CORE PATH Stop 48 / 106

Polycentricity and subsidiarity October 17, 2021 Updated: September 11, 2026 14 min read THY-REALITY

How Do We Measure Whether Decentralisation Is Real?

Decentralisation is not the number of municipalities, boards or providers. This article turns it into a measurable profile of decision authority, finance, ownership, data, supply routes, exit costs, redundancy and corrigibility of power.

the articles from “Monopoly, Plutocracy and the Concentration of Economic Power” through “Institutional Capture: When the System Starts Serving Itself” identified four different pathways by which power can reconcentrate: economic concentration, delegate drift, emergency centralisation and institutional capture. Yet all four can emerge inside a system that still describes itself as decentralised. If decentralisation is measured only by counting municipalities, boards, firms or servers, we can end up with a very distributed façade and a highly centralised core.

This article therefore treats decentralisation not as an identity or slogan but as a measurable property of a system. The question is not only how many units exist, but where decisions are actually made, who controls revenues and assets, who holds the data, how many independent supply routes exist, how difficult exit is, and whether the system survives the failure of one centre.

The most important methodological choice is not to search for one magic index. OECD work on fiscal decentralisation shows that different dimensions of autonomy can move independently: a large local spending share does not necessarily imply strong tax or decision autonomy. The same is true more broadly. Real decentralisation is a profile of several measures, not a single number.

Formal and functional decentralisation are not the same thing

Formal decentralisation describes structure: how many levels, offices, legal entities or local units exist. Functional decentralisation asks how much autonomous power those units actually possess. A local authority may exist while being unable to change rules, budgets, staffing or providers. Ten companies may exist while all depend on the same platform, bank, logistics hub or data interface.

The World Bank therefore distinguishes political, administrative and fiscal decentralisation. The Council of Europe’s local self-government framework likewise refers not only to local institutions but to their own competences, structures and financial resources. Decentralisation is therefore not a question of what the organisational chart looks like, but of how much real discretion exists.

This article’s first test is simple: if the lower level cannot make an important decision, secure part of the necessary resources and correct an error without permission from the centre, decentralisation is probably more formal than functional.

It is tempting to give a system one score from 0 to 100. But such a number can conceal exactly what we need to see. A system may be highly decentralised in service delivery yet completely centralised in finance. Ownership may be dispersed while data infrastructure remains singular. There may be many suppliers that all rely on the same upstream source.

This article therefore proposes a multidimensional dashboard. Each axis measures a different form of concentration and the results are read together. If one axis shows extreme centralisation, it should not disappear inside an average of the others. The most fragile axis is often the one that matters most.

In practice this means preferring seven or eight intelligible measures with explanations over one impressive but opaque number. The purpose of measurement is not to manufacture a political slogan, but to reveal where one point can still decide for everyone else.

Axis 1: where does real decision authority sit?

The first axis measures decision concentration. For every important function, record who can make the final decision, who can veto it, who sets the agenda, and how much discretion the lower level has to adapt implementation to local conditions. It is not enough that a local body is nominally 'responsible' for a service if all key parameters are fixed elsewhere.

A useful measure is the share of functions in which the closest capable level can change implementation without prior approval from a higher level. Another is the number of mandatory approvals or vetoes required for a significant decision. A third is whether the unit can change provider, implementation standard or timetable where common minimums allow it.

The European Charter of Local Self-Government stresses that local powers should normally be full and exclusive and that local authorities should have room for initiative within the law. That is a useful orientation: measure discretion, not merely assigned tasks.

The second axis is fiscal autonomy. A local level can spend a great deal of money and remain highly dependent if most revenue arrives as earmarked funds whose use is determined by the centre. OECD measurement therefore distinguishes revenue and expenditure shares from actual tax and spending autonomy.

Measure at least four things: the share of own-source revenue, the share of revenue over which the unit can set the rate or base, the share of expenditure over which it has genuine discretion, and the share of transfers that are tightly earmarked. Record unfunded mandates separately: responsibilities moved downward while money and rules remain above.

This prevents a common optical illusion: decentralised spending without decentralised authority. If a local unit merely executes someone else’s budget under someone else’s rules, it is closer to an operating branch than an autonomous decision centre.

Axis 3: who owns critical assets?

The third axis measures ownership concentration over assets without which the system struggles to function: land, housing, productive capacity, energy, payment infrastructure, logistics or key communication channels. This continues “Monopoly, Plutocracy and the Concentration of Economic Power”, but this article no longer asks why concentration emerges; it measures how much of it exists.

Simple measures can include the share held by the largest owner, the share held by the top three or five, an HHI or similar concentration index, and the share of a critical function controlled by one legal or effective group. Connected ownership matters here; different names on the façade do not necessarily mean independent control.

It is important to distinguish concentration of ordinary wealth from concentration of critical dependency. Control over a large share of a luxury market is not the same as control over most routes to housing, energy, basic payments or local employment.

Axis 4: who controls data and the metrics themselves?

The fourth axis measures information and data concentration. This article deliberately stays narrow because “Who Controls the Data About Us?” will examine data in depth. For decentralisation measurement, it is enough to ask: who holds the only complete copy of critical data, who defines the data model, who can terminate access for others, and whether the data can be moved to another system without disproportionate loss of time or functionality.

OECD work and the EU Digital Markets Act treat portability and interoperability as mechanisms that can reduce switching costs and lock-in. That means the number of apps or providers is not sufficient. If all of them need permission from the same data gatekeeper, the information layer remains centralised.

This article therefore measures the share of critical data that affected users or local units can export in a usable format, the time and cost of migration, and the number of independent systems able to perform the same function without privileged access from a single gatekeeper.

Axis 5: how many genuinely independent supply routes exist?

The fifth axis measures supply decentralisation. “Trade and Supply Chains: Mutual Benefit or a New Dependency?” showed that several direct suppliers do not necessarily create a distributed network if they all depend on the same upstream source. Measurement therefore needs to look at least one or two layers deeper than the first contract.

Measure the share of critical supply coming from the largest route, concentration by actual origin, the number of mutually independent transport or production routes, and the time required to switch to an alternative source. OECD supply-chain resilience work treats concentration as relevant because of vulnerability to disruption, not because every global link is inherently undesirable.

The key issue is correlated failure. Three routes that all cross the same bridge, port, processor or legal intermediary are not three independent routes. Redundancy exists only when the failure of one does not also disable the others.

Axis 6: how expensive is exit or switching?

The sixth axis measures exit costs. “Institutional Competition: Does Choice Limit Power?” and “Exit, Voice, and Loyalty: What Do We Do When a System Fails?” already established that choice is real only when it can actually be used. This article converts this into a measurable question: how much time, money, data, social capital or rights does a person lose by switching provider, community, bank, platform or service operator?

OECD competition guidance pays particular attention to switching costs because they can lock users into an incumbent even where nominal alternatives exist. Measure average switching time, direct financial cost, loss of history or data, repeated identity or qualification checks, and the share of users who can move without interruption of a critical service.

This is one of the strongest tests of functional decentralisation. If leaving an institution means losing nearly everything, its power is not meaningfully constrained by choice even if competitors formally exist.

Axis 7: can the system survive the loss of one centre?

The seventh axis measures redundancy and resilience. Decentralisation is not only distribution across a map; it is also the absence of a single point of failure. NIST resilience modelling explicitly includes multiple ways of providing the same input, backup storage and portable capacity as forms of redundancy.

A practical test is scenario-based: choose the largest centre, provider, data node or funding source and remove it for 72 hours. What share of the critical function remains? How long does switching take? Is there an alternative route controlled by a genuinely different operator?

A well-decentralised system is not one in which nothing ever fails, but one in which the failure of a single node does not stop the whole. If every backup plan ultimately has to call the same centre, the redundancy is not real.

The most common measurement illusions

The first illusion is counting units. One hundred local providers mean little if all use the same financier, data system or supplier. The second is counting budgets: a large local budget does not imply autonomy if it is almost entirely earmarked. The third is counting providers without measuring switching costs.

The fourth illusion is geographic dispersion. Servers or warehouses can be spread around the world and still be under one operator. The fifth is legal dispersion: several companies or agencies can sit under common ownership or effective control. The sixth is averaging, which can hide one critically centralised axis.

Every result should therefore answer one more question: what happens if the largest centre disappears tomorrow or simply refuses to cooperate? If all routes then lead back to the same point, the decentralisation was largely optical.

A practical dashboard: twelve measures worth tracking

Measurement should also capture the corrigibility of power: who can trigger review, replacement, reauthorisation or a return of authority to a lower level. The final dashboard measure is therefore institutional rather than merely technical — how long the system takes to actually correct or replace a power-holder who drifts from the mandate.

A community, network or project does not need an academic laboratory. Once a quarter or once a year it can track twelve basic measures: (1) share of key decisions made at the closest capable level, (2) number of external vetoes over local decisions, (3) share of own-source revenue, (4) share of discretionary spending, (5) concentration of ownership of critical assets, (6) concentration of critical data, (7) share of supply from the largest independent route, (8) number of independent backup routes, (9) provider switching time, (10) total exit cost, (11) share of function surviving loss of the largest node, and (12) time to effective correction or replacement of a power-holder.

Each measure can be marked green, amber or red for its own context, but thresholds should not be invented as universal constants. Critical water supply needs different redundancy from a cultural activity; a small voluntary group needs different financial reserves from a regional network.

The most useful signal is not the absolute score but the trend. Are more decisions moving upward over the last two years? Is earmarked funding growing? Is exit becoming more expensive? Are supply routes merging? Measurement becomes an early-warning system for reconcentration.

the articles from “Monopoly, Plutocracy and the Concentration of Economic Power” through “How Do We Measure Whether Decentralisation Is Real?” together form Stage 9’s immune system against reconcentration: first identify economic power, then delegate drift, emergency exception, institutional capture, and finally test whether the system is actually distributed. The institutional arc thereby moves from principles to observable indicators.

The next step is more specific. Among all this article’s axes, one has become important enough to require its own article: data. Whoever controls data can see, classify, predict, condition access and, in digital environments, often determine who can participate at all.

“Who Controls the Data About Us?” will therefore ask: who controls the data about us? this article leaves behind the baseline test: decentralisation is real only to the extent that not merely names and buildings, but decisions, resources, information, pathways and the ability to exit are genuinely distributed.