If you ever visit Rome, and wander through the Colosseum or Circus Maximus, it’s hard not to be struck by a sense of fragility and impermanence. Here are the remnants of the most powerful and complex of ancient European societies, now reduced to ruin and rubble. How did this once proud and mighty empire crumble?
Joseph Tainter’s 1988 book The Collapse of Complex Societies has an answer to that question, and to similar questions you might ask about the collapse of other ancient societies (Mayan, Incan, Babylonian etc). His book is widely cited and discussed among those who are interested in the topic of civilisational collapse. Having now read it, I can see why. Tainter presents his views with a logical simplicity that is often lacking in these debates, only setting out his own theory after having exhaustively categorised and dismissed alternative theories. What’s more, his own theory is remarkably easy to state and understand: societies collapse when they hit a point of rapidly declining marginal returns on their investments in problem-solving capacity.
Despite this, I have yet to see a really good summary of his theory online. I want to provide one in this post. I’ll try to focus on the essential elements of Tainter’s theory, and not on his dismissal of rival theories or his detailed case studies. I’ll also aim to be critical of the theory where needed, and to provide some reflections on whether it can be applied to contemporary societies at the end. These reflections will be somewhat idiosyncratic, tied to my own interests in democratic legitimacy and technology.
1. What does Tainter Mean by the ‘Collapse’ of ‘Complex Societies’?
Tainter’s theory is premised on the assumption that societies do indeed collapse. That’s the motivation for his entire book: the fact that he thinks is in need of explanation. You might suppose that this is a relatively uncontroversial assumption. Tainter starts his book with a comprehensive list of examples of societies that have collapsed. They include: the Western Chou Empire, The Egyptian Old Kingdom, The Hittite Empire, the Western Roman Empire, and Lowland Classic Maya. What could be more conclusive than that? And yet, despite its seeming obviousness, the notion that societies do undergo something akin to ‘collapse’ is actually controversial among scholars of the ancient world.
I’ll give examples of this later on, but we can see why this controversy might arise by considering some of Tainter’s definitions. His book is about the ‘collapse’ of ‘complex societies’. So it bears asking: what does he mean by the words ‘collapse’ and ‘complex society’? Fortunately, he provides some answers. He defines the terms in the following way:
Collapse = A society can be said to collapse when it undergoes a rapid and substantial loss of an established level of socio-political complexity. This, according to Tainter, is always a political process. It stems from the destruction and decay of social organisations and institutions. He gives a list of the kinds of things you can expect to see less of in a society undergoing collapse. These include: less social stratification and differentiation, less economic specialisation, less centralised control, less trading and economic activity and less production of ‘cultural epiphenomena’ such as monuments, buildings, and artworks (Tainter 1988, 4)
Complex Society = A society is complex when it has a lot of heterogeneity, specialisation and centralised control. In other words, when there are many different institutions, organisations, social functions and social roles, bound and integrated together into some centralised government. Complexity, according to Tainter, is not an all or nothing proposition. It comes in degrees and the precise delineation between a complex society and a ‘simple’ society is disputable. The creation of the ‘state’ is one of the key steps in the development of complexity, but when exactly a society transitions from being a band or tribe to a state can be tricky to pin down. It follows from this, incidentally, that the loss of social complexity also comes in degrees. Thus, collapse, like complexity, is itself a ‘continuous variable’ (Tainter 1988, chapter 2).
How exactly do complex societies emerge? Tainter spends some time discussing this. I won’t rehash the discussion here. There are really only two key points that emerge from it — at least from the perspective of understanding his theory. The first is that the emergence of complex societies has something to do with problem-solving. Complex societies emerge and sustain themselves by bringing benefits to their members, i.e. by addressing and overcoming barriers to their existential and psychological needs. This doesn’t mean that the benefits are evenly shared. Quite the contrary. Complex societies are often marked by profound inequality and hierarchy: some people benefit much more than others.
The second point is that, once they have emerged, complex societies are always threatened by some degree of internal dissent and rebellion (e.g. some people who feel that they aren’t getting the most out of the problem-solving bargain). They have two basic strategies for dealing with these threats: (i) boosting their perceived legitimacy or (ii) coercion and control. The latter requires no explanation. The former is a somewhat subtle idea. One way to boost legitimacy is to increase problem-solving capacity (and share the benefits to the dissenters). The other way is to inculcate the belief that the institutions of the state are sacred in either a religious (divinely mandated, righteous etc) or non-religious (just, moral, democratic) sense. A combination of both is probably required for legitimacy.
That’s enough by way of stage-setting. The bottom line is this: Tainter’s theory tries to explain the collapse of complex societies, where complexity is a function of the institutional heterogeneity of a society, and where collapse arises when there is a rapid and substantial decline in this complexity.
2. The Declining Marginal Returns Theory
Now that we know what it is that Tainter is trying to explain, we can turn our attention to the explanation itself. In chapter 4 of The Collapse of Complex Societies he sets out his explanatory theory. In chapter 5, he applies it to three historical case studies: the collapse of the Western Roman Empire, the collapse of classical Mayan civilisation, and the collapse of the Chacoan society (a North American society based around modern-day New Mexico). I’ll be skipping the case studies to focus on the theory itself. I do, however, recommend looking at them to get the full sense of how his theory works. You can do so in this article or this video lecture
(P1): Human societies are problem-solving organisations that generate benefits (B) for their members and, in order to develop and sustain themselves, they must continue to solve problems and generate benefits.
(P2): Like all organisations, human societies must capture and expend energy (C) in order to sustain their problem-solving capacity (classically, societies have captured energy by foraging, farming, burning fossil fuels, and also through war and imperial expansion)
(P3): Therefore, in order to sustain themselves, societies face a basic cost-benefit equation: the benefits of increased energy expenditure on problem-solving capacity must exceed the costs (i.e. B must be > C)
(P4) Increased investment in socio-political complexity usually yields higher returns (B ↑) but this usually comes at an increasing per capita cost so that, at a certain point, the marginal benefits (mB) on increased investment are outweighed by the marginal costs (mC).
(P5) If mC > mB at an increasing rate, societies will collapse (i.e. experience a rapid and significant decline in socio-political complexity).
The first three propositions are conceptual in nature. Indeed, they almost have the quality of truisms or platitudes. I don’t see how P2 could be denied: it is like a natural/physical law. P1 might be questionable, but it looks basically right to me. One thing you might worry about is whether problem-solving capacity alone is what sustains a society. As highlighted earlier, societies can also sustain themselves by boosting perceived legitimacy and through coercion. So maybe, at a certain point in their evolution, societies don’t need to keep solving problems in order to survive. That, however, seems unlikely and when you think about it a bit more both strategies really end up confirming the central importance of problem-solving capacity. After all, legitimacy is, at least in part, determined by problem solving capacity and coercion is itself a costly, problem-solving exercise and thus confronts the same basic cost-benefit equation alluded to in P3.
It’s really P4 and P5 that are the heart of Tainter’s theory and most of his attention in chapter 4 is dedicated to proving that P4 is true, i.e. that complex social organisations really do confront declining marginal benefits (chapter 5 looks to whether P5 is confirmed by the historical record). Tainter gives four examples of P4 playing out in human societies.
The first example has to do with agricultural production. Using work done originally by Ester Boserup, Tainter documents how (prior to mechanisation) agricultural intensification resulted in declining marginal benefits (i.e. declining marginal yields). People in foraging societies usually don’t have to work all that hard to secure their basic calorie needs. We see this by looking at contemporary equivalents where as little as 2.5 hours of work per day per person is all that is required. Contrast that with non-mechanised agricultural societies. Back-breaking work from dawn til dusk is often required and yet it generates diminishing marginal returns per unit of labour. Tainter uses a case study from farming communities in Northern Greece to illustrate this. The data from this case study suggests that “labor applied at an annual rate of about 200 hours per hectare is roughly 15 times more productive (in returns per hour of labor) than labor applied at 2000 hours per hectare” (Tainter 2000, 11).
The second example has to do with knowledge production and information processing power (broadly speaking the “intelligence” of human society). Tainter uses lots of examples to illustrate this problem, all broadly having to do with the returns on education and research and development. He documents how the early years of education generate far greater returns than the later years, and how most industrial societies seem to be undergoing a decline in their rates of innovation. This is actually something that has become better documented since Tainter wrote his book. A recent paper from Nicholas Broome and colleagues, for example, suggests that we are currently undergoing a significant marginal decline in the returns on research and development spending. Exponentially more money is being invested in the ‘education-research-industrial’-complex in order to sustain existing levels of growth. Michael Nielsen and Patrick Collison also have a nice article summarising this problem that looks at several different indicators of marginal decline. None of this should be surprising. There is a ‘low hanging fruit’ principle at play in most fields of research and development: the easy solutions/theories/insights are developed first and then it becomes progressively more difficult.
The third example has to do with bureaucratic power and control. Tainter gives several well-documented examples of administrative bloat and mission creep in bureaucratic organisations, particularly the military and Navy. These organisations are often central to societal problem-solving, but they tend to proliferate (i.e. more and more specialised organisations are developed to deal with discrete problems) and grow in size to cope with the challenges. This usually results in declining returns as more costly administrative staff are hired to manage the complex organisations compared to relatively fewer front-line workers/soldiers/doctors etc. We see this problem of administrative bloat and declining returns crop up over and over again in human history. Modern universities are a good illustration.
The fourth example has to do with declining rates of overall economic productivity. As societies increase in complexity they might initially generate high rates of economic growth but, over time, as they grow more complex, they tend to sustain relatively inferior rates of economic growth and have to invest more money, time and effort to sustain these rates. This is often linked to the problem of declining returns on education and research. Indeed, this is something that Nicholas Broome and his colleagues document quite well in their paper on declining rates of innovation.
Some people might complain that these examples are cherry-picked and that not all societies confront the problem of declining marginal returns. Tainter responds by arguing that this doesn’t matter. As he puts it:
The fact that such examples can be compiled does not mean that all economic trends follow the same curve, nor that socioeconomic processes in complex societies follow only the law of diminishing returns. It also does not mean that such trends occurring in any specific sphere are irreversible. Where this law is operative, however, serious consequences can be expected, and those are the situations of interest here.
Comments like this provide some reassurance, particularly when it comes to assessing our own era. Maybe we can buck the trend? Still, Tainter’s point seems to remain valid. Complex societies can run into the problem of declining marginal returns and when they do the prospect of collapse looms large. He sums it up by drawing a curve that he thinks maps out the common relationship between the benefits and costs of increased, and captures the dilemma confronted by all societies that must invest in increased complexity.
This curve is not based on any data. It is simply a mental model that we can use to understand the process of collapse. The idea is that societies often do well in the early stages of complexity but then they peak (at point B2, C2 in the diagram above) and enter a state of decline. At this point, things are often still getting better, but at a declining rate and if this rate of decline speeds up, the overall benefits of complexity will be outweighed by their costs. Societies will have to spend more resources to ensure stability, either through boosting legitimacy or increased coercion, but these solutions are often only temporary, stop-gaps. At a certain point, the decline becomes unstoppable.
That, in a nutshell, is Tainter’s theory.
3. Criticisms of Tainter’s Theory
There are many ways to criticise Tainter’s theory. The most obvious would be to challenge its empirical bona fides. As Tainter himself notes, it’s very hard to prove or confirm (or even falsify) a theory of collapse. We can’t run controlled experimental tests. We are forced, instead, to interpret the historical evidence and see whether it fits the theory. This is always a problematic exercise as people are inclined to twist or exclude facts in order to promote their preferred theory (they may do this unintentionally or without malice). That said, there have been some attempts to develop computerised models of human societies that can test these different theories. I know of at least one attempt to do this with Tainter’s theory which suggests that it can explain societal collapse. But I suspect that these model-building exercises are also fraught with difficulty, particularly when it comes to deciding what goes into the model.
More generally, and perhaps more seriously, it is worth noting that there are archaeologists and historians who doubt the very idea of societal collapse. I mentioned this earlier. After Jared Diamond wrote his best-selling book Collapse — which is somewhat similar to Tainter’s in that it tries to document and explain historical examples of societal collapse — a group of scholars got together to write a detailed rebuttal of the examples he used. They argued that human societies are actually remarkably resilient: they shift, evolve and adapt rather than collapse. This is a theme taken up by Guy Middleton in his 2011 book Understanding Collapse which reviews all the infamous examples of societal collapse and suggests that the evidence for actual collapse in each case is limited: aspects of a given society might decline, certain buildings or settlements might be abandoned, and particular regimes/states might be overthrown, but this does not mean that a civilisation as a whole has collapsed. Middleton wrote a short summary of his book’s thesis for Aeon magazine that you can read online.
Tainter might be able to escape these criticisms by arguing that, for him, collapse is a matter of degree. He is not claiming that societies disappear or self-destruct. He is claiming that they experience rapid declines in complexity. But this raises another problem. There is a certain ‘fuzziness’ to each of the concepts Tainter uses to support his theory and that makes it a bit too easy to massage historical data to fit the theoretical parameters. One example of this is the idea that collapse arises from a ‘rapid’ and ‘substantial’ decline in socio-political complexity. But what is a rapid and substantial decline? Some of the examples Tainter uses involve societies that changed over the course of centuries. Is that ‘rapid’? I’m not so sure. Furthermore, by zooming in on particular examples of apparent decline in complexity, there is a sense in which the theory misses the bigger picture, namely: that human civilisation really seems to be defined by a remarkable increase in socio-political complexity over time. Sure, there have been pockets of decline here and there, but in the long sweep of history things have gotten much more complex.
This brings me to what I think is the most important and interesting criticism of the theory. Remember, Tainter claims that society’s woes are caused by a decline in the marginal return in problem-solving ability. You might argue that this can be averted by some new technological innovation — one that boosts society’s problem solving ability. This technological breakthrough would allow society to shift to a new cost-benefit curve and avoid the problem of declining returns. A classic example would be the shift from agricultural society to industrial society. This was made possible by a series of technological breakthroughs that enabled us to more effectively harness the energy from fossil fuels (initially, then other power sources were discovered). This boosted Western European society’s problem-solving ability by massively increasing its energy capture, as well as the efficiency with which it could utilise that energy. This is something documented at length in Ian Morris’s work on societal development.
Tainter is aware of examples like this. Indeed, he discusses the industrial revolution in some of his work, noting that it was, to some extent, a lucky breakthrough because English society was encountering the problem of declining marginal returns after its shift from charcoal-burning to coal-burning (Tainter 2000). The steam engine came at just the right time. Furthermore, and more generally, he argues that even if a technological breakthrough is made it is often a temporary solution at best. Ultimately, the problem of declining marginal returns will be encountered by this technological breakthrough.
I wonder about this though. It puts me in mind of debates about ‘peak oil’ and other about finite resources. These debates rely on the same, basic logic as Tainter’s theory. They all assume that we are overly dependent on a particular resource (oil, coal, gas etc) and that this resource is finite. Granting this finitude, we are eventually going to ‘peak’ in our exploitation of the resource and enter a period of terminal decline. This logic is ineluctable: if the resource in question is finite it simply must be true that if we continue to exploit it we will reach a point of decline. But the problem is that, unless we actually know what the full stock of the relevant resource is, we don’t know where we are on the relevant exploitation curve. In his book The Wizard and the Prophet, Charles Mann points out that people have repeatedly claimed that we have reached (or are about to reach) ‘peak oil’. This has been true since the 1800s. These people have been repeatedly been proved wrong. We are producing and consuming more oil today than ever before. We don’t know where we are on the exploitation curve. It seems to me that Tainter’s theory encounters the same basic problem: maybe all innovations in problem-solving ability will eventually encounter the problem of declining returns. But how confident are we at our ability to predict when this might be? Furthermore, there are more optimistic views out there. The economist Julian Simons, for example, famously argued that human problem-solving ability is an infinite resource, not a finite one. So the ineluctable logic of the ‘peak oil’ argument may not apply.
That said, I’m not pollyannish about the prospects of technological solutions to our problems. Technological innovations might help us to avoid collapse, but they themselves often create new problems that need to be solved. This is not something that Tainter makes much of in his 1988 book, but it is something that greatly vexes contemporary researchers, particularly those focused on the topic of ‘existential risk’. Nick Bostrom’s paper on the ‘The Vulnerable World Hypothesis’ neatly summarises these fears, suggesting that technology can both solve our problems and make our world more vulnerable to existential collapse.
4. Conclusion: Whither Humanity?
All of which leads to the inevitable question: how do Tainter’s theories apply to the present day? After all, that’s really what motivates the historical inquiry, isn’t it? Perhaps some of us are interested in understanding the decline of the Roman Empire for its own sake, but lurking in the back of most people’s minds is the fear that our own civilisation is about to collapse.
Well, as you might imagine given my previous remarks, I’m somewhat sceptical about the wisdom of asking this question. The logical simplicity of Tainter’s theory, combined with its fuzziness, makes it both tempting and dangerous to apply it to our present day. Certainly, there are plenty of warnings out there about the current limits in our problem-solving capacity and the environmental challenges we face in the coming decades (Charles Mann’s aforementioned book The Wizard and the Prophet is an excellent primer on these challenges). But it’s hard to say where exactly we are on Tainter’s cost-benefit curve, or whether we might shift to a new curve with some technological breakthrough. Nevertheless, I do want to end with two reflections on the application of Tainter’s theory to the modern era.
The first has to do with democracy. Democratic institutions of governance are currently the most favoured problem-solving system in the world. And yet many people perceive them to have entered a state of stagnation and decline. This is a theme taken up in David Runciman’s book How Democracies End (far better than the more hyped American equivalent How Democracies Die), which argues that mature democracies have entered a ‘mid-life’ crisis. A democracy’s primary tools for increasing its problem-solving capacity and its perceived legitimacy is to enlarge the voting franchise and increase transparency and accountability. But many mature democracies have hit the point of declining marginal returns on these mechanisms. They may very well be at Tainter’s inflection point.
This brings me to the second reflection, which has to do with the role of technology, particularly artificial intelligence, in boosting our problem-solving capacity. According to some of its advocates, a significant breakthrough in AI could be the deus ex machina we need to solve our emerging problems. Instead of relying on imperfect, squabbling human intelligence we could rely on a more perfected artificial form. This is an argument that Miles Brundage makes explicit in his paper on the case for ‘conditional optimism’ about AI, and is implicit in other work on AI and algorithmic governance. The problem with this, however, is that technological solutions of this sort suffer from a significant legitimacy crisis. We see this everyday in scandals around tech companies. It could be that we face a hard tradeoff in the not-too-distant future: either we have legitimate problem-solving institutions that face the problem of rapidly declining marginal returns or we have fast, efficient and technologically-mediated institutions that suffer from a legitimacy deficit. Although I did not express it in these terms, this was effectively the argument in my paper “The Threat of Algocracy”.