Why Centralization Sometimes Makes States Weaker

At the Escorial, the vast monastery-palace complex outside Madrid that Philip II built as the administrative center of the Spanish Empire, the King worked. He worked relentlessly, obsessively, in ways that alarmed his court physicians and exasperated his ministers. He read every dispatch from governors in Peru, the Netherlands, Italy, and the Philippines. He annotated documents in the margins — sometimes correcting the Latin, sometimes approving a sentence and rejecting the next, sometimes countermanding a decision a provincial governor had made six months earlier based on conditions that no longer existed. He is reported to have said that he preferred slow death to disorder. He achieved both.

The Spanish Empire under Philip II was the largest political structure the world had yet seen, stretching from the Philippines to Peru, from Sicily to the Netherlands. It was also, by most measures, the most centralized. And it was, despite its enormous resources, consistently less effective at translating those resources into political outcomes than smaller, more administratively diffuse competitors. The Dutch Republic, a loose confederation of seven provinces with no central executive worth the name, fought the Spanish Empire to a standstill and eventually to a negotiated truce. England, with perhaps a tenth of Spain’s imperial revenues, became a major naval power. The pattern demands explanation.

The explanation is information economics. And it applies not just to Philip II’s Spain but to every political structure that has tried to govern complex environments through central control.

The Information Problem at the Heart of Centralization

The fundamental problem with centralized decision-making is not that central authorities are stupid or corrupt, though they may be both. It is that they are informationally limited in ways that are structural and unavoidable. A central authority can only receive, process, and act on a fraction of the information that exists in the system it is trying to govern.

This point was made famously by Friedrich Hayek in the context of economic planning, but Hayek’s argument applies with equal force to political governance. Hayek’s observation was that economic knowledge is not concentrated — it is dispersed across millions of individual actors who each possess small fragments of information about their local circumstances: the price of a particular input, the availability of a particular labor skill, the demand for a particular product in a particular location. No central planner can possess this knowledge because it is not the kind of knowledge that can be aggregated and transmitted. It is tacit, local, and constantly changing.

The same is true of governance. A provincial governor in sixteenth-century Peru possesses information about local conditions — the reliability of local allies, the state of the roads, the mood of the indigenous population after a recent epidemic, the particular character of a local dispute — that cannot be fully conveyed in a dispatch to Madrid. Philip could read the dispatch. He could not possess the contextual knowledge that would make his marginal annotation appropriate to conditions on the ground. His decisions, made on the basis of incomplete and dated information, were often worse than the decisions his governors would have made with local discretion.

The deeper problem is not just that Philip lacked information but that his attempt to centralize decision-making degraded the information-processing capacity of the system as a whole. When governors know that their decisions will be reviewed and potentially countermanded by the center, they lose the incentive to develop and exercise the judgment that good governance requires. They become administrators of instructions rather than genuine decision-makers. The information that would have informed good local decisions — the knowledge that can only be acquired by the person closest to the problem — is no longer acted on, because acting on it requires discretion that the system does not grant.

The Soviet Planning Failure as Limiting Case

The Soviet Union’s experience with central economic planning is the most extensively documented case of central control failing because of information economics, and it is worth dwelling on because it reveals the mechanism with unusual clarity.

Soviet central planning attempted something that Philip II’s empire did not: comprehensive coordination of an entire economy, specifying quantities of every output, allocating every input, setting every price. The Gosplan bureaucracy employed tens of thousands of people, processed millions of pieces of data, and produced five-year plans of extraordinary detail. The ambition was total — not just to govern but to substitute the planning system for the market as the economy’s information-processing and coordination mechanism.

The failure was complete, and the mechanism was exactly what Hayek predicted. The information required to make good production decisions — what inputs are available at what quality, what technologies are most efficient at a given plant, what consumers actually want rather than what planners assume they want — was distributed across the entire economy. It could not be aggregated and transmitted to a central planning bureau without massive distortion. By the time information traveled from the factory floor to Gosplan and back as instructions, conditions had changed, the local knowledge embedded in the original information was lost, and the instructions were no longer appropriate.

The Soviet system developed characteristic pathologies that are predictable from the information problem. Enterprises hoarded inputs, because inputs not formally requested through the planning system were unavailable when needed. Managers met quantity targets at the expense of quality, because quantity was observable and verifiable while quality was not. Planners relied on the previous period’s outputs as the baseline for the next period’s plan (the “ratchet effect”), because they had no other way to set targets, which meant successful enterprises were punished with higher targets and unsuccessful ones were rewarded with lower ones. The information problem generated perverse incentives throughout the system.

The Soviet Union grew rapidly in the 1930s through 1950s by mobilizing underutilized resources — labor, land, and capital that had been left idle or poorly allocated under the previous system — and directing them toward priority sectors. This is what centralized systems can do well: mobilize and redirect resources when the direction of mobilization is known, the mobilizable resources are identifiable, and the coordination problem is essentially logistical. Building steel mills, electrifying the countryside, constructing railways — these are problems where central direction works reasonably well because the goals are clear and the means are technically known.

The system failed when the easy mobilization phase was complete and productivity growth required innovation — the development and implementation of better methods, better products, better technologies. Innovation requires local experimentation, tolerance for failure, and responsiveness to feedback about what is actually working. Central planning could not provide any of these. The information required to manage an innovative economy was too distributed, too tacit, and too dynamic for any central authority to process.

The Ottoman Success Through Delegation

The Ottoman Empire at its height (roughly 1450 to 1650) offers a different data point — a large, successful political structure that achieved its success partly through deliberate delegation rather than centralization.

The Ottoman system managed a vast and diverse territory through a layered structure of intermediaries. Timar holders (cavalry soldiers granted revenue rights over agricultural land) managed rural taxation and defense in Anatolia. Christian millet communities managed their own internal legal affairs under their own religious law. Armenian merchants, Jewish financiers, and Greek shipowners managed specific economic functions under arrangements that gave them autonomy in their domains of competence. Janissary commanders managed garrisons with significant local discretion. The system was held together by a powerful but thin central authority — the Porte in Constantinople — that maintained overall direction while delegating most actual governance to local intermediaries who possessed the local knowledge to exercise it.

This was not decentralization by default or failure of state capacity. It was a deliberate organizational choice, reflected in sophisticated legal and administrative doctrine. The Ottoman state knew what it needed to control centrally — the military capacity to suppress rebellion, the revenue system to finance the court and army, the appointment of high officials who served at the Sultan’s pleasure — and delegated everything else. The system worked because the division of control followed the division of information: the things that required local knowledge were managed locally by people who had it.

The Ottoman decline in the eighteenth and nineteenth centuries is partly a story of attempting to modernize by centralizing — replacing the diffuse, locally-adapted intermediary system with a more Western-style centralized bureaucracy. The Tanzimat reforms of the 1839–1876 period were explicitly aimed at replacing the old intermediary structures with a more uniform, centrally controlled administrative system. The reforms succeeded in some respects — they created a more legible state, a more uniform legal code, a more professional officer corps. But they also destroyed the local knowledge and flexibility that the old intermediary system had provided, without fully replacing these with the administrative capacity of a genuinely modern bureaucratic state. The result was a state that was in some ways more centralized and in some ways weaker.

European Competition as Institutional Laboratory

One of the more robust arguments in economic history — associated with scholars like Joel Mokyr and David Landes — is that Europe’s political fragmentation from roughly 1000 to 1800 CE was actually a driver of institutional and technological innovation rather than a weakness. The argument runs as follows: because Europe was divided into many competing political units, no single authority could suppress experiments in governance, religion, technology, or commerce that challenged existing arrangements. Failed states could be abandoned; successful institutions migrated across borders. Competition among states for taxable population and mobile capital drove institutional improvement.

This argument is sometimes overstated, but the core insight is sound. The comparison between European fragmentation and Chinese centralization is instructive. China under the Qin unification (221 BCE) and its successors achieved remarkable administrative uniformity over a vast territory. This uniformity enabled the scale effects that made China the world’s largest economy for most of the two millennia following unification. But it also eliminated the local experimentation and institutional competition that European fragmentation permitted.

The Chinese examination system, which provided meritocratic access to the bureaucracy based on mastery of classical texts, was a genuine institutional innovation — it created a more capable administrative class than hereditary aristocracy would have. But it was also a centralizing institution that homogenized the ideas and values of the governing class across the empire, reducing the institutional diversity that innovation requires. When the examination system calcified — when the texts to be mastered became fixed and the curriculum became increasingly divorced from practical governance — it became a mechanism for reproducing failure rather than generating success.

European universities, by contrast, competed. The University of Bologna competed with Paris competed with Oxford competed with Leiden. No single authority controlled what could be taught or debated across this network. The result was a more chaotic intellectual environment than the Chinese examination system produced — and a far more innovative one. The scientific revolution and the subsequent industrial revolution emerged from the European competitive system, not despite its political fragmentation but partly because of it.

What Modern Organizational Research Shows

The trade-off between centralized coordination and local knowledge is not just a question for historians. It is a live question in organizational design, and the empirical research on it is reasonably clear.

Large organizations — corporations, government agencies, nonprofits — consistently face the same trade-off that Philip II faced: central control allows coordination and prevents the local deviations that create chaos, but it also concentrates decision-making in the hands of people with less local knowledge than those closest to the problem. The management literature has explored this trade-off extensively, and the findings are consistent: the optimal degree of centralization depends on the nature of the task.

For tasks requiring close coordination — manufacturing processes where components must fit precisely together, financial reporting where consistency matters, safety-critical systems where deviation can be catastrophic — centralization works. The cost of local variation exceeds the benefit of local knowledge. For tasks requiring responsiveness to local conditions — sales, service delivery, product development, research — decentralization outperforms centralization. The cost of information loss at the center exceeds the benefit of coordination.

The most successful large organizations are those that have correctly identified which decisions require centralization and which require local discretion — and have resisted the managerial temptation to centralize everything because centralization is more comfortable for those at the top. Top-down control provides legibility: the center can see what is happening and intervene. Decentralization requires trusting people whose decisions you cannot fully monitor, which is psychologically and politically difficult. The bias toward centralization, even where it degrades outcomes, is predictable from the psychology of leadership.

The Limit of Any Centralized System

The conclusion from four hundred years of evidence — from the Escorial to Gosplan to modern organizational research — is that centralization solves coordination problems while creating information problems, and the limit of any centralized system is the amount of information its decision-makers can process.

Philip II’s Spain was not defeated by superior Spanish strategy — Spain didn’t have any. It was defeated by the compounding costs of making bad decisions at the center based on information that was too incomplete, too outdated, and too distorted by the transmission process to support good governance. The Dutch Republic, which made most of its military and economic decisions at the provincial and municipal level, was faster, more adaptive, and more capable of responding to local conditions precisely because it did not try to route every decision through a central authority.

The most successful large political structures — the British Empire at its functional peak, the Ottoman Empire before the Tanzimat reforms, the United States federal system before its twentieth-century centralization — figured out the division of governance that follows the division of knowledge. They centralized what required centralization: military force, monetary policy, foreign relations, the legal framework. They left local what required local knowledge: commercial decisions, municipal administration, the management of local disputes.

The paradox of centralization is not really a paradox. It is a straightforward consequence of information economics. States that try to control everything can only process a small fraction of the information required to control everything well. States that correctly identify what to control and what to leave alone process the relevant information through the actors who actually possess it. The first kind of state looks more powerful from the center. The second kind of state is more powerful in practice. Philip II’s marginal annotations changed nothing in Peru. He would have been better served by appointing a good viceroy and leaving him alone to govern.

The lesson is available to any state willing to learn it. Most prefer the comfort of control to the discipline of delegation. The history of governance suggests that preference is expensive.