The Collapse Problem

Why Do Empires Fall?

Few questions in history have attracted so many answers. In 1984 the German historian Alexander Demandt gathered the explanations that scholars, preachers, and pamphleteers had offered for the fall of Rome alone, and counted two hundred and ten of them, from Christianity and lead poisoning to inflation, bureaucracy, soil exhaustion, moral decline, and the attacks of the Germans and the Huns. The list is a warning as much as a catalogue. When a single event can be explained in two hundred ways, the difficulty is not a shortage of causes but an abundance of them.

The difficulty deepens when the question is widened. Rome is only the most studied case among many. The Akkadians, the Hittites, the Assyrians, the Han, the Abbasids, the Mongols, the Aztecs, the Mughals, and the Ottomans all rose, and all of them, in one sense or another, fell. Their worlds differed in nearly every respect: in technology, religion, geography, economy, and political culture. If anything links their endings, it must lie below the level of particular events, in the structure of empire itself.

The search for such a structure is old. Writing in North Africa in 1377, Ibn Khaldun proposed in the Muqaddimah that dynasties pass through a cycle. A group bound by strong solidarity, which he called asabiyya, seizes power from a settled and softened predecessor. Its descendants enjoy the fruits of rule, lose the cohesion that won it, and come to depend on clients and hired soldiers, until a more cohesive challenger overthrows them in turn. A dynasty, he judged, rarely outlasts three generations: about a hundred and twenty years.

Six centuries later the archaeologist Joseph Tainter proposed a different mechanism. Complex societies, he argued in The Collapse of Complex Societies (1988), solve their problems by adding complexity: more officials, more soldiers, more infrastructure, more regulation. Each addition costs more and returns less, until a society finds itself paying ever more for ever less. Collapse, which Tainter defined as the rapid loss of an established level of social and political complexity, then becomes less a catastrophe than an economical choice.

Neither framework explains every case, and this essay does not propose a single theory. Its aim is narrower: to compare how several great empires actually came apart, to identify the vulnerabilities that recur among them, and to test the comparison against the Institute's own data on how long the rise and the fall of more than three hundred states and empires took.

Rome: A Fall in Slow Motion

The Western Roman Empire is the case against which all others have been measured, and the long debate over it shows how much depends on the question asked.

For much of the late twentieth century the most influential interpretation stressed continuity. Peter Brown's The World of Late Antiquity (1971) described an age not of decline but of transformation, in which Roman law, Christian institutions, and classical learning were reshaped rather than destroyed, and in which the end of the western emperors mattered less to most people than the rise of the bishop and the holy man. Seen in this light, 476 was a date whose significance was clearer to later chroniclers than to most of those who lived through it.

Two books published in 2005 pushed back. Peter Heather's The Fall of the Roman Empire argued that the West was brought down from outside, by the arrival of the Huns on the steppe frontier and by the Gothic, Vandal, and other groups they drove into Roman territory. Bryan Ward-Perkins's The Fall of Rome and the End of Civilization turned to archaeology and found, across much of the former western provinces, a collapse in material life: fewer coins, cruder pottery, smaller buildings, roofs of thatch where there had been tile, even smaller cattle. Whatever it meant to bishops and theologians, the end of Roman order meant a sharp fall in the standard of living of ordinary people.

The sequence of events supports both views. In 376 the Goths, fleeing the Huns, were admitted across the Danube; two years later they destroyed an eastern Roman army and killed the emperor Valens at the Battle of Adrianople. At the end of 406 the Vandals, Alans, and Suebi crossed the Rhine, and in 410 Alaric's Goths sacked Rome itself. Yet the decisive blow fell not on the city but on the treasury. In 439 the Vandals took Carthage, and with it the provinces of North Africa, the richest tax base left to the western government. Without African revenues the West could no longer pay for the armies it needed to recover what it had lost, and each loss of territory reduced the means of defending what remained. By the time Odoacer deposed the boy emperor Romulus Augustulus in 476, the western state had been shrinking for a generation into something closer to an Italian kingdom with an imperial title.

The eastern half of the same empire, sharing the same institutions, religion, and law, survived the same century and lasted, as the Byzantine Empire, for nearly a thousand years more. The contrast is instructive. Constantinople had shorter frontiers, a richer and more urban tax base in Anatolia, Syria, and Egypt that the invaders of the fifth century could not reach, and land walls that no army breached until 1453. Structure rather than culture made the difference: the same Roman system, under different pressures and with different resources, produced two different outcomes.

Kyle Harper's The Fate of Rome (2017) has since added a third layer, the environment. The empire's greatest age coincided with a warm, stable, and wet phase of climate that historians call the Roman Climate Optimum. Its troubles coincided with pandemics, the Antonine Plague from the 160s, the Plague of Cyprian in the middle of the third century, and the Plague of Justinian from 541, and with a cold phase in the sixth century that followed a series of great volcanic eruptions. Harper does not replace political explanations with natural ones. He shows how environmental shocks struck a system that was already stretched to its limits.

The Institute's curve for the Roman Empire captures the slow shape of the story. Its strength peaks in 117, the year of Trajan's death and of the empire's greatest extent. It falls to half that strength only in 235, when the murder of Severus Alexander opened the half century of disorder known as the Crisis of the Third Century. And the curve does not reach zero until 476, three hundred and fifty-nine years after the peak.

The Abbasids: Power Without Provinces

If Rome's fall was slow, the decline of the Abbasid Caliphate was slower still, and in one respect stranger: the institution outlived the state by centuries.

The Abbasids came to power in 750 by overthrowing the Umayyads, and in 762 the caliph al-Mansur founded Baghdad as their capital. Within two generations, under Harun al-Rashid (r. 786–809) and al-Ma'mun (r. 813–833), the caliphate presided over the richest and most sophisticated society in western Eurasia. Its armies were less remarkable than its tax registers, its markets, and its libraries. Yet the Institute's curve of Abbasid power falls to half within forty-one years of its peak in 820, and the reasons for the descent are visible in the sources.

The first was the army. Al-Ma'mun's successor al-Mu'tasim (r. 833–842) built a new force of Turkish slave soldiers from Central Asia, loyal to him personally rather than to the older Arab and Khurasani constituencies of the caliphate, and moved the court to a new capital at Samarra to house them. The solution worked for a generation; then the soldiers discovered their own power. In 861 Turkish officers murdered the caliph al-Mutawakkil, and in the nine years that followed, remembered as the Anarchy at Samarra, four caliphs were raised and removed by the troops who were supposed to serve them.

The second was the provinces. A caliph in Baghdad could govern distant regions only through governors, and governors with armies and revenues of their own had every reason to keep both. Ahmad ibn Tulun, sent to Egypt in 868, made it a hereditary dominion within a few years. The Saffarids took eastern Iran by force; the Samanids ruled Transoxiana and Khurasan as nominal servants of the caliph and real sovereigns of their own lands. In southern Iraq the Zanj rebellion of enslaved labourers (869–883) laid waste the heartland itself. Each province that kept its taxes was a province that no longer paid for the centre. In 945 the Buyids, a dynasty of soldiers from the mountains of Daylam south of the Caspian, entered Baghdad and reduced the caliph to a figurehead in his own capital.

What remained was legitimacy. Buyid emirs and, after 1055, Seljuk sultans sought the caliph's investiture, and his name continued to be pronounced in the Friday sermon across much of the Sunni world. In the twelfth century, as Seljuk power waned, caliphs such as al-Nasir (r. 1180–1225) rebuilt a modest state in Iraq. The Mongol sack of Baghdad in 1258, in which the last caliph of Baghdad was put to death, ended that revival. Even then the office survived, as a shadow caliphate under the protection of the Mamluk sultans in Cairo, until the Ottoman conquest of Egypt in 1517.

The Abbasid case reveals a pattern that recurs wherever empires decline slowly. Fragmentation is rarely a single event. It proceeds province by province, as local holders of military and fiscal power discover that the centre needs them more than they need the centre. The symbols of unity, the title, the coin, the name in the prayer, can survive long after the substance has gone, and their survival can disguise how far the decline has progressed. In the Institute's data the Abbasid profile is one of the most lopsided of all: seventy years from foundation to peak, and nearly seven hundred from the peak to the end of the office.

The Mongols: The Price of Speed

At the opposite extreme stands the Mongol Empire, the largest contiguous land empire in history and among the fastest built. When Temüjin was proclaimed Chinggis Khan at an assembly in 1206, the Mongols were one people of the eastern steppe. By 1241, on the Institute's curve, their power stood at its height. The Jin state of northern China had fallen in 1234; Central Asia and much of Iran had been overrun; and in a single week of April 1241 Mongol armies destroyed a Polish army at Legnica and the army of Hungary at Mohi. The rise had taken thirty-five years.

The speed was itself a source of fragility. The empire was held together less by institutions than by a family and a custom: the descendants of Chinggis Khan, gathered in an assembly called a kurultai, chose each new Great Khan. Every succession therefore halted the empire's business, and sometimes its campaigns; the death of Ögedei in December 1241 is traditionally held to have drawn the Mongol armies back from central Europe the following year. When the Great Khan Möngke died in 1259, the custom failed altogether. His brothers Kublai and Ariq Böke were each proclaimed Great Khan by rival assemblies in 1260, and the war between them lasted until 1264. Kublai won, but the western branches of the family never again obeyed a single ruler as they had before. The empire became four: the Yuan dynasty in China, the Ilkhanate in Iran, the Chagatai Khanate in Central Asia, and the Golden Horde on the western steppe.

The successor states soon went their own ways, adopting the religions, languages, and administrative traditions of the peoples they ruled. The Ilkhan Ghazan became a Muslim in 1295, and the Ilkhanate disintegrated after the death of Abu Sa'id without an heir in 1335; the Yuan were driven from China in 1368. The Golden Horde lasted longer, but its last remnant was destroyed in 1502. On the Institute's curve the Mongol Empire's strength halves fifty-three years after its peak, by the death of Kublai in 1294, and the curve ends in 1368 with the fall of the Yuan.

The Mongol case shows a vulnerability that charisma and conquest cannot cure. A state whose unity rests on a dynasty must solve the problem of succession in every generation, and a state that has grown faster than its institutions must solve it with nothing but family loyalty. Yet the Mongol case also complicates the word “fall”. The empire divided rather than vanished, and its successor states ruled much of Eurasia for another century. What collapsed was the unity, not the power.

Sudden Endings

Not every empire declined. Some simply ended.

The Aztec Empire, the Triple Alliance centred on Tenochtitlan, was at the height of its power in 1519, and the Institute's curve peaks in that very year, when Hernán Cortés landed on the Gulf coast of Mexico. The curve reaches zero two years later, with the fall of Tenochtitlan in August 1521. The Inca Empire, the largest state in the Americas, peaked around 1515 by the Institute's estimate; by the end of 1533 Francisco Pizarro had executed its ruler Atahualpa and entered its capital, Cusco.

In both cases a few hundred Spanish soldiers were the visible agents, but they were not the whole cause. Smallpox reached Tenochtitlan in 1520. In the Andes an epidemic, probably smallpox, killed the Inca ruler Huayna Capac and his chosen heir in the late 1520s and opened a civil war between his sons Huáscar and Atahualpa, which was ending just as Pizarro arrived. In Mexico, Cortés's victory depended on tens of thousands of indigenous allies, above all the Tlaxcalans, who had their own reasons to destroy the Mexica tribute state. An empire that drew tribute from resentful subject peoples, struck by a disease it had never known, found that many of its subjects were ready to change masters.

The Achaemenid Empire, the largest the world had yet seen, ended almost as quickly. Alexander of Macedon crossed into Asia in 334 BCE; within four years he had defeated the Persian armies in three great battles, burned the ceremonial capital at Persepolis, and seen Darius III murdered by his own officers. Two centuries of Persian rule over lands from the Aegean to the Indus passed to a new master in a single campaign.

The Late Bronze Age offers a collapse on a larger scale. Around 1200 BCE, within a few decades, the palace states of the eastern Mediterranean fell one after another. Hattusa, the capital of the Hittite Empire, was abandoned and partly burned; the great trading city of Ugarit on the Syrian coast was destroyed and never reoccupied; the Mycenaean palaces of Greece were burned, and with them went the Linear B script. Egypt survived, diminished, after Ramesses III fought off the invaders whom modern historians call the Sea Peoples. Eric Cline's 1177 B.C. (2014) argues that no single cause can explain the collapse. Drought, famine, earthquakes, invasion, internal revolt, and the breakdown of the long-distance trade on which the palace economies depended combined into a “systems collapse”, in which the failure of each part made the failure of the others more likely. The more tightly the kingdoms were bound to one another by trade and diplomacy, the more surely the fall of one dragged down the rest.

The fall of the Assyrian Empire was almost as sudden. In 671 BCE, when Esarhaddon conquered Egypt, Assyria was the greatest power the world had yet seen, and the Institute's curve stands at its peak. Within six decades it had vanished. After the death of Ashurbanipal a struggle over the succession weakened the state, Babylon rose under Nabopolassar, and the Medes and Babylonians took Assur in 614, sacked Nineveh in 612, and defeated the last Assyrian attempt to retake Harran in 609. An empire that had ruled by terror and mass deportation found few defenders.

Sudden endings share a profile. The empires that fell fastest were often those that ruled subject peoples by tribute and fear rather than by incorporation: the Assyrians, the Mexica, and to a lesser degree the Inca. When such an empire suffered a severe shock, a disease, a disputed succession, a determined coalition, an invader of unfamiliar strength, its subjects did not rally to defend it. Coercion is efficient while it works, but it builds no reserve of loyalty for the day it fails.

Recurring Vulnerabilities

Set side by side, these cases suggest not a single cause of collapse but a small number of recurring vulnerabilities, which appear in different combinations from one empire to the next.

The first is fiscal. Empires are expensive, and the cost of holding territory tends to grow faster than the revenue it yields. Philip II of Spain, master of the silver of the Americas, suspended payments to his creditors four times, in 1557, 1560, 1575, and 1596. When expenses outrun revenue for long enough, states debase their coinage, sell their offices, or hand their revenues to local power-holders in exchange for immediate support, as the second essay in this series described, and each expedient solves one crisis by weakening the state's capacity to face the next. This is the historical form of Tainter's diminishing returns. Paul Kennedy's The Rise and Fall of the Great Powers (1987) gave it a modern name, “imperial overstretch”: the condition of a state whose commitments have outgrown the economic base that sustains them.

The second is succession. Every empire that depended on a single ruler had to transfer that ruler's authority at least once a generation, and every transfer was a moment of danger. The empire of Alexander the Great did not survive his death in 323 BCE as a single state. The Carolingian Empire was divided among Charlemagne's grandsons by the Treaty of Verdun in 843. The Mongol Empire split in the war of 1260 to 1264. Institutions that settled the succession by rule rather than by contest, even imperfect ones, conferred a durability that personal greatness could not.

The third is the army. Empires repeatedly entrusted their defence to soldiers whose loyalty was owed to commanders, paymasters, or their own corps rather than to the state: the Praetorian Guard, which in 193 sold the Roman throne to the highest bidder; the federate armies of the late Roman West; the Turkish guards of Samarra. A military that can make and unmake rulers will eventually do so.

The fourth is the elite. Peter Turchin and Sergey Nefedov's Secular Cycles (2009) traced, in England, France, Rome, and Russia, a recurring sequence in which long periods of population growth and prosperity multiplied the number of people who expected elite positions, until competition among them outran what the state could provide. Factional conflict, fiscal crisis, and civil war followed, until a period of disorder reduced both the population and the elite and the cycle could begin again. In this view the danger to an empire lies not only in its external enemies but in its own success, which creates more claimants to power than the system can satisfy.

The fifth is the environment, including disease. The Roman pandemics, the epidemics of the Americas, the droughts of the Late Bronze Age, and the Black Death of the fourteenth century, which struck the Golden Horde and the Mamluk Sultanate alike, show how natural shocks can strike a political system at its weakest point. They rarely destroyed healthy states on their own. They turned strains into crises.

What these vulnerabilities have in common is that each is a failure of renewal. Every empire faced fiscal strain, contested successions, ambitious soldiers, restless elites, and natural disasters. Those that lasted were not those that escaped such shocks, but those that possessed institutions able to absorb them and recover.

The Lesson for Institutions

The comparison suggests a reframing of the question with which this essay began. To ask why empires fall is, in a sense, to ask the wrong question, for the fall of empires is the historical norm. Every one of the ninety-four empires the Institute profiles has come to an end. The more useful questions are why some lasted so much longer than others, and why some declined over centuries while others vanished in a few years.

The answer that emerges from the cases is consistent with the argument of the first essay in this series. Durability belonged to empires whose power was embedded in systems rather than persons: in fiscal institutions that could raise revenue in a crisis, in rules of succession that could survive a weak ruler, in armies tied to the state rather than to their commanders, and in a relationship with subject peoples that gave them some stake in the empire's survival. The Eastern Roman Empire, which possessed most of these, outlived the Western by nearly a thousand years. The Mongol Empire, which possessed few of them, divided within two generations of its founding.

Decline, moreover, is rarely visible from the inside. The Abbasid caliphs kept their titles, their ceremonies, and their place in the Friday prayer long after their armies and revenues had passed to others. The western Roman government went on issuing laws for provinces it no longer controlled. Institutions in decline tend to preserve their forms most carefully at the very moment their substance is draining away, because the forms are what remains. The lesson for the historian, and for anyone responsible for an institution, is to look past the forms to the working parts: who collects the revenue, who commands the soldiers, who decides the succession, and whether the people being governed have any reason to want the government to survive.

What the Data Shows

The Institute's profiles give each state and empire a power curve: an estimate of its strength over time on a scale of zero to ten, measured against its own peak. The curves are interpretive judgements drawn from the historical record rather than measurements, and the dates at which states begin and end are often matters of convention. Taken together, however, they allow the questions of this essay to be asked of hundreds of cases rather than a handful.

Of the profiles in the Institute's collection, 342 describe states and empires that have come to an end: 94 empires and 248 other states. Their median lifespan is 320 years. The median time from foundation to peak is just under 160 years; the median time from the peak to the end is 118 years. In 56 per cent of cases the fall was faster than the rise. Among the great empires the proportion is lower, 47 per cent, and among the other states it is higher, 60 per cent, which suggests that the largest empires, once established, tended to decline more slowly than they had grown, cushioned by the very scale and institutions that made them empires.

The pace of the turn is more striking still. Measured from the peak to the first point at which a state's strength had fallen to half, the median interval is just 41 years, roughly the span of a single career. The Tang dynasty's curve halves in fifteen years, between its height in 740 and the outbreak of the An Lushan rebellion in 755, although the dynasty itself survived until 907. The Abbasid curve halves in 41 years, between 820 and the murder of al-Mutawakkil in 861. The Roman Empire, at 118 years from Trajan to the crisis of the third century, is among the slower cases.

The data therefore separates two things that are often confused: losing power and coming to an end. Most states lose the greater part of their strength within a lifetime of their peak, often abruptly. But the institutions, titles, and dynasties they leave behind may endure for generations or centuries afterwards, as the Abbasid caliphs, the later Tang emperors, and the last Mughal emperors in Delhi did. Collapse, in the sense of a sudden fall from the height of power, is common. Disappearance is usually slow.