properly set Harberger tax, if you somehow manage to deceive the world into thinking your house is worth 5% more, then when you sell it, you will gain an extra 5%, but before that, you will have to pay an additional 5% in taxes, or someone will quickly buy it from you at the original price.

The drawbacks are smaller than they seem. For example, a natural drawback is that it exposes owners to uncertainty, as someone could buy their property at any time, but this is hardly an unknown, as this is a risk that renters face every day. However, Weyl and Posner do propose more gentle methods of introducing the tax that do not have these issues. First, the tax could apply to types of property currently owned by the government; it is a potentially better option than continuing government ownership and traditional full privatization. Second, the tax could apply to forms of property that are already "industrialized": radio spectrum licenses, domain names, intellectual property, etc.

The Rest of the Book

The remaining chapters present similar ideas that are spiritually akin to the discussion of the Harberger tax, using modern game theory principles to create mathematically optimized versions of existing social institutions. One of the proposals is about quadratic voting, which I summarize as follows.

Imagine you could vote multiple times as needed, but voting requires "voting tokens" (assume each citizen is allocated voting tokens each year), and it costs tokens in a nonlinear way: your first vote costs one token, your second vote costs two tokens, and so on. If someone feels more strongly about something, they would be willing to pay more for a vote; quadratic voting leverages this by perfectly matching the number of ballots to the cost of voting: if you are willing to pay up to 15 tokens for a vote, you will continue to buy ballots until the last one costs 15 tokens, so you will have cast a total of 15 votes. If you are willing to pay up to 30 tokens for a vote, you will continue to buy ballots until you can no longer purchase more ballots for less than or equal to 30 tokens, so you will ultimately cast 30 votes. Voting is "quadratic" as the number of votes rises proportionally.

After this, the book describes a market for immigration visas that could greatly increase the number of accepted immigrants while ensuring local residents benefit, while adjusting incentives to encourage visa sponsors to choose immigrants who are more likely to succeed in the country and have a lower likelihood of committing crimes, followed by strengthening antitrust laws, and finally the idea of establishing personal data markets.

Markets Are Everywhere

There are many ways to respond to each individual suggestion made in the book. For example, I personally think that the immigration visa plan proposed by Posner and Weyl is well-intentioned, looking at how it could improve the status quo, but it is also overly complex; to me, a plan to auction or sell visas each year, with an additional requirement for immigrants to obtain liability insurance, seems simpler. Robin Hanson recently proposed greatly expanding liability insurance mandates as an alternative to various regulations; while imposing new mandates on society as a whole seems unrealistic, a newly expanded immigration plan seems like an ideal place to start considering them. Paying for personal data is interesting, but there are also concerns about adverse selection: to put it mildly, those willing to sit there all year submitting vast amounts of data to Facebook to earn $16.92 (Facebook's current annual revenue per user) are not the same people advertisers are willing to spend hundreds of dollars trying to pitch Rolexes and Lamborghinis to. However, I find the general principles that this book tries to promote more interesting.

Over the past century, a great deal of research has indeed been conducted to design economic mechanisms with ideal characteristics that outperform simple bilateral trading markets. Some of this research has been implemented in certain specific industries; for example, combinatorial auctions are used for airports, radio spectrum auctions, and several other industrial use cases, but it has not really penetrated into any broader policy designs; the political institutions and property rights we have are still roughly the same as they were two centuries ago. So, can we leverage modern economic insights to reform foundational markets and politics so deeply, and if so, should we?

Typically, I like markets and clean incentive mechanisms, dislike politics, bureaucracy, and ugly hacks, I love economics, and I very much enjoy using economic insights to design better-functioning markets so that we can reduce the bureaucracies and ugly hacks in politics and society. Therefore, it is natural that I like this vision. So let me be a good intellectual and oppose it as best as I can.

The complexity of economic incentive structures and markets is limited because users' ability to think and reassess and continuously measure the valuation of things is limited, and people value reliability and certainty. Quoting Steve Waldman’s critique of Uber's surge pricing:

Ultimately, we need to consider the problem of economic calculation. In macroeconomics, we sometimes trade off between constantly rising and unpredictable variable price levels and full employment. Whether wise or not, our current policy is to stabilize price levels, even at the cost of short-term output and employment, because stable prices allow for more long-term economic calculation. The kind of fuzzy benefits that are not visible on the supply and demand graph are considered worth a significant sacrifice. The same concerns exist in a microeconomic context. If the "ridesharing revolution" really takes hold, many of us will have to decide whether to own a car or rely on the world's Sidecars, Lyfts, and Ubers to get us to work every day. To make these calculations, we need something like predictable pricing. Commuting to our minimum wage jobs (which has already ended on average!) through Uber might be okay under standard pricing, but not so much during surges. In the desperate utopia of "free market economists," there is always a way to solve this problem. We could define a futures market for Uber rides, thus hedging against price volatility risks! In practice, this is unlikely…

And:

It is clear that in many cases, people have a strong preference for price predictability over immediate access. The vast majority of services we purchase and consume are not priced in any finely grained way. If your barber or mechanic is busy, you will be scheduled for next week…

For the same reasons, strong property rights are valuable: aside from the arguments about allocative and investment efficiency, they also provide psychological convenience and planning advantages of predictability.

Notably, even Uber itself does not implement surge pricing in the "market-based" way recommended by economists. Uber is not a market where drivers can set their own prices, and passengers can see the available prices and weigh their choices between price and wait time. Why doesn’t Uber do this? One argument is, as Steve Waldman puts it, "Uber itself is a cartel," which wants the ability to adjust market prices, not only for efficiency but also for profit maximization, strategically setting prices to drive out competing platforms (as well as taxis and public transport) and public relations. As Waldman further points out, one of Uber's competitors, Sidecar, did allow drivers to set prices, and I would add that I have seen ride-sharing apps in China where passengers can offer higher prices to drivers in an attempt to entice them to get a car faster.

One possible counterargument Uber might give is that drivers themselves are not as good at setting optimal prices as Uber's own algorithms, and that people generally value the convenience of a one-click interface over the psychological complexity of considering prices. If we assume Uber has fairly won its market dominance over competitors like Sidecar, then the market itself has decided that the economic benefits of marketizing more things are not worth the psychological transaction costs.

At least from my perspective, the Harberger tax seems likely to lead to these exact types of problems multiplied by ten; people are not experts in property valuation, and they will have to spend a lot of time and energy figuring out their self-assessed value of their house, and if they accidentally undervalue it, they will complain more when they suddenly find their house is gone. If the Harberger tax also applies to smaller property projects, people will need to conduct a lot of psychological assessments of everything. Similar criticisms may apply to various personal data markets, if implemented in full form, and could even apply to quadratic voting.

I could respond by saying, "Ah, even if this is true, it’s the 21st century, we can have companies build AI to make pricing decisions on your behalf, and people can choose to use the AI that seems to work best; it could even become public choice"; Posner and Weyl themselves think this is likely the way to go. This is where the interesting conversation begins.

A Story from Crypto Land

One reason this discussion particularly interests me is that the cryptocurrency and blockchain space itself faces similar challenges in some cases. In the case of the Harberger tax, we actually did consider a nearly identical proposal in the context of the Ethereum Name Service (our decentralized DNS alternative), but the proposal was ultimately rejected. I asked the ENS developers why it was rejected. To paraphrase their response, the challenges are as follows.

Many types of ENS domain names only have two classes of participants who would be interested: (i) the "legitimate owner" of a given name, and (ii) scammers. Moreover, in certain specific cases, legitimate owners are particularly underfunded, and scammers are particularly dangerous. One special case is MyEtherWallet, an Ethereum wallet provider. MyEtherWallet provides an important public good for the Ethereum ecosystem, making Ethereum easier to use for thousands of people, but it can only capture a small portion of the value it provides; thus, its budget for domain bidding is low. If a scammer gains control of that domain, users who trust MyEtherWallet can easily be tricked into sending all their Ether (or other Ethereum assets) to the scammer. Therefore, since any domain name typically has a clear "legitimate owner," a purely property rights system would almost not cause allocative efficiency losses, and the public has a strong overriding interest in the stability of references (i.e., that a legitimate domain name won't redirect to a scam the next day), the potential downsides of Harberger tax levels may outweigh the benefits.

I suggested to the ENS developers that the Harberger tax be applied to short domains (e.g., abc.eth) but not to long domains; the response was that having two classes of names was too complicated. That is to say, perhaps there are some versions of the proposal that could meet the specific constraints here; I would love to hear Posner and Weyl's feedback on this specific application.

Another story from the blockchain and Ethereum space, with a more radical market conclusion, is transaction fees. The concept of psychological transaction costs, even considering whether some small payments for a given digital good are worth it, is often used as an argument for why widespread adoption of blockchain technology would be difficult: every transaction incurs a small fee, and figuring out what fee to pay is a major usability barrier in itself. These arguments were further amplified at the end of last year when Bitcoin and Ethereum briefly saw transaction fees soar over 100 times due to high usage (talk about surge pricing!), and those who accidentally did not pay a sufficiently high fee saw their transactions stuck for several days.

That said, we can now say that this problem has largely been overcome. After the peak at the end of last year, Ethereum wallets developed more advanced algorithms to choose which transaction fees to pay to ensure transactions are included in the chain, and today most users are happy to simply comply with them. From my personal experience, the psychological transaction costs of worrying about transaction fees do not exist, just as car drivers do not worry about the gasoline consumed with every turn, acceleration, and braking.

Personal pricing AI for interacting with public markets: has already become a reality in the Ethereum transaction fee market

We are considering a third "radical market" implemented in the context of the Ethereum consensus system, which is a market incentivizing validator nodes to decentralize to prove stake consensus. Decentralization in blockchain is important, similar to the challenges that antitrust laws attempt to address, but the tools we can use are different. Posner and Weyl's antitrust solution, which prohibits institutional investment funds from holding shares in multiple competitors in the same industry, is too subjective and reliant on human judgment to work in blockchain, but for our specific case, we have different solutions: if a validator node makes a mistake, it will be penalized in proportion to the number of other nodes that made a mistake around the same time. This incentivizes nodes to set themselves up in such a way that their failure rate is maximally uncorrelated with everyone else's, thus reducing the chances of many nodes failing simultaneously and threatening the integrity of the blockchain.

In summary, I am cautiously optimistic that the various behavioral issues in implementing "radical markets" can be addressed with good default settings and the help of personal AI, although I do believe that if we are to push this vision, the biggest challenge will be finding increasingly larger and more meaningful places to test it and prove that the model works. I particularly welcome using the blockchain and crypto space as a testing ground.

Another Radical Market

The entire book tends to focus on centralized reforms that can be implemented economically from the top down, even if their intended long-term effects are to push more decision-making power to individuals. These proposals involve a massive restructuring of how property rights work, how voting works, how immigration and antitrust laws work, and how individuals view their relationships with property, money, prices, and society. But it is also possible to leverage economics and game theory to propose decentralized economic institutions that can be adopted by smaller groups at a time.

The most famous examples of decentralized institutions in the fields of game theory and economics might be (i) guarantee contracts, and (ii) prediction markets. Guarantee contracts are a system where some public goods are funded by allowing anyone the opportunity to stake funds, and only when the total amount staked exceeds a certain threshold is the stake collected. This ensures that people can contribute because they know they will either get their money back or there will actually be enough money to achieve some goal. One possible extension of this concept is Alex Tabarrok's dominant guarantee contract, where if a given guarantee contract does not raise enough funds, the entrepreneur proposes to refund participants more than 100% of their deposits.

Prediction markets allow people to bet on the likelihood of events occurring, and it is even possible to condition it on taking some action (I bet $20 that if candidate X wins the election, the unemployment rate will go down). There are some techniques for subsidizing markets for those interested in information. Any attempt to manipulate the probabilities shown by prediction markets will only create opportunities for people to earn free money by betting against the manipulator (yes, I know, risk aversion and capital efficiency, etc.; still close to free).

Posner and Weyl do indeed provide an example of what I would call a decentralized institution: a game that chooses who gets the assets in the case of divorce or company splits, where both parties provide their valuations, and the person with the higher valuation gets the item, but they must give the loser an amount equal to half the average of the two valuations. Some economic reasoning suggests that this solution, while not perfect, is still close to a mathematical optimal solution.

I have been interested in a class of decentralized institutions that improve the incentives for content publishing and content management in social media. Some of my ideas include:

• Proof-of-stake conditional hash cash (when you send someone an email, if they think it’s spam, you can have them burn $0.50)
• Prediction markets for content moderation (using prediction markets to predict the outcomes of content moderation votes, thus encouraging a fast content pre-moderation market while penalizing manipulative pre-moderation)
• Conditional payments for paid content (after you pay for a piece of downloadable content and view it, you can decide afterward whether you should pay the author or proportionally refund previous readers)

And my ideas in other contexts:

• Calling up guarantee contracts
• DAICO (a more decentralized and safer alternative to ICOs)

Others' general ideas about decentralized institutions include:

• TrustDavis (adding skin in the game to e-commerce reputation by providing e-commerce ratings to protect others from fraud by rating recipients)
• Circles (decentralizing basic income through local alternative coin issuance)
• CAPTCHA service markets
• Digitized peer-to-peer rotating savings and credit associations
• Token-managed registries
• Crowdsourced smart contract truth oracles
• Using blockchain-based smart contracts to coordinate unions

I would love to hear Posner and Weyl's thoughts on these "radical markets" that people can initiate and start using without potentially controversial societal-wide changes to politics and property rights. Can such decentralized institutions be used to address the key decisive challenges of the 21st century: promoting beneficial scientific progress, developing public goods of information, reducing global wealth inequality, and the big issues behind fake news, government-driven and corporate-driven social media censorship, and cryptocurrency product regulation: how do we ensure quality assurance in an open society?

In summary, I strongly recommend Radical Markets to anyone interested in these kinds of issues (by the way, I also recommend Eliezer Yudkowsky's Inadequate Equilibria) and look forward to the discussions that the book will spark.