← What the Universe Wants
A page from What the Universe Wants — the ruin written in the ledger

Why Civilizations Fall

or, how a society solves itself to death — and why the simpler thing on the other side can feel like relief

Jump to the simulation: the society that keeps solving its problems

Hold a Roman coin from the year 270 and you are holding a confession. A denarius struck under Augustus, three centuries earlier, was a solid little disc of silver. The one from 270 is mostly copper with a thin silver wash over it, a skin of value the mint flashed on so the thing would still ring true in a market stall. Pull a few hundred of them out of the ground in sequence and you can watch the silver bleed away year by year, emperor by emperor, like a fever chart of a body that is quietly failing. Nobody in the mint woke up one morning and decided to wreck the money. Each debasement was a sensible patch on a real problem — an army to pay, a frontier to hold, a grain dole to keep a restless city fed. The coin did not fall because anyone willed it to fall. It fell the way the cod off Newfoundland fell, the way a building settles: a great many reasonable decisions, summing.

We tell the story of Rome’s end as an invasion — the Visigoths at the gates, the smoke, the barbarians who finally kicked in a rotten door. And the barbarians were real. But the barbarians are almost always real; there is always someone at the frontier. They are the event, not the cause. The interesting question is not why the door was kicked but why, after a thousand years, it had gotten so cheap to kick. For that you want less a battlefield historian and more an accountant. You want Joseph Tainter.

Tainter is an anthropologist who, in 1988, wrote a book called The Collapse of Complex Societies that did something the lurid “decadence and barbarians” stories never managed: it gave collapse a mechanism general enough to fit Rome and the Maya and the Ancestral Puebloans of Chaco Canyon all at once, with no need for any of them to have been wicked, or stupid, or unlucky in some special way. His account is almost dry, and that dryness is its power. Collapse, he argued, is not a moral event. It is an economic one. It is what happens when a particular kind of investment stops paying off.

The machine that solves problems

Start where Tainter starts, with a definition shorn of drama. A society has collapsed, he says, when it undergoes a rapid, significant loss of an established level of complexity. Not necessarily death — the people are mostly still there afterward — but a sudden simplification: fewer layers of administration, fewer specialists, smaller-scale everything, less of the elaborate apparatus that a moment before had seemed permanent. To understand why that apparatus comes down, you have to see what it was for in the first place.

A society, in Tainter’s eye, is a machine for solving problems. A drought, a rival on the border, a city that has outgrown its wells, a population that needs feeding — each problem gets met with a little more complexity: a new office, a new tier of officials, a longer aqueduct, a standing army where a militia used to do, a tax to pay for all of it. Complexity is the universal tool. It genuinely works; that is why societies reach for it again and again. The Romans, their own farmland thinning, solved the problem the direct way an empire can: they conquered the neighbors and took their surplus — the grain of Egypt, the silver of Spain, the bodies of the enslaved. Every conquest was a windfall, a great slug of someone else’s stored energy poured into the Roman account at once.

But here is the hinge the whole thing turns on, and it is so simple it is easy to walk past. Complexity has to be paid for, forever. A conquest is a one-time deposit; the province you took is a standing bill — roads to garrison, officials to post, a border now longer than it was. The windfall lands once; the upkeep is owed every year after. And societies, being sensible, solve their cheapest, highest-return problems first. The first aqueduct waters a thirsty city and pays for itself ten times over. The tenth aqueduct reaches a little farther for a little less. The hundredth is mostly maintenance. Each new layer of complexity buys a little less than the layer before it, while costing at least as much to keep.

Output saturates. Upkeep does not. That gap is the whole story, and the universe will not let you out of it.

That sentence is the engine of the simulation below, so it is worth making concrete. Picture two curves drawn against the growing complexity of a society. One is what the complexity produces — order kept, food grown, problems solved. That curve rises fast at first and then bends over and flattens, because the easy wins get used up: it is a curve with a ceiling. The other is what the complexity costs to maintain — and that one just keeps climbing, close to a straight line, because every office and aqueduct and legion you have ever added is still there with its hand out. A rising curve with a ceiling, and a rising curve without one. They start far apart, with output far above cost — that gap is the golden age, the surplus you build monuments with. But a bounded thing cannot stay ahead of an unbounded thing forever. The cost curve climbs into the flattening benefit curve and, at some unremarkable moment when the temples still look magnificent, crosses it. After that the society is running at a loss, papering the gap with debased coin and deferred maintenance and the seed grain. It can limp like that for a while. What it can no longer do is absorb a shock. And there is always, eventually, a shock.

So before we name the three societies that ran this exact program into the ground, build one yourself and watch it happen. Below is a society that does what societies do: it answers each problem by getting a little more complex. You set its temperament — how readily it reaches for complexity, how big an energy windfall it starts with, how much slack it keeps in the granary against a bad year. Press Begin, and watch the two curves on the right while the structure on the left grows.

The Experiment

Experiment — the society that keeps solving its problems
85
how readily the society answers each problem by adding a new layer
0
the size of a one-time bonanza — conquest, a new coal seam — that then depletes
30
how big a reserve it banks — how deep a deficit it can carry before it breaks
decade 0 complexity — output 0 upkeep 0 treasury 0 return on the last layer —
Left, the society — each band is a layer of complexity it has stacked up; it stands gold while it runs a surplus and goes ashen as it slides into deficit; the bar beside it is the granary. Right, the two curves over time: what the complexity produces (it has a ceiling) and what it costs to keep (it does not). Where the cost curve crosses above the output curve, the returns have gone negative — and a shock the granary can’t cover topples the whole stack at once.

Things to try:

Try 1 of 6

Just press Begin and leave the sliders alone. For a long while it looks like a triumph — the structure climbs, the output curve races out ahead of the upkeep curve, the granary fills. Then watch the upkeep curve catch up and the gap close. The moment the cost line crosses above the output line, the society is running at a loss it can’t see yet. A few decades later a bad year arrives that the granary can’t cover, and the whole stack sheds at once. That is collapse: not a murder, an arithmetic.

Try 2 of 6

Run it again and this time watch only one number: return on the last layer. It starts large and positive and ticks down every single decade. It crosses zero while the structure is still tall and the granary still full — long before anything looks wrong. The collapse is legible in that one figure a generation before it shows up in the skyline. Diminishing returns is not the alarm; it is the slow leak under the floorboards.

Try 3 of 6

Reset and push Energy windfall up high — this is Rome taking Egypt, Britain finding coal. The golden age stretches out; the windfall lifts the whole output curve and holds the crossing off for decades. But keep watching: the ceiling slowly sinks as the bonanza depletes, the upkeep curve climbs to meet it anyway, and the reckoning arrives late instead of never. A windfall buys time. It does not buy out.

Try 4 of 6

Now pull Appetite for complexity down toward the bottom. The society answers fewer of its problems by bolting on another institution; its complexity settles at a lower level and stops climbing — below the crossing point — so output stays comfortably above upkeep and it simply… keeps going. No collapse, no golden age either: a quieter, thriftier thing that can pay its own bills. The one durable exit is to stop solving every problem with more structure.

Try 5 of 6

Try to cheat the arithmetic: crank Appetite to the top and the windfall with it. You get a magnificent, towering run — and it still comes down, just later and harder. Output can rise no higher than its ceiling; upkeep has no ceiling to rise to. A bounded benefit cannot outrun an unbounded cost, however rich you start. Wealth changes the date of the reckoning, not the fact of it.

Try 6 of 6

After a collapse, take your hands off the controls and watch what the simpler society does. With its top layers gone, its upkeep has fallen off a cliff while its land and people remain — so output sits above cost again and the treasury refills. For the people who had been paying for all that grandeur, the fall reads less like an apocalypse than like a tax cut. Then, problem by problem, the climb begins again. Collapse isn’t the end of the curve. It’s the curve starting over.

1 of 6

What you have been watching is not a model of villainy. Nobody in that society is corrupt; every layer was added to solve a genuine problem, and at the moment it was added it paid for itself. The trap is that the layers accumulate and the returns do not. The society is doing the one thing it knows how to do — getting more complex to meet what comes — right up until that very competence is what sinks it. You did not watch a society fail to solve its problems. You watched it solve its problems all the way into the ground.

Three societies that ran the program

Rome is the case Tainter draws most fully, and it fits the curves almost embarrassingly well. The early Republic and Principate were the steep, profitable part: conquest poured in windfalls, and a relatively cheap apparatus turned them into roads and grain and order. Then the conquests ran out — there were no more rich neighbors within reach worth the cost of taking — and the empire was left holding the standing bill with no new deposits coming in. The third century is the curves crossing: barracks emperors, civil wars, that hemorrhage of silver out of the coinage. Diocletian and Constantine answered it with more complexity, not less — they roughly doubled the army and the bureaucracy, fixed sons to their fathers’ trades by law, and pressed taxes so hard that, the chronicles say, farmers walked off their land because the upkeep of being Roman had finally exceeded the benefit. When the Western provinces did slip away, many of the people in them felt it as a lightening.

The Maya of the southern lowlands ran a version with no barbarians in it at all. Over centuries the great cities — Tikal, Copán, Calakmul — competed by building: taller temples, larger courts, bigger monuments, more elaborate ritual and warfare and the priesthoods and laborers all of it required. Each king answered the problem of a rival king with another increment of grandeur, and the increments stacked into a complexity the thin tropical soils and the drought-prone rains could no longer underwrite. When the bad droughts came in the ninth century, the cities had no slack left to absorb them. The people did not vanish — the Maya are still here — but the courts, the kings, the calendar of monuments, the whole towering apparatus, came down and was not rebuilt. The jungle took the temples.

And the Ancestral Puebloans of Chaco Canyon ran the smallest, clearest version of all: a network of great houses in the high desert, knit together over generations into something elaborate and centralized and dependent on good rain at the margin of where rain can be counted on. When a long drought came late in the thirteenth century, the elaborate version could not pay for itself, and the people did the rational thing — they walked away from the great houses and dispersed into simpler, smaller, more self-sufficient settlements. Their descendants are the Hopi and the Puebloan peoples of today. They did not die. They de-complexified.

Three societies on three different continents with nothing in common — not their gods, not their crops, not their century — tracing the same curve to the same place. That is the kind of thing this site keeps pointing at: a pattern that recurs across systems that never met, because the same arithmetic sits underneath them. The tragedy of the commons was one such trap, where private gain and shared cost summed to ruin. This is its slower cousin, where yesterday’s solution is today’s fixed cost, and the bill comes due across generations instead of seasons.

Is there a door?

There is, but it is narrower than we want it to be, and the simulation is honest about which exits are real. The seductive one — the one our own civilization has been walking through for two hundred years — is the windfall. A vast new energy subsidy resets the curves: it lifts the output ceiling so high that the upkeep line has a long way to climb before it catches up, and the golden age stretches out across centuries. Coal and then oil were a windfall on a scale Rome could not have dreamed, which is exactly why we have been able to keep adding complexity — more specialists, more institutions, more global supply chains threaded through more intermediaries — for so long without the returns visibly turning. But a windfall is a deposit, not an income. You can watch in the simulation what the ceiling does as it depletes. It sinks. The crossing comes late instead of never. This is the part of Tainter’s argument that is genuinely unsettling, and the honest thing is to name it as a real worry rather than a certainty: the same diminishing returns may be operating right now, hidden, on the most complex society that has ever existed, postponed but not repealed by the richest energy subsidy any society has ever found.

The other door is the one almost nobody chooses willingly, and it is the only one that doesn’t end in a fall: spend less complexity. Solve fewer problems by building another layer, and shed the layers that have stopped paying before they bankrupt you. It sounds like decline because we have been taught that more complexity is always progress, but there is at least one society that took this door on purpose and bought a thousand years with it. After the West came down, the Eastern Roman Empire — Byzantium — in its worst seventh-century crisis did the unthinkable: it deliberately simplified. It let much of the money economy lapse, shrank and decentralized the army into self-supporting districts of soldier-farmers, thinned the bureaucracy, and pulled its ambitions in. It chose a lower, cheaper level of complexity over a higher one it could no longer afford — and survived, in that leaner form, for eight more centuries. Simplification, taken deliberately and early, is not collapse. It is the thing that collapse is the failure to do in time.

That is the difference between Try 4 and Try 1 in the simulation, and it is not a small one. The thrifty society and the magnificent one are running identical machinery; the only thing that differs is whether they keep their complexity below the line where the returns go negative. One of them never gets a golden age. The other gets a golden age and a grave.

And here is the note the page wants to end on, because it is the spine of this whole site held up to a hard light. Complexity is one of the most beautiful things the universe does — it is the eddy swimming up against the entropic current, the cathedral built out of nothing but local rules and patience. This site spends most of its pages admiring it. But the same machinery that builds the cathedral builds the trap, and the universe runs the program with the same blank competence either way. A pattern that recurs across Rome and the Maya and Chaco is not thereby blessed; it is simply cheap for the rules to make. The universe permits the diminishing return. It does not warn you when you cross the line, and it does not care that the temples are lovely on the day the cost curve finally catches the output curve.

What it leaves to us is the only part that was ever ours: to read the curve while there is still slack in the granary. To notice that the return on the last layer has quietly gone negative while everything still looks magnificent. To find the nerve to simplify on purpose, early, the way Byzantium did and almost no one else ever has — rather than the way Rome did, which was to keep solving its problems, beautifully and reasonably, all the way to the bottom. The frog on the bank does not get to vote on whether the pond fills. We, who can see the curve, do.

Sources & Further Reading