Seshat: What 10,000 Years of Data Actually Shows

In 2010, a mathematical biologist named Peter Turchin made a prediction.

Based on patterns he'd identified in historical data, he said America would experience major political instability around 2020. Not because of any specific policy or leader—not because of who would be president—but because of structural dynamics that recur across civilizations when certain conditions align.

He didn't know about COVID. He didn't know about George Floyd. He didn't know about January 6th. He predicted the timing and type of instability purely from historical patterns—the same patterns that preceded civil wars in Rome, revolution in France, and dynastic collapse in China.

The prediction aged well.

Turchin isn't a mystic or a Nostradamus figure making vague prophecies. He's a population ecologist who turned his mathematical tools to human history and asked: Does history have structure? Can we find the regularities? Can we predict?

To answer that, he built the largest database of quantified historical information ever assembled. It's called Seshat, after the Egyptian goddess of writing and measurement. Over 200,000 data points covering 500+ political entities across 10,000 years. Population, territory, government hierarchy, military organization, taxation, religion—all coded, standardized, and ready for statistical analysis.

When you crunch the numbers, patterns emerge. Some of them are very uncomfortable.

The Cliodynamics Project

Turchin coined the term cliodynamics—the application of mathematical models and statistical methods to historical data. Clio was the Greek muse of history. The "-dynamics" part signals that history might follow laws, like physics does.

This was controversial. It still is. Historians generally resist the idea that history has laws. Every situation is unique, they argue. Context is everything. The French Revolution happened because of the French Revolution's specific causes—not because it fit some universal pattern.

Turchin's response: fine, let's test it. If every situation is unique, then we shouldn't find patterns. But if patterns exist—if the same variables predict instability across different centuries and cultures—then history does have structure, and we can potentially identify it in advance.

His background made him well-suited to this project. In ecology, you study how populations rise and fall, how predator-prey dynamics cycle, how systems reach equilibria and then destabilize. The math isn't simple, but the principle is: complex systems have regularities. You can find them with data.

Why would human societies be any different?

What Seshat Measures

The database codes societies across dozens of dimensions:

Social scale. Population. Territory. Hierarchical levels of government. Capital city size. Number of administrative levels between ruler and commoner. How big is the society, and how elaborately organized?

Information systems. Does the society have writing? Record-keeping? Calendars? Standardized weights and measures? How much does the society know about itself?

Economics. Currency. Market integration. Taxation mechanisms. Trade networks. How is value created, moved, and extracted?

Military. Army size. Professional soldiers vs. citizen militia. Fortifications. Weapons technology. How does the society project force and defend itself?

Religion. Doctrinal beliefs. Priestly hierarchies. Supernatural punishment for moral violations. Does the society have moralizing gods who care whether you lie and cheat?

Governance. Legal codes. Bureaucratic structures. Merit-based selection for office. Specialized administrative roles. How does the society coordinate collective action?

Each variable is coded according to careful protocols. Multiple experts review controversial cases. Ancient Rome and Han China and the Aztec Triple Alliance and medieval France all go into the same framework, measured by the same criteria.

Is this comparing apples to oranges? Maybe. But if you find that the same variables predict the same outcomes across societies as different as Rome and China and the Aztecs—that's telling you something about the underlying structure of complex societies.

The Patterns

What emerges from 10,000 years of data?

Secular cycles. Political stability and instability follow cycles of roughly 2-3 centuries. Turchin calls these "secular cycles"—secular meaning long-term, not nonreligious.

The cycle goes like this: An integrative phase—population growing, economy expanding, elites cooperating, institutions functioning. Then a disintegrative phase—population pressure, elite competition, institutional breakdown, instability and violence. Then a reset—often through war or revolution—and the cycle begins again.

The integrative phases can last a century or more. The disintegrative phases are shorter but more dramatic. And the timing is remarkably consistent across very different societies.

Elite overproduction. This is Turchin's signature finding and the driver of the disintegrative phase.

During good times, elites multiply. More elite children survive to adulthood. More families invest in their children's education. More people acquire the credentials and connections that mark elite status. More people expect high-status positions commensurate with their training and ambitions.

But economic growth eventually slows. The number of elite positions doesn't keep pace with the number of elite aspirants. The ratio of would-be elites to actual positions worsens.

What do you get when you have more trained, ambitious, entitled people than the system can absorb? Frustration. Radicalization. Counter-elites who turn against the system that failed to reward them. Political entrepreneurs who mobilize discontent. Polarization. Violence.

Turchin found this pattern preceding virtually every major period of instability in the historical record. Late Republican Rome. Pre-revolutionary France. Pre-Civil War America. Antebellum China before dynastic collapse. When you produce more credentialed aspirants than your system can absorb, something breaks.

Warfare builds states. One of Seshat's most robust findings: intense inter-group warfare strongly predicts the emergence of complex political organization.

This sounds backwards. War is destructive. Shouldn't peace build states?

But the data says otherwise. The societies that developed elaborate bureaucracies, standing armies, and large-scale coordination mechanisms were the ones under existential threat. Compete or die. Organize or be conquered. The crucible of warfare selected for institutional complexity.

This doesn't mean war is good. It means that states—complex, hierarchical, powerful—are in some sense war machines. They emerged to fight and survive. That origin shapes what they are.

Moralizing gods scale with society. Small-scale societies have spirits and ancestors who don't care much about human morality. They might help you or harm you—but not because you cheated or lied. Morality isn't their domain.

Large-scale societies—ones that need to coordinate strangers who will never meet face to face—develop gods who watch, judge, and punish. Belief in supernatural surveillance and moral enforcement correlates with social scale.

Why? Probably because large societies face cooperation problems that small societies don't. In a village of 100 people, everyone knows everyone. Reputation enforces morality. In a city of 100,000, you interact with strangers. Who enforces morality when no human is watching?

The gods, apparently. Societies that needed moral enforcement at scale developed religions that provided it.

Complexity ratchets. Once societies cross certain complexity thresholds, they rarely regress below them except through catastrophic collapse.

Writing, once invented, doesn't get un-invented. Professional bureaucracies, once established, don't dissolve back into village councils. Standing armies, once formed, don't return to seasonal militias. Complexity accumulates, ratchets up, and only declines through systemic breakdown.

This connects to Tainter's framework. Complexity accumulates because it's hard to remove. The ratchet means societies tend toward ever-greater elaboration—until they hit constraints (like EROI decline) that force simplification.

The 2020 Prediction

Let's return to that 2010 prediction, because it's the test case for whether cliodynamics actually works.

Turchin's argument went like this:

America in the late 20th century massively expanded higher education. College enrollment soared. Graduate and professional degrees multiplied. The society produced an enormous cohort of people with credentials and expectations—people who thought of themselves as elite, or at least elite-adjacent.

But the economy didn't create commensurate opportunities. Wages for most jobs stagnated after 1970. The cost of the markers of middle-class success—housing, healthcare, education—exploded. The positions that matched elite expectations became brutally competitive.

The result: a surplus of frustrated elites. Lawyers working as baristas. PhDs driving Ubers. MBAs in entry-level positions. All of them trained for success, expecting success, not receiving success. All of them increasingly angry at a system that promised them outcomes it didn't deliver.

These frustrated elites, Turchin argued, would become counter-elites. Some would radicalize left. Some would radicalize right. All would polarize politics. All would attack institutional legitimacy. All would drive conflict.

The historical pattern suggested crisis around 2020. Not because Turchin knew what specific events would occur, but because the structural conditions matched what had preceded instability in every previous case he studied.

Then 2020 happened. Pandemic. Protests. Contested election. Capital stormed. Trust collapsed. Polarization maximal.

Was this confirmation of the pattern, or coincidence? You can argue either way. But the prediction was specific (around 2020), the reasoning was explicit (elite overproduction following historical patterns), and the outcome matched.

The Controversies

Seshat has critics. Serious ones, with serious concerns.

Data quality. Historical data is uncertain. Can you really code "hierarchical levels of government" consistently across Han China and the Aztec Triple Alliance? The measurements might be too noisy to support statistical claims.

Variable selection. What you find depends on what you measure. Seshat's categories might impose modern Western concepts on diverse societies. The framework might create patterns rather than discover them.

Correlation isn't causation. Seshat finds correlations. The causal interpretations are contestable. Does warfare create states, or do proto-states make warfare more organized and therefore more visible in the record? The data shows association, not mechanism.

Turchin's priors. Much of Seshat's framing comes from Turchin's specific theoretical commitments. Is the project testing hypotheses or confirming what he already believed?

These concerns are legitimate. Quantitative history is young. The methods are evolving. The findings should be held provisionally.

But the alternative—not attempting quantification at all, relying solely on narrative—has its own limitations. Stories can support any conclusion. Data, even imperfect data, creates the possibility of being wrong. You can check your claims against evidence. You can make predictions and see if they hold.

And the 2020 prediction was right.

What the Data Suggests Now

Turchin's 2023 book End Times applies the Seshat framework to the present moment.

His diagnosis: America is deep in a disintegrative phase. Elite overproduction is severe and intensifying. Political polarization is extreme. Institutional legitimacy is collapsing. Trust between groups is at historic lows.

The indicators match what preceded instability in Rome before the civil wars. In France before the Revolution. In England before the Civil War. In China before dynastic collapse.

His prediction: the instability will likely continue and potentially intensify through the 2020s. Structural conditions have momentum. They don't change quickly. The cycle has to run its course—or be interrupted by deliberate action.

His prescription: reduce inequality, absorb surplus elites through productive channels (new institutions, frontier expansion, whatever works), rebuild cooperation across factional lines. The pattern can be interrupted—societies have pulled back from the brink before—but it requires recognizing the structural nature of the problem rather than attributing everything to the other side's wickedness.

Critics argue Turchin is too confident. That America is unique. That the 2020 prediction was lucky or self-fulfilling or not actually that specific.

Maybe. But here's the thing: Turchin has made other predictions too. He's predicted continued instability through the 2020s. We'll see if that holds. He's predicted that without structural intervention—actual changes to elite pathways, actual reduction in inequality—the disintegrative dynamics will continue.

The scientific method works by making predictions and checking them. Turchin is doing that. You can evaluate his track record in real time.

What Cliodynamics Gets You

Even if the specific predictions prove wrong, Seshat offers something valuable: a different way of thinking about why societies fail.

The conventional explanation for instability is moral. One side is evil. One side is destroying the country. If only the right people were in charge, things would be fine.

The cliodynamic explanation is structural. Elite overproduction happens regardless of who's in office. Secular cycles turn regardless of which party controls Congress. The dynamics emerge from the interaction of variables that no single actor controls.

This is simultaneously more pessimistic and more optimistic than the moral explanation.

More pessimistic: you can't solve the problem by defeating the bad guys. The bad guys—on both sides—are products of structural conditions. New bad guys will emerge as long as the conditions persist.

More optimistic: if the problem is structural, structural solutions might work. You don't need to convert people's hearts. You need to change the incentive gradients. Create opportunities for surplus elites. Reduce inequality. Rebuild cooperation mechanisms.

None of this is easy. But it's at least conceivable in a way that "make the other side stop being evil" is not.

The Fundamental Question

The deepest question Seshat raises isn't about Rome or 2020. It's about history itself.

Does history have structure? Are there regularities we can discover? Or is every situation unique, every context unrepeatable, every prediction impossible?

Seshat's answer—provisional, contested, but increasingly supported by data—is yes. History has structure. Patterns recur. Variables predict outcomes across times and places that had no contact with each other.

If that's true, then the past isn't just a collection of stories. It's a dataset. And datasets can teach you things you didn't know.

Maybe—just maybe—in time to matter.

Further Reading

- Turchin, P. (2023). End Times: Elites, Counter-Elites, and the Path of Political Disintegration. Penguin. - Turchin, P. et al. (2018). "Quantitative historical analysis uncovers a single dimension of complexity that structures global variation in human social organization." PNAS. - Seshat Databank: http://seshatdatabank.info/

This is Part 5 of the Collapse Science series. Next: "Resilience Theory: The Adaptive Cycle."