Since Montesquieu, Europe’s success has been credited to its political fragmentation. Yet for most of recorded history the richest region on earth was China — and China was usually one empire.
Between AD 0 and 1800, historical China, roughly the landmass from the Mongolian steppe to the South China Sea answered to a single authority for 1,008 years. Fewer than 80 states ruled any part of China across those eighteen centuries. Europe, over the same span, went from 37 sovereign states in AD 600 to 61 by 900 and 114 by 1300 — and stayed that way.
Interstate war was ordinary in Europe and comparatively uncommon in China. But the conflicts China did fight were far more destructive, because they were not border adjustments between neighbours — they often involved the collapse of an entire empire.
Only two European wars before 1750 are estimated to have killed more than five million people. Five such wars are recorded for China, and each one had a nomadic dimension.
Before 1800 every major invasion of China came from the north. Mountains, rainforest and ocean sealed the other three flanks. Medieval Europe had no such luck: Vikings from the north; Arabs, Berbers and Turks from the south and south-east; Magyars and Mongols from the east.
That is the asymmetry the model turns on. A threat that always arrives from one direction can be met by one army standing in one place. Threats arriving from several directions cannot.
The continent is the interval [0, 1] with population spread evenly along it. A threat of size Λ strikes a frontier and decays inland at rate α. A regime picks a capital G and buys military M at cost θM²; its strength at distance t from the capital is M − βt². If the regime fails to shield a fraction δ of the continent, it is overthrown.
Capitals and armies below are the closed-form equilibrium of Propositions 1 and 2; the shielded region and the damage that gets through are integrated across the continent. Move the controls and watch which arrangement survives.
Seven results fall out. Two are about who survives; the rest are about what survival costs the people living there.
China offers 2,060 years of decadal data. The frequency of nomadic attacks comes from Bai and Kung (2011), who show that steppe incursions tracked exogenous swings in rainfall. Political unification comes from Wei (2011).
Intuitively, we expect an attack in the current decade to make unification less likely: a state under assault fractures. But as the model predicts, the long-run effect is consolidation. An attack in the previous decade makes unification more likely.
| Specification | Contemporaneous | One decade later | Long-run effect | Obs. |
|---|---|---|---|---|
| ADL, no controls | −1.20 pp** | +1.96 pp*** | +6.3 pp | 203 |
| VAR, with controls | — | +1.76 pp*** | +8.3 pp | 203 |
| IV — droughts & floods | −8.74 pp | +13.4 pp** | +20.4 pp | 203 |
Europe cannot be tested the same way — state counts exist only at century intervals. But the narrative fits. Rome fell when its military edge eroded and threats multiplied along a very long border. Justinian’s reconquest collapsed under the Avars and Persians, then the Arabs. Charlemagne’s empire was pulled apart by Magyars, Vikings and Muslims arriving on three different fronts.
After the eleventh century the Viking and Muslim threats receded — and, as the model predicts, fragmentation simply persisted. The Mongol invasion of the thirteenth century was too brief to provide a sustained impetus toward European unification.
If the model is right about why these political systems existed, it should also be right about how they behaved. Three predictions, three bodies of evidence.
| 1700 | 1750 | 1780 | Total 1780 (t. silver) | |
|---|---|---|---|---|
| Dutch Republic | 210.6 | 189.4 | 228.2 | 466.8 |
| England | 91.9 | 109.1 | 172.3 | 1,627.3 |
| France | 43.5 | 48.7 | 77.6 | 1,962.0 |
| Spain | 28.6 | 46.2 | 59.0 | 642.5 |
| European average | 52.1 | 58.0 | 77.3 | 711.5 |
| China | 10.4 | 11.8 | 9.2 | 2,769.3 |
Before AD 400 — when both ends of Eurasia were ruled by empires — the two population series behave alike. The divergence in volatility appears only after China consolidates and Europe fractures. That timing is the point.
Higher-frequency Chinese data from Cao (2000) makes the mechanism visible directly: the falls of Xin, Han and Sui, the An Lushan Rebellion, the Jurchen and Mongol invasions, the ends of Yuan and Ming — each leaves a crater. Between 1223 and 1292 China’s population fell by roughly 30 million, a third of the total. Europe had exactly one continent-wide collapse after Rome: the Black Death.
We are explicit in the paper that the steppe is not the whole story, and we name two things our own mechanism cannot account for.
Our argument is about sequencing, not monocausality. Multisided threats fragmented Europe before 1100. That initial fragmentation is what let a rival Church and independent city-states entrench themselves — and those institutions then kept Europe divided long after the invaders stopped coming. Tilly’s city-states and Hoffman’s papacy are consequences of the first-millennium collapse as much as causes of what followed.
Political centralization was stable in China and unstable in Europe, and the difference paid off in opposite directions on different time horizons. Unity was efficient in the short run — no wasteful arms race, lighter taxes, faster growth. Fragmentation was robust in the long run.
The start-stop character of growth under empire — long expansions erased by catastrophic contractions — is offered as one route into the larger puzzle: why the most populous economy in the world came within a hair’s breadth of modern growth and did not get there.