Why civilisations fall: from Polybius to Turchin

The question Gibbon left behind

Edward Gibbon closed his third volume in 1781 with a short essay, the General Observations. It handed the tradition its best-known line: the decline of Rome was "the natural and inevitable effect of immoderate greatness". Then he turned the question over. The surprise, he wrote, was that Rome "had subsisted so long". Falling is cheap. Holding an empire together for four centuries, against fiscal arithmetic, distance and its own generals, is the expensive part. A theory of decay that can't price the holding is a theory of nothing.

Two thousand years of argument about political decay sort into three families. The first says orders rot from inside and the rot is moral: character fails, institutions follow. The second says decay is structural, a matter of solidarity, energy and arithmetic. It would happen to a nation of saints. The third says the question is malformed, because "a civilisation" is not the sort of object that can fall. Polybius wrote in the second century BC. Peter Turchin published Ages of Discord in 2016. The distance between them is smaller than you might think.

"Prosperity ripened the principle of decay." Edward Gibbon, General Observations, 1781.

Polybius and the wheel of constitutions

Polybius gave decay a shape in Book VI of the Histories: anakyklosis politeion, the cycle of constitutions. Six forms follow one another in fixed order. Kingship (basileia) rots into tyranny, aristocracy into oligarchy, democracy into ochlokratia or mob rule, and out of the wreckage a strongman starts the round again. The engine is generational. Heirs raised inside inherited power never paid for it, so they spend it. His remedy is the politeia symmikte, the mixed constitution, which for him is Rome: consuls, senate and people jamming the wheel by checking each other.

Book VI is two texts welded together. F. W. Walbank's commentary, published at Oxford between 1957 and 1979, prises them apart: an inherited Greek cycle, Stoic in flavour, and an empirical account of Rome that doesn't fit it. The objection is older. Aristotle makes the same objection in Politics V.12, against the sequence in Republic VIII, where Plato's kallipolis degenerates through timocracy, oligarchy and democracy into tyranny. Each regime is produced by the soul-type that the last one bred. The order is neither historically nor logically necessary. Tyranny doesn't wait politely for its turn. Plato's scheme is beautiful. It also predicts nothing you could catch it getting wrong.

Sallust, Machiavelli, Ortega: the moral case at its best

Sallust's Bellum Catilinae, written around 42 BC, dates Rome's ruin to a victory. While Carthage stood, fear of a rival held citizens to virtus and hard work. He calls that fear metus hostilis in the Jugurtha. Carthage fell in 146 BC, and in his telling "fortune began to grow savage", and ambitio and avaritia moved in. Fear was the discipline. Ronald Syme's Sallust (1964) reads the whole scheme as inherited rhetoric, a Thucydidean and Catonian topos doing the work of an explanation. Syme also notes who is preaching: a senator whom the censors expelled from the Senate in 50 BC.

In the Discorsi, Machiavelli takes over the six regimes and throws out the fatalism. He wrote it between about 1513 and 1519; it was printed in 1531, four years after his death. A city left to itself would cycle through the forms until a neighbour swallowed it. So a founder mixes them on purpose. Lycurgus did it at a stroke in Sparta; Rome blundered into it. The scandal is chapter I.4, on the tumulti, the street fighting between plebs and the Senate. That quarrel, Machiavelli says, made Rome free and gave it its laws. Conflict is the medicine. Republics live by virtù and die of corruzione, and his cure is a periodic, violent return to first principles. That is a doctrine of permanent emergency, and it's the strongest argument I know against him.

The argument reappears in 1929. José Ortega y Gasset serialised La rebelión de las masas in the Madrid paper El Sol from October that year. The book followed in 1930. His figure is the hombre-masa, the mass man. He inherits a civilisation the way an heir inherits a house, using the plumbing with no idea who laid the pipes. His sharper name for the type is the señorito satisfecho, the spoilt young gentleman. That's Polybius' generational engine wearing a suit. As an explanation, it carries the old defect, a description of the outcome dressed up as its cause.

Ibn Khaldun and the half-life of asabiyya

Ibn Khaldun completed the Muqaddimah in 1377. It opens by announcing a new science, ilm al-umran, the science of civilisation. Its first principle is that royal authority (mulk) rests on asabiyya, the group feeling produced by kinship and shared hardship. Asabiyya runs strongest where life is hard, among tribes and desert peoples (badawa), and weakest in settled cities. Conquest follows that gradient. A group that still has asabiyya takes the state from a dynasty that has spent its asabiyya, then settles and acquires luxury (taraf). Its own solidarity drains away across roughly four generations. By then someone hungrier is at the gate.

What makes this structural rather than moralising is that nobody has to sin. Luxury is what winning produces, and softness is what security produces. Nobody had to lose their nerve. Asabiyya behaves like a quantity with a half-life. The strongest objection is Aziz al-Azmeh's, in Ibn Khaldun: An Essay in Reinterpretation (1982). The model universalises a Maghribi contest between tribe and city into a law of history. And Ibn Khaldun barely applies it in the narrative volumes of the Kitab al-Ibar, the work the Muqaddimah was written to introduce. That last point stings. A founder who won't use his own instrument on his own material has told you something about the instrument.

Tainter's arithmetic and Turchin's surplus elites

Joseph Tainter treated decay as an accounting problem in The Collapse of Complex Societies (1988). Societies solve problems by adding complexity: officials, tiers, records, garrisons, written procedure. Complexity costs energy. The cheap problems get solved first, so each new layer returns less than the last, until the marginal return on complexity flattens and turns negative. Collapse becomes rational at that point, a rapid, deliberate simplification by populations no longer getting value from the apparatus. His three cases are the Western Roman Empire, the Maya lowlands and Chaco Canyon.

Secular Cycles, written by Peter Turchin with Sergey Nefedov and published by Princeton in 2009, gives that mechanism a demographic model. Instability tracks a few measurable variables: popular wellbeing, state fiscal health, and above all elite overproduction, meaning more elite aspirants than there are elite positions to hold them. Ages of Discord (2016) runs the same machinery over American history since 1780. The surplus produces counter-elites. Counter-elites recruit the discontented, and states come apart through competition inside their own upper class rather than through luxury or cowardice. You can even count law degrees per judgeship.

The data are the soft spot. Seshat, the historical databank Turchin co-founded, came under heavy criticism after a 2019 Nature paper on moralising gods by Harvey Whitehouse and colleagues, Turchin among the authors. Edward Slingerland, M. Willis Monroe and a group of historians replied in the Journal of Cognitive Historiography. They reported that 61 per cent of the relevant data points were unknowns that had been recoded as absences. The authors rejected the criticism. Nature retracted the paper in July 2021. A research programme whose selling point is measurement cannot afford to be casual about how that measurement was done.

Spengler's seasons, and why I leave them alone

Oswald Spengler published the first volume of Der Untergang des Abendlandes in Vienna in 1918 and the second in Munich in 1922. Cultures, on his account, are organisms. Each has a soul, and each runs about a thousand years through spring, summer, autumn and winter. The terminal season is Zivilisation: Kultur hardened into form, its creative spring dry, expressed now in technique, money, world-cities and imperialism. The West is Faustian, defined by a craving for infinite space. Its winter, he says, brings Caesarism.

The scheme does not survive inspection. Otto Neurath's Anti-Spengler (Munich, 1921) went at the method. The correspondences between Egyptian, Classical, Arabian and Western "seasons" are asserted, never demonstrated, and the vocabulary is built so that no fact can embarrass it. Where the particulars can be checked, they are frequently and grossly wrong: the history of mathematics and his treatment of cultures whose languages he could not read. Adorno's essay Spengler after the Decline appeared in English in Studies in Philosophy and Social Science in 1941. It is the rare hostile reading that keeps part of the account of Caesarism and junks the biology.

Two books put his politics in the record: Preußentum und Sozialismus (1919) and Jahre der Entscheidung (1933). The Völkischer Beobachter attacked the second because Spengler had not saluted Hitler. He did reject Nazi race biology. None of that has stopped the radical right from quoting him, from the 1930s to now. Take it all seriously and the book still fails on its own terms. A morphology that accommodates every outcome predicts none. I don't use it. I'd say the same if his politics had been immaculate.

Burckhardt's centaur and the late-antiquity turn

Jacob Burckhardt lectured at Basel between 1868 and 1871 on what became Weltgeschichtliche Betrachtungen, printed posthumously in 1905. He rejected the genre outright. Philosophy of history, he says there, is "a centaur, a contradiction in terms", because history coordinates while philosophy imposes a hierarchy. He wanted the historian to study constants, the interactions of state, religion and culture, and to give up the plot. No grand arc. His target was Hegel, and behind Hegel the temptation everyone in this essay falls prey to: knowing the ending in advance.

The best-documented case came apart in the 1970s. Peter Brown's The World of Late Antiquity (1971) recast the years from AD 150 to 750 as a transformation, a stretch with its own religious and artistic life where Gibbon had seen ruin. Walter Goffart's Barbarians and Romans (Princeton, 1980) asked whether the fifth-century invasions were invasions at all, reading the settlements as an extension of Roman taxation. What replaced Gibbon's fall was a set of regional stories with different endings. Then Bryan Ward-Perkins published The Fall of Rome and the End of Civilization (2005) and pointed to the material record: pottery quality, roof tiles, cattle bones, and literacy. All of it drops hard in Britain and Italy. Calling that a transformation is a euphemism.

Where I come down, and what would move me

I come down with the structural party, and by a wide margin. My reason is unglamorous: moral accounts are written after the result is known, and the vices they name are just as visible in states on the rise. Sallust dates the rot to 146 BC. Rome then took Gaul, annexed Egypt in 30 BC, and reached its greatest extent under Trajan, who died in 117. That is two and a half centuries of expansion inside the alleged decline. A theory whose clock runs backwards against the power curve is doing something other than explaining.

Inside the structural camp I'd rank Ibn Khaldun and Tainter above Turchin, because their mechanisms are simpler and demand less of evidence we do not have. The strongest objection to my own side is that complexity and returns are almost never measured independently of the collapse they are called on to explain. That is Spengler's sin in a lab coat. Tainter's Roman chapter at least leans on countable things, coin debasement and the size of the army. Much of the rest doesn't. And Burckhardt keeps his force, because the unit itself, a civilisation, is our category rather than history's.

Four things would move me. A pre-registered out-of-sample test is the first. Fit Turchin's structural-demographic indicators to states before 1900, predict instability after 1950, and see what comes out. If the forecasts do no better than a trend line, elite overproduction goes back on the shelf. The second is series data. If coin fineness, tax yield per official, aqueduct maintenance and grain shipment volumes stay flat right up to several well-documented collapses, Tainter's curve is decoration rather than mechanism.

Show me a measurable index of elite behaviour: oath-breaking, desertion rates, defaults on public debt. If it moves first in several independent cases while the material series remain flat, character sits on the causal side of the causal arrow, where Sallust put it. That is the third test, and it would convert me. The fourth belongs to Burckhardt. A rigorous statistical analysis of a large set of collapse events, showing no shared structure across them at all, and I stop. Ward-Perkins counted roof tiles and cattle bones. That is the standard. Most of this tradition doesn't come close to it.