Why I Am not a Market Radical
E. Glen Weyl
December 15, 2020
This essay is the latest in a series I have written clarifying my views and contrasting them with other schools of thought. In many ways, these essays have nearly all been self-critical, as at some point in time I identified closely with nearly all the ideologies I critique in those pieces. Yet the position I am most publicly identified with is the views laid out in my joint book with Eric Posner, Radical Markets: Uprooting Capitalism and Democracy for a Just Society (RM henceforth), which came out two and a half years ago. In this essay, I want to make clear what I see as the most important limitations and failings of the views expressed there.
Before doing so, it is important to clarify the scope of this essay. I continue to believe that many of the ideas expressed in RM were fundamental advances in social technology (the ways we organize our politics and economy). I have been pleasantly surprised by the wide adoption of the ideas and have been quite amazed that, to my knowledge so far, no large-scale harm or dissatisfaction has resulted from these applications. Given all the weaknesses in the ideas, which I will highlight below, this is more than I reasonably could have hoped for. I remain a strong supporter of the efforts by groups like the RadicalxChange Foundation and others to experiment with these ideas and harness them as part of cultural, political, technological, and academic movements for transformation.
I also continue to believe that elements of the sprit behind them, such as deeply fusing markets/freedom with democracy/equality and approaching political economic institutions with the same creativity we approach technology, are crucial to human survival on this planet. The dominant existing ideologies of today, such as statism and capitalism, depend even more heavily (as I highlight in my critiques of those ideologies) on the problems with the ideology in RM than does the vision there and I think RM made critical strides in moving beyond those ideologies. In fact, it is hard for me to know whether the more nuanced understanding I’ve now come to would basically change the arguments I laid out there, which served an important purpose when they were made and managed to catalyze significant social energy.
However, it is precisely my dedication to these principles that leads me to write in a much more critical vein today, for a few reasons:
- If one takes seriously the view of political economy as social technology, one must hope it will continue to advance and that the ideas in RM are a few among many other tools needed to continue to advance in social organization. No matter how much a text is written in such a spirit, there is always a tendency for ideas to ossify as they gain momentum, a tendency I wish to head off with self-critique, to ensure continued evolution.
- While I am not aware of any serious wrong turns with applying the book’s ideas, this seems to me an inevitable consequence of scaling absent highlighting failure modes, of which I have learned a lot from critics. With enough eyes and time, all bugs are shallow, and I want to help direct eyes and time by focusing on what strikes me today as the greatest failure modes.
- Some of the greatest enthusiasm for the ideas has come from quarters (including the Effective Altruism movement, the Rationalist community, the blockchain community, and the technology community more broadly), which tend to have an attitude towards the role of formal systems in society that I would call technocratic and have critiqued elsewhere. While folks from these communities have tremendous intellect and creativity to offer as designers, their general technocratic attitude worries me that some may advocate for top-down, rather than experimental, application of some RM ideas that could be dangerous. I hope this critique will help avoid such harms.
- Perhaps most importantly, many (religious) Conservatives, establishment moderate small “c” conservatives, and communitarian-inclined folks on the far left have been rightly “frightened off” by the type of interpretations among technocrats that also concern me. RadicalxChange needs allies from these communities, with whom I personally feel deep kinship. It is thus important to clarify where I, and hopefully most people in the RadicalxChange community, depart from some of these interpretations of RM.
With these explanatory remarks in mind, I turn to the substance of this essay, the tone of which is almost entirely (self-)critical, though I will gesture towards some potential directions for addressing these criticisms with a next generation of designs. The fundamental weakness of RM is that it is profoundly economistic and suffers from the two leading confusions that beset the field of economics: an atomistic individualism and an authoritarian technocratic perspective on social change. This is particularly paradoxical given that the book uses this economistic perspective to take on fundamentally social issues, like public goods and common ownership, and to promote decentralized experimentation with fundamentally different kinds of living. In this sense, the book essentially assumes itself away. Its mechanism design approach to social technology is grounded in an assumption that the common goods it focuses on hardly exist. In what follows I will lay out this critique abstractly and then detail how it affects designs proposed in the book.
More than a substantive field, economics can most easily be understood as the application to social analysis and design of a particular branch of enlightenment thought that I will call Atomistic Liberalism and Objectivist Naïve Epistemology (ALONE). Central to ALONE is a binary between Individuals, conceptualized as largely presocial, independent ultimate loci of value/preference/good/belief (well-being for short), and some global coordination device variously referred to as the social planner, objective truth, the modeler, the mechanism designer, the impartial observer, God or, most commonly and how I will refer to it, The State.
Social structures that intermediate between these two poles are of secondary importance to ALONE and are usually either abstracted from, mechanized, or sharply reduced. The primary problem in the ALONE world view is how The State can be structured so that individuals acting in their selfish interests will maximize some aggregate of well-being. It is from this binary that the initially paradoxical combination of extreme individualism and authoritarian technocracy that characterizes economic thought arises. By largely assuming away intermediate social structures, ALONE makes the only natural locus for social change interventions by The State.
ALONE has been a highly tractable and thus widely applied framework. Much of the success of economics in gaining social power is attributable to the tractability of ALONE to formal analysis. The framework has been useful, especially through its inspired social systems, like the nation-state and capitalism, in realizing significant gains in standards of living, life expectancy, population, that were impossible when cooperation was more tightly bounded by space and community boundaries.
However, the ALONE model of the world is wildly naïve and inaccurate, as is understood by anyone who has spent even modest time reflecting on it. In fact, the central, founding insight of a whole range of academic fields, from continental philosophy to sociology, is how utterly mistaken ALONE is. It is outside the scope here for me to try to argue for this, though I do recommend to those skeptical of this a brilliant essay by David Foster Wallace. Let me just list a short and selective, but illustrative list of canonical insights from this intellectual tradition:
- The production and consumption of most value, and the origin and standards of ideas/beliefs/truth, are fundamentally social (as opposed to individual), but not universal, in nature. Instead, they emerge from diverse and often intersecting social contexts.
- Much of culture can be understood in terms of the way people communicate and convey meaning in different contexts. Because these ways are always partial, symbolic, imperfectly credible, etc. cultures will always have dysfunctions and forms of oppression/waste built into them.
- The possibility of meaningful individual “agency” or “autonomy” is of relatively recent origin and is grounded in the proliferation of diverse social settings and communities. This novelty has been born of modern political economic institutions, such as the city. In more closed and narrow societies, individuality, “selfishness”, and the like had little meaning as nearly all interacting members had the same loci of value.
- Social conflict, which formal institutions may help resolve, results not from selfishness but from the diversity characteristic of modern societies. Because in modern societies typically interacting individuals have only a few loci of value in common, they may come into conflict. Reconciling these diverse social commitments is a primary challenge of politics.
Taking any of these points serious fundamentally challenges the mechanism design methods we used in RM. While all, and many more, are important, let me focus on one that is particularly transformative and relatively easy to explain in economics language.
A central focus of RM is on the importance of collectively created value, public goods, and more broadly what economists would call “increasing returns” (IR): any situation in which many people or units of production can collectively achieve more than the sum of what they might individually achieve. Colloquially, any case where the whole is greater than the sum of the parts. Quadratic Voting (QV) is designed to near-optimally govern the collective choices that such IR settings constantly present; the Common Ownership Self-Assessed Tax (COST) is designed to broadly share collectively created value and to ensure flexible reallocation of properties to allow for IR activities that could otherwise be blocked by holdout. The book repeatedly asserts that such situations are of fundamental and systemic importance and that addressing them more appropriately could be transformative. This claim is not especially controversial within economics; essentially every major analysis that has considered the issue has emphasized the centrality of IR; for more on all this see another recent essay of mine.
Yet, ironically, while IR settings are core to the motivation of RM, the methodology of mechanism design central to the book assumes they are of minimal importance. That methodology assumes what economists call “quasi-linear utility”: put briefly, every participant cares only about how much money (or other currency) she has herself and the outcome of the allocation problem at hand. Such an assumption is (approximately) valid if most of the world is constituted of “private goods” that bring benefit only to a single individual and are thus well-summarized by the money that person has, there are almost no instances of IR and the economy is near equilibrium, among other assumptions.
On the other hand, if IR is prevalent, this assumption is far from correct. In a world of prevalent IR, individuals in the economy (effectively) engage in collaborative consumption and production. To the extent that agents interacting in some setting are part of IR processes together, even if they are “selfish” in the sense of maximizing their own utility, individuals will care about the allocation of resources to others as others will use those resources towards common causes. To take an extreme example, given that most consumption of a romantic couple is shared, the allocation of resources to one member of the couple is almost equivalent to the allocation to the other. Other examples are weaker and richer, but we should expect those living in the same city, working for the same business, etc. to all care about the allocation of resources to those they have close social ties to.
The standard quasi-linearity assumption will therefore only hold in the uncommon case where everyone participating in the mechanism is at an equal social remove from others participating in it and from whatever social group receives the payments collected in the mechanism. Beyond this special case, an approximately optimal mechanism will require as input not just individual identity and payments, but also the social relations of common value creation among the participants. That is, it will have to take as input social identity and value rather than individual identity and money. It will have to view the basic problem not as reconciling individual selfishness with the common good, but rather achieving cooperation across lines of social difference. Otherwise, all its conclusions will be badly wrong.
A canonical example to consider is an auction. In standard economic theory, auctions are the prototypical efficient way to allocate a resource, because they place all potential purchasers in a symmetric position of competition with one another forcing all to compete on a level playing field. Yet, in a world of increasing returns, an auction does not put all participants on an even competitive footing. Suppose bidding is for a work of art. Two bidders are art collecting co-workers at the same office in New York City, a third is a public gallery in the city towards which the first two bidders make charitable contributions and a fourth is bidding for the government of a developing country which believes the work should be repatriated. The first two bidders will not simply be indifferent over who wins the artwork assuming it is not them individually, nor over how the winning bidder pays. The co-workers will view the other winning as nearly as good as their winning (possibly even better), assuming both will hang the work in their joint office. And they will view the public gallery winning, and saving money, as highly desirable relative to the developing country, or seller.
In short, the auction will not lead to an impartial or competitive allocation of the artwork unless the social relations in which the participants stand with respect to one another feature in the design. More broadly, the primary conflict that will arise in designing mechanisms of this sort will not be selfishness, but rather tendencies to favor one’s in-group. To overcome this, mechanisms will need to not be so much compatible with individual incentives, but rather to foster cooperation across lines of difference.
To highlight the incoherence of many of the proposals in the book with the world of prevalent and diverse IR that the book suggests, it is instructive to consider the ideas individually (though briefly):
Common-Ownership Self-Assessed Tax (COST): The idea in the book that most clearly builds on the concept of an auction was the common-ownership self-assessed tax with a right of compulsory purchase. In justifying and explaining the idea, we made repeated and quite vague references to the different way in which the value is usually the result of community effort; yet community structure and public goods are almost completely absent from the proposal. Yet, upon reflection, it is clear that any plausible/workable large-scale version of the COST should be intimately and inextricably intertwined with community, as the very name of the proposal cries out.
The book roughly imagines the tax applying to individual property and all revenues being rebated by some sort of global super state back to individuals. Yet this is obviously not the way such a thing should or could work. Most wealth is not held directly by individuals; it is held in a variety of collective entities such as corporations, nations, etc. While the precise nature of these entities might well change in the future, the book imagines, it seems even more likely that in that future the relevant owners will be such entities. One can easily imagine nested or overlapping structures in which a variety of collective entities are all partial common owners of a property, with the imagined cosmopolis in the book owning only a small share.
Optimal taking rules (viz. who is eligible to take at what price, with what bundle), tax rates and revenue disbursements in such a world seem likely to be quite different from a universal basic income. Income would accrue to a variety of entities discussed above and likely be used primarily to finance public goods of various sorts, which themselves are the source of the value of the assets collected by the tax. One can imagine all sorts of potential schemes (e.g. ones operating at many levels of traditional government or ones based around “voluntary” associations like condo boards) consistent with the general principles behind the COST. In practice, in both the longer term and even in many emerging applications, precisely the way community structure interacts with COST will be critical to whether it is a successful, appealing and legitimate social technology, something completely overlooked in the book.
Quadratic Voting (QV): While aimed specifically at facilitating the creation of public goods, QV was itself based firmly on ALONE. Votes are tallied as the sum over participating individuals of the square root of the tokens each contributes (in favor of or against a particular option). This treats each participant as an isolated atom, symmetrically positioned relative to every other isolated atom in the system, giving extra weight to small value from many atoms. As many insightful critics pointed out, this is preposterous and leads to endless paradoxes. Should romantic partners be inside or outside the square root with regards to a vote? Should co-workers? Should members of the same religion? The answer will probably almost always be “somewhere in between”. Otherwise, the process is likely to be dominated by tight social cliques (much like today’s “special interest groups”) rather than a broader reading of the public good. Yet by imagining the only primitives in the model as individual identity and a token, QV cannot account for any of these critical social relations, raising a range of issues in practice that get categorized as “collusion” and are in fact simply the natural working of the kinds of sociality that QV aims to enable.
Visas between Individuals Program (VIP): In its very name, VIP was imagined as a contract between two individuals, and it was thought that the right to offer the visa belonging to individual citizens would be sufficient to ensure the rents associated with migration would end up in the hands of these individuals. Furthermore, it was designed as a right to issue a visa, rather than a share of auction revenues, precisely because it wanted to give individuals some agency over the social relations created through migration. Yet an atomistic individualist design like this was clearly inconsistent with these goals, given the analysis in the rest of the book.
Individuals would end up relying on matching engines and vetting processes to help them select those they sponsor. Most social interactions happen at community or larger scales, not at the individual level, so attempts to create agency over non-pecuniary aspects of migration cannot simply occur at the individual level. While the book vaguely alluded to both issues, it treated these central problems as if they would simply resolve themselves “by the magic of the market”, the precise magic that the rest of the book was about ensuring was reliable. Without further design, or at least thought about these social structures, it seems entirely plausible that the capture by capital of the rents from migration would simply reemerge and that social disruption from migration would become significantly worse. Thoughtful versions of things like the VIP, as proposed recently by the Canadian government and the campaign of Democratic Presidential Candidate Pete Buttigieg, focused heavily on harnessing community organization to avoid such outcomes, something entirely missed in the book. Danielle Allen eloquently highlights this critique and these superior alternatives in her forthcoming book, Justice by Means of Democracy, which has deeply influenced my thinking here.
Antitrust: While celebrating the radical potential of antitrust, the book almost completely ignored the issue of increasing returns and economies of scale that make antitrust an issue in the first place. The idea of Data as Labor in a separate chapter takes for granted that scale issues make antitrust mostly a non-starter in addressing the power of technology platforms. However, the analysis of the financial industry and labor market in the antitrust chapter completely ignored these issues and assumes that market power simply arises from thin air. This, without any serious analysis, put one set of issues into an antitrust bucket and another into a countervailing power bucket and made no attempt to show how these can work together or what the appropriate sphere of each is. A more serious attempt to integrate these frameworks would doubtless have highlighted the role of unions and other countervailing power in at least partially addressing the labor and financial market power issues which we treated as purely antitrust questions. In work with Zoë Hitzig, Michelle Meagher and André Veiga, I have begun to attempt such an integration.
Data as Labor (DaL): Part of the purpose of the “data as labor” analogy was to conjure the history of the labor movement and the importance of collective organization and collective bargaining. Yet, simultaneously, DaL’s overwhelming focus on economics, on bargaining for a share of the digital pie, rather than on reconceptualizing the loci of power in the digital economy over issues such as information quality (“fake news”), information flows (“privacy”), and attention missed many of the concerns that were most pressing for collective organization to address. This was most glaring in the opening vignette, which ludicrously chose about as extreme an example as possible of one participant being induced to share a datum that mostly pertained to others to an artificially intelligent overlord…about as far from a vision of labor collectively empowered as one can imagine. This example ignored perhaps the strongest case for organizations like data unions, namely that most data is inherently social and not individual, and thus cannot meaningfully be owned by individuals. Future evolutions of the framework must heavily incorporate the social nature of data and the other bargaining issues that platform users face, and increasingly have been doing so.
RM aimed to help imagine and set in motion a transformation of formal political economic systems to foster cooperation and equality. To accomplish such a goal, those inspired, both positively and negatively, by its example will have to move far beyond the approach taken in the book. They will need to build designs far more liberated from ALONE. The coming years will tell whether building such movement for liberation through social innovation is possible, and what role if any RM will have played in inspiring it. I am, however, hopeful, that the best uses of these ideas, such as those in Taiwan, which have used the ideas in the most thoughtful ways, will be the greatest example to the future.