16.2 — Why Civilizations Collapse

Southern Iraq was the most advanced, most productive and most literate place on Earth for over a thousand years.

It is now a salt flat with mounds in it (Chapter 4.1).

The starting observation of this chapter is that being the centre of the world confers no protection, and the useful question is what the collapses actually have in common.

What collapse means

Joseph Tainter's definition is the working one: a rapid, significant loss of an established level of sociopolitical complexity.

Which is deliberately narrow, and the narrowness matters.

Collapse is not the disappearance of a population. The Maya did not vanish; millions of Maya speakers live in the same region today (Chapter 4.5). The Indus people did not vanish; they de-urbanised and moved (Chapter 4.3). Rome's western provinces continued to be inhabited by the same people speaking evolved versions of the same language (Chapter 5.5).

Collapse is the loss of the structure — the administration, the long-distance trade, the writing, the monumental building, the specialists.

And it is frequently better for ordinary people than the alternative would have been. A peasant paying tribute to a palace that then stops existing may be materially better off. The historian's sense of loss is partly a sense of loss of evidence: complexity leaves records and simplicity does not.

And the qualification matters too. Bryan Ward-Perkins's evidence on post-Roman western Europe (Chapter 5.5) shows measurable material decline — worse pottery, smaller cattle, less trade, less literacy — for centuries. So collapse is not always liberation.

The cases

Six, from this volume.

The Bronze Age system, around 1200 BCE (Chapter 4.6). Interconnected, specialised, no redundancy; hit by a multi-decade drought, probably earthquakes, and population movements; every state's failure removed a resource the others depended on.

The Western Roman Empire, over three centuries (Chapter 5.5). Fiscal exhaustion, plague, climate deterioration, loss of the richest tax provinces, and dependence on armed groups whose loyalty required payments it could no longer make.

The Classic Maya, around 800–900 CE (Chapter 4.5). Severe multi-decade droughts on a landscape already stressed by deforestation and soil erosion, with endemic warfare and elite competition for prestige.

The Indus civilization, from around 1900 BCE (Chapter 4.3). A river system that dried, a weakening monsoon, and a contracting trade network — with de-urbanisation rather than destruction.

The Soviet Union, 1991 (Chapter 12.6). Economic stagnation, an unaffordable military burden, a resource price collapse, an unwinnable war, and a legitimacy failure that was invisible until it was total.

And Easter Island, which deserves a paragraph because it is the most cited and the least reliable.

The standard story — that the inhabitants deforested their island to move statues, collapsed into warfare and cannibalism, and starved — has been substantially challenged. Current evidence suggests deforestation was substantially caused by rats introduced with the settlers eating palm seeds; that the population was managing a difficult environment reasonably; and that the catastrophic collapse followed European contact, disease and slave raiding. It is a good case study in how a morality tale gets attached to an archaeological site, and it should be retired from the standard list.

What they share

Four features appear in most of them.

Multiple stresses arriving together. No case in the list has a single cause. Each involves a climatic or environmental stress plus a political or fiscal weakness plus an external pressure, and the interaction is what does the damage.

A loss of the capacity to absorb shock. Complex societies handle problems by adding complexity — more administration, more specialists, more infrastructure — and each addition costs. Tainter's central argument is that the returns to added complexity diminish while the costs do not, and eventually a society is spending enormously to maintain a structure whose benefits no longer justify it. At that point, collapse can be the rational outcome — the population stops paying for a structure that is no longer worth its cost.

Elite competition and extraction increasing as the base weakens. This appears repeatedly: Maya kings building larger monuments during the drought; Roman elites obtaining tax exemptions as revenue fell; Mughal nobles extracting harder from smaller jagirs (Chapter 6.12).

Peter Turchin's structural-demographic work formalises it. His proposed mechanism is elite overproduction: when the number of people with elite aspirations exceeds the positions available, competition intensifies, factions form, and the resulting instability weakens the state. The theory is contested and it names a real and observable phenomenon.

And a failure of information. Chapter 16.1's eighth pattern. In almost every case, the problem was visible to someone and could not travel to where the decisions were made.

The single-cause explanations, tested

Environmental determinism — that societies collapse because they destroy their resource base. Jared Diamond's Collapse is the popular statement.

Its evidence is real — Mesopotamian salinisation (Chapter 3.7), Maya deforestation, Easter Island's palms — and it consistently understates the political dimension. Societies facing identical environmental stress respond differently, and the difference is institutional. The Dutch built a functioning state around water management; Mesopotamia did not solve salinisation.

Moral decline — the oldest explanation and not an explanation at all (Chapter 5.5). Every generation of conservative commentary has described its own society as decadent, including during periods of expansion.

Barbarian invasion — usually a symptom. Chapter 5.5 showed the Germanic peoples had been inside the empire for centuries as soldiers and settlers, and that the crisis was fiscal.

Climate — real and never sufficient. The 4.2-kiloyear event, the drought behind the Bronze Age collapse, the Maya droughts and the Late Antique Little Ice Age are all documented. And societies with resilient institutions survived the same events.

And plague — a genuine multiplier, never sufficient alone (Chapter 7.7).

The honest general finding: collapse is what happens when several stresses coincide in a system that has lost its slack, and no single-cause account survives the cases.

What prevented collapse

The more useful question and it is asked far less often.

China reassembled itself after every collapse (Chapter 5.6): a shared script across many spoken languages, a shared canon and examination system producing a governing class with common training, a doctrine of legitimacy that survives dynastic change, and a geography that rewards unity.

Byzantium survived when the western empire did not (Chapter 7.3): richer provinces, a shorter defensible frontier, and a city that could not be taken.

Japan preserved legitimacy separately from power (Chapter 7.9), which allowed it to change government repeatedly without changing its legitimating principle — and which is what made the Meiji transformation possible as a restoration rather than a revolution.

And the modern instances. Britain after 1945: an empire lost, an economy exhausted, and a state that adapted rather than broke. Germany and Japan after 1945, rebuilt by occupiers who had learned from 1919 (Chapter 12.1).

The common features of survival: a mechanism for transferring power without a crisis; institutions that outlive individuals; a way for bad information to reach decision-makers; and enough slack to absorb a shock.

Does it apply now

The uncomfortable section.

What is different about the present. The system is global rather than regional, so a collapse would not have an outside to recover from — which cuts both ways: no external refuge, and also enormous redundancy in knowledge and capability across continents. Knowledge is recorded and distributed at a scale no earlier civilisation approached. Agricultural productivity has margins earlier societies did not have. And the specific mechanism that ended several earlier cases — a regional drought destroying the food supply of a society with no ability to import — is much weaker in a world with global trade.

What is similar. Chapter 4.6's efficiency-fragility trade, at global scale. Environmental limits being approached simultaneously (Chapter 14.6). Elite competition and inequality rising in several large societies. And the specific failure mode of information: a media environment in which the incentive is attention rather than accuracy.

And what is genuinely new. Two capabilities that no earlier civilisation possessed: the ability to destroy itself deliberately in an afternoon (Chapter 11.10), and the ability to change the planet's climate as a side effect of its ordinary operation (Chapter 14.1).

The honest position. Modern global civilisation is far more robust than any earlier one to the things that destroyed earlier ones, and it has failure modes none of them had. Which means the historical pattern is a poor guide and it is the only guide available.

The useful conclusion

Not that collapse is coming, which is unknowable, and not that it cannot happen, which is contradicted by every case in this chapter.

What the cases actually support is narrower and more usable.

Maintain slack. A system optimised to the last percent has nothing left when conditions move.

Keep information channels open. The most consistent feature of failure is that somebody knew and could not be heard.

Watch the ratio of what elites extract to what the base produces.

And treat the institutions that transfer power peacefully as the load-bearing structure, because in every case in this chapter, once succession or legitimacy failed, everything else followed.

What the next page covers

Chapter 16.3 asks the positive version of the question: what makes a society rich, free and stable? It gathers the findings from every Part that touched it — the institutional evidence of Chapter 9.4, the East Asian sequence of Chapter 9.13, the Indian ledger of Chapter 6.28, and the political findings of Part 13 — and states what the evidence supports and where it runs out.