Castles

Medieval Europe was littered with castles, and most of them were not the king's. They belonged to barons. Judged by the standard of the modern state, this is a textbook failure: no monopoly of violence inside the territory. Strong kings, the story runs, pull down baronial castles; weak kings let them multiply.

Desierto and Koyama argue the conventional reading has the causation backwards. Drawing on James C. Scott's Seeing Like a State, they treat the feudal realm not as a proto-state but as a coalition — a king and a set of barons, each holding their own military and economic resources. In that world the baronial castle is not friction. It is the instrument that makes the king's promises worth anything.

The state-centric story does not fit the data

If castles were an obstacle that centralizing rulers progressively cleared away, the share of castles in royal hands should trend upward. Across England, France and Castile it does not. Over four centuries the royal share is close to flat — while the absolute number of castles, especially in France, grows enormously.

There is no march toward monopolization to be found. Kings acquired castles and lost them; they granted castles away as readily as they took them. Something other than a slow victory of the centre is going on.

Any king worth his salt … would look to the castles of his realm, look after his own, seek to ensure, by the exercise of his huge power of patronage, that as many castles as possible were in the hands of those he could trust. R. Allen Brown, Allen Brown's English Castles, p. 166

Note what the ambition is: castles in the hands of men the king could trust — not castles in the king's own hands. That distinction is the whole argument.

The king cannot rule alone. He needs the barons' men, horses and money, and he buys them with a share of the realm. But nothing binds him: there are no courts above him, no constitution, no mechanism that makes a royal promise enforceable. He can always renege and take a larger share for himself.

So what keeps him honest? The baron's castle. A rebel baron retreats behind his walls and takes with him whatever he can defend. The castle makes those resources illegible — visible, perhaps, but not appropriable. Reneging therefore costs the king twice: he loses the withdrawn resources, and he fights the next baron with a smaller war chest. Valuable baronial castles make royal promises credible.

Turn the crank

Below is the model's comparative statics made operable. One dial: how much a castle is worth militarily. Everything else — who joins, whether the king keeps his word, what kind of realm you end up with — follows.

Four kinds of realm

Theorem 2 sorts feudal polities on two axes. A realm is large when every baron in the territory joins the coalition, and consolidated when the king honors every bargain he has made. Both depend on what the castles are worth — and on how castle strength is distributed among the barons.

The prediction that follows is blunt: territories thick with valuable baronial castles form large realms that rebel rarely. Territories with few form small realms that rebel often. Castles and political order move together — in the same direction, not opposite ones.

Henry II of England and Philip Augustus of France were about as formidable as twelfth-century rulers came. Both expanded their domains enormously. Neither made any serious attempt to bring the castles of the realm into royal hands — and Philip did the opposite when it suited him.

Not only did Philip acquire and build fortresses to consolidate the royal domain and his new acquisitions, he also gave them away when it best served his interests. John Baldwin, The Government of Philip Augustus, p. 301

In 1189 he pledged Montboissier to the count of Auvergne; in the Auvergne he handed three castles to Bertrand de la Tour, four to the bishop of Clermont, five to the bishop of Le Puy. On conquering Normandy he granted castles out to his vassals. This is not a ruler monopolizing military power. It is a ruler who needs lords with the capacity to hold ground on his behalf, and who must give them something worth holding it for.

The bridle on the king

The mechanism has a sharp empirical implication for the moment when the bargain broke down. In 1215 a coalition of English barons forced Magna Carta on King John. Which barons joined? On this account, a baron weighs the non-appropriable strength of the coalition he would be joining — and castles are what is non-appropriable.

Using the universe of English barons alive in 1215 and their family networks, Desierto, Hall and Koyama find that the number of castles held by rebels in a baron's family network strongly predicts his joining the rebellion.

Note that this sits alongside the model rather than against it. Lemma 2 works at the level of the realm: more castle value, more royal restraint, fewer rebellions. The Magna Carta result works within a rebellion once one has broken out: given that the realm is coming apart, the barons with the most rebel castles around them are the ones who move.

Show the partial-regression plot from the paper

Fig. 4 — Reproduced from the paper. Partial-regression plot, conditional on the full set of controls. Data: Desierto, Hall & Koyama.

If castles were so functional, why did they go? Not because centralizing kings pulled them down. Because gunpowder made them cheap to take.

Cannon mattered from the fifteenth century, once iron shot arrived. In 1453 the Ottomans breached the land walls of Constantinople, previously thought impregnable. In 1464 Bamburgh became the first English castle reduced by artillery. The model's prediction for a fall in castle value is unambiguous: the feudal equilibrium destabilizes — more reneging, more rebellion. That is what the record shows.

Compare two English civil wars two and a half centuries apart. The First Barons' War was a war of sieges. The Wars of the Roses was not.

Contemporaries did not see it coming. At the outbreak the English government ordered a survey of castles and fortified towns to check they were in repair, and the early 1460s brought a run of campaigns for the Lancastrian castles of the north. Then Bamburgh fell to cannon, and sieges stopped mattering. By the end of the century England was invaded repeatedly because the coastal castles stood "in decay, ungarrisoned and unmunitioned."

In truth to one who did not see it, the operation would seem incredible; here they are fighting more with military engines than with manpower. A contemporary on Louis XI's artillery, 1465

What replaced the castle was not the castle. The trace italienne was vastly more expensive, sat on frontiers rather than scattered through the interior, and could only be built and garrisoned by a state with regular tax revenue. The decentralized fortification — the thing a single baron could hold with a small garrison and use to threaten an army's supply lines — simply stopped existing. The feudal equilibrium did not fall to a stronger state. Its technological foundation was taken out from under it.

William of Newburgh called castles regni ossibus — the bones of the kingdom. The paper explains why the phrase is exact. In a world with no courts above the king, no constitution, no device for binding a ruler's future self, the baronial castle solved the commitment problem of medieval governance. It made reneging expensive, and so made promises credible.

There is a modern rhyme, and it is worth being careful about it. Large firms and wealthy individuals hold assets the state cannot easily see or seize — offshore accounts, layered corporate structures, mobility. Like castles, these can serve both parties where they help a state commit not to expropriate. But the difference is decisive: a holding company cannot be garrisoned. We live under states with a genuine monopoly on violence and high capacity. Private wealth is not a check on government in the way a baron with a castle was.

Scott was skeptical of "the hegemony of the nation-state as the standard and nearly exclusive unit of sovereignty." Read in that spirit, the paper asks us to take the feudal world on its own terms rather than scoring it against a standard it was never trying to meet. Do that, and what looked like state failure turns out to have been a working — and rather durable — form of political order.