Complexity As a Fundamental Diseconomy of Scale

I want to begin by praising Apolito’s piece. It both tackles the strongest argument against anarchism, the problem of achieving coordination at scale, and models the problem using insights from the cluster of fields that lie at the base of complex systems. Whatever my disagreements with them over markets, this is a welcome addition to the discourse. There is an obvious connection between the ideas of anarchism and the ideas in complex adaptive systems,1 but few people have rigorously teased them out. Apolito’s piece is clearly a step in that direction.

The clarity that they bring however is in considerable contrast with how they respond to markets. Claims like “I view the market mechanism as running on a steepest descent towards a cost/energy minimum, in an attempt to maximize profit, which inevitably singles out the least valuable options, while wiping out anything that is of any value … along the process” are made without engaging with any evidence from left market anarchists to the contrary (hell, it doesn’t even engage with counter arguments that you’d pick up from an introductory course on economics).

Now for what it’s worth, such dismissals are made within a context of over a century of left wing critique of markets. I’m familiar enough with enough of these that I can respond to Apolito’s content-less argument. But the dichotomy between Apolito’s brilliant articulation of what I see as the core problem for any functional anarchist political order (coordination at scale) and their critique of markets is maddening.

Part of the problem is that critiques of markets vary significantly in both their quality and implications. The underlying reasons why economic processes work the way they do cannot be separated from other social questions. As such any economic critique or analysis must come with non-economic ramifications that are explicit or implicit. 

I suspect that the implications of the Austrian critique of non-market forms of economic organization were a part of the reason the left struggled to make progress on the economic front in the postwar era, as well as why adopted its “weirdly skewed” perspective towards science and technology. As society and the field of social struggle became more complex, the frameworks that had previously sufficed to describe the dynamics at play began to be less and less relevant. But because it had wedded itself to a position in which markets were a fundamental enemy it could not.2 To put forward a coherent picture of the world would be to admit that “bourgeoisie economists” had gotten something right.

What’s frustrating about all this is that the domain the left failed to tackle properly was that of communication and information processing, which as it turns out has consequences for pretty much everything. As such until the left builds a theoretical framework that can grasp these concerns, I have my doubts about its capacity to effect systemic change in the 21st century. While it may benefit from secular trends that happen to favor it, it will not be able to anticipate such dynamics, much less put in the work to make things happen.

Thankfully there is a significant difference between markets and capitalism and we have good reason to think that the arguments against economic planning also cut sharply against runaway accumulation, especially under conditions of broader environmental dynamism and individual empowerment. These make the runaway accumulation that we have seen under capitalism is contingent on systemic state intervention since the beginning.

Now before we begin there are some preliminary concerns that must be addressed. First of all it must be acknowledged that some level of inequality is acceptable. Even Marx, contrary conservative stereotypes, wrote in The Critique of the Gotha Program that there would be some inequality in even a communist society resulting from inequities in capability and differences in choice and preferences. We do not want equality achieved through forcing everyone down, but rather equality of options in terms of what each person can do.

As such, slight levels of inequality are acceptable, especially since the demarcation point at which inequality can result in a positive feedback loop is unclear and because wealth is inherently subjective. What we should be far more concerned about is inequality that gives people the capacity to dominate one another.

Finally there are concerns leftists have with markets beyond inequality. The most common are the negative externalities of commodity production and the values that they promote in the people who partake in them. However I believe that inequality trumps these as a concern. If markets do not inexorably create inequality, these concerns can be remediated. A wealthy, egalitarian society can both afford and politically achieve less environmentally impactful ways of coordinating economic activity. Likewise if markets do corrupt the values of the people who engage in them, the wealth they create can be used to create spaces outside of markets that encourage better values.3

Complexity as the root of fundamental diseconomies of scale

The most fundamental diseconomy of scale emerges out of the difficulties that arise out of the challenges of communication between individuals. The capital of a rich person does just not automatically see returns, it must be invested in a particular way so as to see guaranteed returns. This allocation requires communication. It is here that we see the first diseconomy of scale.

The go to argument for why communication is a problem is that much of what we know is “tacit knowledge,” that we cannot effectively explain the action we do. But it’s important to articulate the underlying reasons as to why tacit knowledge is such a big deal.

The primary reason that tacit knowledge is so difficult to communicate is because of the complexity inherent to our world. Our seemingly normal day-to-day lives are the result of a vast number of tiny processes that aggregate into emergent phenomena. The reason this seems so normal to us is that our brains evolved over hundreds of millions of years to effectively navigate this world by efficiently processing the overwhelming amount of information that we receive through our senses to make both intelligent and timely decisions. And while neuroscience has yet to give us a full picture of how the human mind works, we know that conscious awareness of anything is cognitively demanding. Through exposure and experience we can move our information processing to a subconscious level which makes processing quicker because it does not have to consciously go through the steps of reasoning.

This is an efficient way to process information in contexts we are familiar with, but it comes at the cost of making the broader reasons behind why we arrive at conclusions seem obvious and thus difficult to explain.

Another trick we perform to making processing information easier is to employ simplistic models of the world that leave out dynamics for the sake of tractability. Trying to take literally every single thing into consideration is impossible and as such we must prioritize certain pieces of information over other pieces so we can actually make decisions.4 It is unavoidable, and instead of disparaging simplification and reductionism, we should find more intelligent ways to identify and correct the problems that arise from such simplification.

These two dynamics make effective communication essential when dealing with any remotely complex problem. Even a simple object can be viewed at several different scales of abstraction that reveal insights that would be obscured at other levels of abstraction. Moreover the most valuable cooperation comes from people with considerably different perspectives coming together and combining insights. Being able to effectively communicate and integrate different models of the world is an immense force multiplier.

These are fundamental barriers for any form of organization, but the problems it presents for hierarchical approaches are particularly grave.

The first problem is the structure of simple incentives for information sharing. When people have a stake in something and feel that their feedback will be taken into account, they are more willing to share information. In most firms, management exists purely as a means to direct workers. When the work being done is simple enough this isn’t much of a problem and can be overcome through workplace surveillance and authoritarian management. But as the tasks become more complex and people have to grapple with the inevitable ambiguity, directing people from above becomes difficult-if-not-impossible because of the ambiguity present to language. This is a likely cause of the workplace autonomy being afforded to workers who deal with such ambiguity and complexity like software developers. What this illegibility provides is considerable space for low-level sabotage because of the high cost of detecting whether workers are deliberately slowing down (if it can even be detected in the first place).

To be sure such low-level disruption is not likely to overthrow the existing order. But when we’re talking about the threat of runaway accumulation in the context of a market in which all actors have relatively equal wealth, it can suffice. The communication inefficiencies do not need to ruin would-be capitalists, they merely need to reduce the efficiency of the organization to the point where its returns are lowered to the point where runaway accumulation becomes considerably less likely.

The second problem is that centralized social arrangements create a processing bottleneck. A small set of people must ultimately process all the incoming information and come to a decision. Even if the people in charge are well-meaning, the information they receive must be throttled for them to actually make sense of it. This means they must simplify the information they receive and will inevitably miss important pieces of information. As such even when the people making decisions in the center have good intentions, they will never be able to fully capture what is going on at the edge.

This is why cooperatives are, on average, more efficient when compared to hierarchical firms of similar size. By giving their workers more control over the enterprise, as well as a stake in its success, the cooperative form overcomes the incentive problems to some degree, while also creating formal means by which workers might reconfigure processes.

Now of course an aspiring capitalist might adopt more efficient forms in the end of profit (some tech companies like Valve have done just that).5 But adopting such approaches is in direct tension with runaway accumulation because it places more power in the hands of the worker. When you run along extremely horizontal lines and let people self-organize they can easily withdraw their labor for whatever reason.

But we can go further.

More horizontal firms might be more effective at utilizing knowledge, but they still face limitations in that the boundaries of the firm act as a barrier between workers on the inside and workers on the outside. To truly overcome this problem, workers don’t just need more agency at work, but also agency over who they work with.

An example of how this works in practice can be found in a 1998 paper by Alan Hyde on the labor market in Silicon Valley. He found that a key feature of Silicon Valley that made it overtake other tech hubs within America was that California refused to enforce non-compete clauses. This, combined with a fluid labor market in which workers could easily leave firms to join or start new firms, resulted in workers being able to take the initiative in where they would apply their knowledge and as such generated a general diffusion of information and know-how. As Hyde writes:

What now seems clear, however, is that the phenomenal growth of Silicon Valley rests in part on what I call its “high-velocity labor market.” In such a labor market, employees move rapidly among firms and start their own, and employment relations may easily be created or terminated. In many cases, these relations take unconventional forms that do not constitute “employment” from a legal point of view, such as relationships with consultants or independent contractors.

In such an environment, information naturally disperses because workers in successful companies have the option to experiment for themselves. Whether they do it because they think they can see more financial return then or because they see an alternative use case for the knowledge that they want to try out, the end result is still the same, namely the diffusion of knowledge throughout the economy. As Hyde put it: “[A] secret is only as secure as an employee is happy. Innovations thus diffuse rapidly and are easily imitated, lowering costs to consumers.”

Now of course such freedom is restricted to an elite class of workers. And with the rise of big tech companies this freedom has decreased. The ability of tech workers to walk puts something of a limit on the power of tech companies, but the nature of intellectual property law is a considerable driver of accumulation as we’ve seen with the rise of tech giants. Nevertheless early Silicon Valley serves as an example of how giving workers agency over who they work with puts limits on accumulation by creating strong incentives for trade secrets to be shared. When the cost of market entry is low enough, any worker in a successful firm who wants a larger share of profits can walk and try their luck.

We also see the sharing of knowledge between firms even under capitalism. Market actors will share information when it’s to their long term benefit. A 1999 paper looking at this found something akin to the iterated prisoner’s dilemma between firms when it comes to sharing knowledge. The authors’ explanation for what was going on was: “what matters is not whether others win – it’s a fact of life that they sometimes will – but whether you win …. Sometimes the best way to succeed is to let others do well, including your competitors.”

Decentralized approaches to production are also able to better integrate new discoveries or environmental shifts. Part of this is because of the tacit knowledge dimension. As we’ve already established, trying to explain how something works before you can do it can take a significant amount of time and might be impossible if the situation is sufficiently complex. Giving people the freedom to test their ideas against reality is far more effective than talking things out because it can definitively show whether something works or not.

Such an approach to innovation and experimentation pairs nicely with productive technology that rewards individuals’ agency and creativity. Once upon a time the Marxist argument about the superiority of large scale industry and the necessity of authoritarian management structures that such configuration entail may have been correct. But we’re long past that point. As the world has become more complex, flexibility becomes more important and authoritarian forms of management simply can’t keep up.

This is why anarchists have long favored neo-artisanal tools such as 3d-printers, computers, CNC machines, etc. These technologies give individuals significant agency in terms of work they can do and also require minimal upkeep. While mass production may still be more efficient for producing a particular commodity even when the state is no longer putting its thumb on the scale, the efficiency is only valuable within a particular context. An economy primarily built around decentralized, flexible production can adapt itself to significantly more contexts than a centralized model built around mass production and distribution. This is because such technology can be repurposed far easier than rigid production lines and because it requires significantly less overhead which means downtime or lack of sales is not nearly as bad.

The flexibility offered by such technologies makes runaway accumulation even harder because it significantly increases the range of possible actions. Approaches which run off the dictate of a small group of people cannot effectively evaluate the multitude of ways technologies might be used in conjunction with each other. As the complexity economist Eric D. Beinhocker writes:

[W]e can reinterpret markets as an evolutionary search mechanism. Markets provide incentives for the deductive-tinkering process of differentiation. They then critically provide a fitness function and selection process that represents the broad needs of the population (and not just the needs of a few Big Men). Finally, they provide a means of shifting resources toward fit modules and away from unfit ones, thus amplifying the fit modules’ influence.

In short, the reason that markets work so well comes down to what evolutionary theorists refer to as Orgel’s Second Rule (named after biochemist Leslie Orgel), which says, “Evolution is cleverer than you are.” Even a highly rational, intelligent, benevolent Big Man would not be able to beat an evolutionary algorithm in finding peaks in the economic fitness landscape. Markets win over command and control, not because of their efficiency at resource allocation in equilibrium, but because of their effectiveness at innovation in disequilibrium.6

The fluidity that decentralization and individual empowerment creates means that society can better react to changes in the environment and avoid danger or exploit opportunities more effectively. In their book on complexity science, Gregoire Nicolis and Illya Priogire draw an explicit analogy from the the fluid structure of ant colonies to more dynamic human societies:

[a] permanent structure in an unpredictable environment may well compromise the adaptability of the colony and bring it to a suboptimal regime. A possible reaction towards such an environment is to maintain a high rate of exploration and the ability to rapidly develop temporary structures suitable for taking advantage of any favorable occasion that might arise. In other words, it seems that randomness presents an adaptive value in the organisation of society.7

Decentralization and empowerment are also favored because the complexity of reality means that solutions to problems and possible ways insights can be combined are non-obvious. We see this in a 2011 paper by Joel Lehman and Kenneth O. Stanley found that evolutionary algorithms that optimized for novelty outperformed evolutionary algorithms that optimized for a specific goal in achieving that specific goal.

Now certainly open-ended innovation is possible under more centralized regimes. But it cannot fully unleash innovation. Take, for example, one of the most successful research institutions the world has ever seen, Bell Labs. While it operated, it saw a tremendous amount of innovation and discovery, resulting from the considerable freedom given to the researchers who worked there. But such experimentation and innovation was allowed only for a select few, within a particular context, and required a state-granted monopoly. The institutional approach to innovation is also limited by the fact that institutions need to have some filtering mechanism to decide who is inside and who is outside. This means that the set of people who can cooperate is limited, which means less diversity, which makes solving complex problems more difficult. As Scott Page in The Diversity Bonus writes: 

Diversity bonuses, when they do exist, do so on complex, high-dimensional tasks: solving a problem, predicting an outcome, designing a policy, evaluating a proposed merger or undertaking research. Bonuses arise when these cognitive nonroutine tasks prove too complex for any one individual. No one person can possess sufficient knowledge, tools or understandings to handle this type of task alone.8

Certainly hierarchical institutions can be more flexible and accepting of minorities. One of the reasons that Bell Labs was so successful was that they embraced oddballs. But at the end of the day institutional boundaries have to be drawn and that means an upper limit to the diversity among the people. Even if upper management is genuinely committed to research and diversity, they still can’t get around the information bottleneck that emerges thanks to centralization.

As we’ve already established, centralized structures do not like disruption and so must limit the degree of change that happens under them.

As such, should a firm or handful of firms successfully dominate the market, they will see stagnation. Such a state of affairs may or may not lead to a workers revolution, but what is very likely to occur is something akin to the dynamics described by Joseph Tainter in The Collapse of Complex Civilizations. These hierarchies will start seeing diminishing returns and increasing fragility because they are caught between the need to reorganize and reconfigure aspects of society and the fact that such reorganization would undermine the social formations from which the people at the top derive power, status, and wealth. The system will eventually be reconfigured, either by internal or external forces, into something that can deal with change.

Such inefficiencies and missteps are possible in a more decentralized system. But a virtue of decentralization is that major mistakes can be localized to a small set instead of affecting the whole system. It also necessitates a diversity of actors and strategies acts as insurance against sweeping disruptive change. A sufficiently diverse set of approaches is a good hedge against upheaval because strategies that are counterproductive under one regime may be highly rational under another and vice-versa. This includes both resource and information sharing. This sort of altruism is rational from a self-interested perspective if setting up a state-like entity to extract wealth from subjects is sufficiently expensive. As we’ve already established, empowering others is the most reliable way to see returns for yourself.

There are of course examples of this kind of mutual aid throughout history among actors engaging in “competitive” market behavior. David Graeber’s Debt has many such examples. We see examples of this even under modern capitalism between “competing” market actors. To quote Piore and Sabel’s The Second Industrial Divide:

Both the employer’s association and the unions are active in their members’ ethnic communities. Thus, Jews attend and contribute to fund drives for Italian orphanages, and Italians help out with the United Jewish Appeal. … This effort for the greater community tempers competition within the community: manufacturers and union leaders are always present at the public rituals in the families of their competitors, as well as their colleagues.9

To summarize my argument: as the number of choices available to people increases, we gain more and more value from specialization. More specialization makes communication more important and more difficult. As such, egalitarian networked forms of organization that emphasize individual autonomy become increasingly superior because they facilitate more effective communication between individuals. Moreover as the overall system becomes more complex and dynamic decentralization and autonomy become important to prevent cascading failure and to adapt to changing circumstances. Finally under such dynamic circumstances, the long term sharing of resources and information becomes rational because a state-like entity that can guarantee reliable returns is increasingly difficult to impose thanks to complexity. This general dynamic frustrates both the state and capitalist attempts to accumulate power as both fundamentally rely on stripping choice from individuals to accumulate wealth/power. At the same time the fact that such systems are not singletons and are in competition with other would-be tyrants means that they must allow some degree of agency so they can adapt. This tension between locking society into rigid stasis and allowing for disruptive innovation and adaptation is simultaneously one of the most important dynamics at play in society in the last few centuries and also one of the least remarked upon.

Competition in the service of cooperation and cooperation in the service in competition

This brings me to the primary problem that many leftists have with markets, namely the notion of competition. The counter-accumulative tendencies I’ve described in this essay and the advantages of cooperation can be summarized as competition in service of cooperation. Competitive dynamics over a particular timeframe or at a particular scale facilitate cooperation over a larger time frame or scale by stress-testing the system, pruning unnecessary components, and by giving individuals the capacity to prove their trustworthiness under pressure.

We can find analogies of such dynamics in nature. To give one example, the cells in our bodies engage in what is known as autophagy wherein unnecessary or dysfunctional parts are eaten so as to be repurposed for more effective ends. The immediate competition between individual cells is what enables the emergence of a collective subject that is magnitudes more capable, durable, and interesting than a single cell.

Market dynamics, when they properly work, offer similar benefits.

I should emphatically note that this is not social Darwinism. The reason why autophagy works is because what it recycles is disposable. The range of actions that a simple cell can perform is fairly limited and they reproduce so quickly that their death is nothing to mourn. We want something similar for our economy. The firms, cooperatives, business strategies, artificial agents, etc that make up the economy are akin to cells in that they can be easily spun up and experimented with. Their death is nothing to morun – indeed what you want is to encourage productive mistakes in which useful information can be gleaned from the failure of a particular approach.

This doesn’t just create a churn of approaches to problems, but it also brings about a space in which trust between strangers can be tested and exercised. While non-market spaces and relationships can certainly facilitate such relations, what currency enables that other forms of social connection cannot is cooperation at scale through the creation of a generic token of trust that can be accepted by a much wider circle of people. This acts as a lubricant between people who do not have direct ties and allows for cooperation at a much larger scale. This enables a wider variety of projects to take place by letting people with diverse backgrounds and diverse skillsets to come together. The competition between projects results in better cooperation because nobody is being forced into it, but this also gives us more finely tuned outcomes because people can easily express their preferences and find situations that work for them.

Creating and sustaining such a dynamic space is of immense value to any anarchist social formation for two reasons beyond the obvious economic value it brings. The first is that anything that dynamic and complex is naturally difficult to control. The second is that the cultural shifts brought about by constant change can be beneficial as well. The fact that markets drift not just towards efficiency, but also towards new ways of doing things, creates a population that is psychologically more comfortable with change. Markets with low barriers to entry give incentives for people to reimagine how the world around them could be different, as well as to try to convince others of their vision. Such incentives are clearly in line with left wing prefigurative politics, even if this is not commonly recognized by actual leftists.


This survey of the counter-accumulative tendencies within markets is, of course, incomplete. Nevertheless, what I’ve hopefully demonstrated is that there are powerful counter-accumulative tendencies or diseconomies of scale that hamper wealth accumulation within markets. As the set of choices available to each individual expands, society overall becomes increasingly more complex which makes the control necessary for accumulation ever more difficult. 

As such we have good reason to suspect that markets are not machines for generating inequality and devouring anything that isn’t profitable, but rather are an integral part of a free society. To be sure, we cannot know this with absolute certainty ahead of time. But the evidence I’ve put forward should entail at worst a critical view towards markets, not dismissing them out of hand.

Getting over the outright rejection of markets and moving towards a discourse in which markets were seen as a tool for economic coordination (i.e. something with a lot of power but potential downsides) would be a fantastic development because it would open up a far more interesting set of conversations. Instead of simple arguments about whether markets will inevitably lead us right back to capitalism or not. More communistically minded folk could still retain a highly critical attitude, but could instead talk about how unexamined tendencies within markets that might result in pathological inequality or how strategy around how we go about engaging with capitalist markets without becoming corrupted. The critical eye of communists can be of significant value in ferreting out concerning aspects or tendencies of markets that more market friendly anarchists shy away from or have  overlooked. Going forward, this would make for a far more healthy and productive discourse than rehashing a century old argument that helps obscure some of the most important political questions for our world today.

  1. See Carlos Maldonado, Anarchism and Complexity.
  2. The actual story is no doubt considerably more complicated than the simple tale I lay out here. See Johanna Bockman’s Markets in the Name of Socialism: The Left Wing Origins of Neoliberalism for an in-depth survey of pro-market leftists in the 20th century.
  3. I do not personally believe that markets freed from state violence are worse in terms of externalities or the values they promote than other economic systems, but this is besides the point for my overall argument and as such I see no need to contest it.
  4. See Greg Gigerenzer, Fast and Frugal Heuristics: The Tools of Bounded Rationality for human reasoning. For artificial reasoning look at the intractability of Bayes theorem and heuristic based solutions see Paul Dagum and R. Martin Chavez, Approximating probabilistic inference in Bayesian belief networks is NP-hard. See also John Salvatier’s blog post Reality has a surprising amount of detail for a simple summary of what I’m trying to get at.
  5. The leftist economist Yanis Varoufakis’ Why Valve? blogpost is the best summary I’ve come across of Valve’s “anarcho-syndicalist” structure
  6. Eric D. Beinhocker, The Origin of Wealth.
  7. Gregoire Nicolis and Illya Prigogrie, Exploring Complexity, pg 233.
  8. Scott Page, The Diversity Bonus, pg 43.
  9. Michel Piore and Charles Sabel, The Second Industrial Divide, pg 265-266.

Mutual Exchange is C4SS’s goal in two senses: We favor a society rooted in peaceful, voluntary cooperation, and we seek to foster understanding through ongoing dialogue. Mutual Exchange will provide opportunities for conversation about issues that matter to C4SS’s audience.

Online symposiums will include essays by a diverse range of writers presenting and debating their views on a variety of interrelated and overlapping topics, tied together by the overarching monthly theme. C4SS is extremely interested in feedback from our readers. Suggestions and comments are enthusiastically encouraged. If you’re interested in proposing topics and/or authors for our program to pursue, or if you’re interested in participating yourself, please email or

Anarchy and Democracy
Fighting Fascism
Markets Not Capitalism
The Anatomy of Escape
Organization Theory