The centre can see the whole map. The local operator can see the blocked drain.
That difference explains one of civilisation’s recurring governance problems.
Large systems need common rules, shared infrastructure, pooled resources and decisions that account for effects across regions. Local systems possess detailed knowledge, short feedback loops and the ability to adapt to conditions that no national spreadsheet can fully represent.
The question is therefore not whether centralisation or decentralisation is universally better. It is which responsibilities should be coordinated at which scale, with which resources, information and accountability.
This article uses fictional examples and general institutional analysis. It is not a recommendation for any particular constitution, country or local-government arrangement. OECD research is used as an external reference for the observation that decentralisation outcomes depend substantially on how responsibilities and resources are assigned across levels of government. [1]
The same problem changes when the scale changes
Imagine a fictional city district where one road floods after intense rain. A local team may know the exact drain, the nearby construction works and the households most affected. That knowledge is valuable for immediate diagnosis.
Now imagine that the same rainfall affects twelve districts connected to one river system. Decisions in one district can alter water flow elsewhere. The problem is no longer purely local.
Scale therefore changes the unit of coordination. A local issue can remain locally manageable until its causes or consequences cross boundaries.
The useful design principle is not to move every decision upward when one spillover appears. It is to identify the part of the problem that genuinely requires wider coordination while leaving locally specific work close to the information needed to perform it.
Central coordination is powerful when uniformity matters
Some functions become easier when everyone follows the same basic standard.
Weights and measures, technical interfaces, national identification systems, major transport standards and some safety rules gain value from compatibility.
If every locality uses an incompatible railway gauge or a different definition for the same national statistic, coordination becomes expensive.
A central standard can therefore reduce duplication and uncertainty.
But the existence of one standard does not mean every operational decision belongs at the centre. A common building code can coexist with locally organised inspection. A national curriculum framework can coexist with professional judgement in individual classrooms.
Local autonomy is powerful when context matters
Local actors often know details that are expensive or impossible to transmit upward in real time.
A neighbourhood team may know which route becomes congested after school dismissal. A local clinic may understand which languages patients actually use. A municipal engineer may know that two assets listed separately in a central register are functionally connected.
Local autonomy allows those actors to respond without waiting for every decision to travel through a distant hierarchy.
The strength is not merely speed. It is the ability to incorporate context that a central rule cannot anticipate completely.
The weakness is that local decisions can become inconsistent, under-resourced or blind to effects outside the locality. Autonomy is therefore a capability that also needs boundaries.
Responsibilities must be assigned with the resources needed to perform them
Giving a local authority responsibility without money, staff, data or legal authority does not create decentralisation in a meaningful operational sense.
It creates an obligation without sufficient capability.
OECD work on assigning responsibilities across levels of government emphasises that the effectiveness of decentralisation depends on how responsibilities are designed and implemented. [1] Responsibilities, resources and accountability need to fit together.
A central government can formally transfer a task while retaining the information or funding required to perform it. Conversely, money can be transferred without a clear definition of the outcome the receiving body is responsible for.
The strongest design makes the assignment legible: who decides, who pays, who operates, who measures and who answers when performance fails.
One task can contain several levels of responsibility
Consider a fictional public-bus system.
A central authority might set safety requirements, integrated payment standards and strategic network goals. A regional body might coordinate routes that cross municipalities. A local operator might adjust stop management and service information based on observed conditions.
None of those levels needs to own every decision.
The point is not to maximise the number of layers. It is to place each responsibility where the relevant information, external effects and economies of scale can be managed most effectively.
This is multi-level governance as an operating problem rather than an organisational chart.
Ambiguous responsibility creates the accountability gap
Imagine a resident reports a damaged footpath. The city says maintenance belongs to the district. The district says the asset is nationally owned. The national agency says local access works are the city’s responsibility.
Every statement might contain part of the truth.
The resident still has a broken footpath.
Ambiguous boundaries become dangerous when institutions use them to transfer accountability rather than coordinate it.
The remedy is not necessarily one organisation owning everything. It is a defined route for resolving boundary cases so that a disagreement about jurisdiction does not become a permanent stop in the service.
This connects to How Civilisations Turn Decisions into Public Services: responsibility for the citizen’s journey must survive organisational handovers.
Subsidiarity is a question, not a slogan
A common intuition in multi-level systems is that decisions should remain close to the people affected unless there is a strong reason to coordinate at a higher level.
That intuition is useful, but incomplete.
Some local decisions create costs for neighbours. Some require expertise that is efficient to pool. Some depend on infrastructure that spans large regions. Some need uniform rights that should not vary simply because a person crosses a municipal border.
A practical subsidiarity test therefore asks: what can the lower level perform adequately, and what problem appears if the decision remains there?
The higher level should solve the scaling problem, not automatically absorb the entire function.
Spillovers are a reason to coordinate upward
A spillover occurs when a decision affects people outside the decision-maker’s boundary.
Air pollution, river management, transport networks, infectious disease and economic development can all create cross-boundary effects.
If one locality receives the benefits while another carries the costs, purely local decision-making may underrepresent the wider system.
That is a reason for broader coordination.
But spillovers do not automatically establish that every operational choice should become central. A regional environmental target can coexist with local implementation choices. The governance design should be no larger than the problem requires.
Economies of scale are another reason to coordinate
Some capabilities are expensive to reproduce in every locality.
Specialised laboratories, national data infrastructure, emergency reserves or complex procurement systems may become more efficient when shared.
Pooling capacity can raise quality and reduce duplication.
Yet large shared systems can also become distant, slow or poorly adapted to small communities.
The comparison therefore needs both sides: what cost is saved by central provision, and what local knowledge or responsiveness is lost?
Local experimentation can make the whole civilisation smarter
When every locality is required to use one untested method, a design error can propagate everywhere.
Allowing bounded local experimentation creates variation. Different approaches can be compared.
If results are measured consistently and shared, successful local innovations can spread.
This is one reason decentralisation can function as a learning architecture, not only a political arrangement.
But experimentation needs boundaries. Local freedom should not mean ignoring basic rights, safety requirements or interoperability where those constraints matter.
The strongest model combines stable common floors with room for learning above them.
A common standard can enable local freedom
Standardisation and autonomy are often treated as opposites.
They can support one another.
A common data format can allow local agencies to build different services while still exchanging information. A national qualification framework can allow institutions to design different programmes while preserving recognisable outcomes. A common technical connector can support many local suppliers.
The design question is which layer should be standardised.
Standardise too little and systems cannot interoperate. Standardise too much and useful adaptation disappears.
Local knowledge should be transmitted upward without being flattened
A central dashboard often needs comparable categories.
Comparison requires simplification.
The danger is that unusual local conditions disappear because they do not fit the standard field.
A good reporting system preserves a standard core and an exception path. The centre receives comparable information without forcing every local case into a misleading box.
This is the governance version of the knowledge problem discussed in How Civilisations Coordinate Knowledge Across Institutions.
Central knowledge should travel downward with enough explanation to be usable
Local autonomy does not mean local actors should rediscover national evidence independently.
Central institutions can provide models, standards, research, legal interpretation and shared services.
The transfer becomes weak when the centre sends only an instruction without the rationale, conditions or data needed for local judgement.
If circumstances differ from the assumed model, local actors need a way to identify that mismatch and ask for review.
Information should therefore flow in both directions: broad evidence downward, observed exceptions upward.
Fiscal design shapes real autonomy
An institution dependent on another level for every discretionary expenditure may possess formal authority but little practical autonomy.
Conversely, a locality with wide revenue powers may make decisions that affect national fiscal stability or equity.
Multi-level governance therefore includes the design of revenue, transfers, borrowing rules and spending responsibilities.
This article does not prescribe a fiscal model. The civilisational point is simpler: decision rights and resource rights must be analysed together.
Equal treatment is not always identical treatment
Suppose national policy provides the same budget per resident to every district. That is numerically equal.
One district may face far higher transport costs or a much older infrastructure base.
Identical inputs can therefore produce unequal capability.
The reverse problem also exists. Highly customised allocations can become opaque and politically difficult to justify.
A mature system explains which differences are relevant and why. Fairness requires a rule that can account for legitimate variation without becoming arbitrary.
National rights can protect people from local exclusion
Local autonomy can strengthen participation and adaptation.
It can also allow local majorities or powerful groups to exclude minorities.
This is one reason some rights and minimum standards are placed above the local level.
Uniform floors can protect people from losing basic entitlements simply because of where they live.
The design tension is not between freedom and control in the abstract. It is between different kinds of freedom: local collective choice and individual protection under wider rules.
Centralisation can hide local failure
A national average can improve while one region deteriorates.
If the centre monitors only aggregate performance, local problems may disappear statistically.
Disaggregated data and local reporting are therefore essential to central governance.
Centralisation without local visibility creates a paradox: the system gains the ability to coordinate widely while losing the ability to see uneven failure.
This connects to How Civilisations Measure Progress and Decline: averages need distributions, and system-level success needs local inspection.
Decentralisation can hide system-wide failure
The opposite problem is equally real.
Every locality may optimise its own performance while the wider network becomes incompatible.
One district improves traffic flow by redirecting congestion to another. One hospital reduces waiting by transferring difficult cases elsewhere. One region competes aggressively for scarce staff, weakening neighbouring services.
Local success can create system failure when spillovers are ignored.
That is why the higher level needs visibility into the network, not merely authority over the parts.
Crisis changes the preferred balance temporarily
Emergencies can require faster central coordination of scarce resources, common information and priority decisions.
At the same time, local adaptation becomes even more important because conditions vary rapidly.
A rigidly central response may be slow to recognise local exceptions. A completely decentralised response may compete for the same scarce resource or send contradictory messages.
The best balance can therefore change with the state of the system.
Emergency powers should not automatically become permanent institutional architecture. After the crisis, the distribution of responsibility should be reviewed again.
Polycentric systems contain several centres of decision-making
A civilisation does not need to choose between one centre and complete fragmentation.
It can contain multiple centres with overlapping but defined responsibilities.
Local governments, regional bodies, national agencies, courts, professional institutions and civil-society organisations can each govern different aspects of the same social system.
The design challenge is the interface between them.
Multiple centres increase adaptability only when rules for coordination, conflict resolution and information sharing exist.
OECD multi-level governance research emphasises design over slogans
OECD studies of multi-level governance describe reforms that reorganise powers, responsibilities and resources across levels of government and note that outcomes depend on how these arrangements are designed and implemented. [2]
This is an important caution against treating decentralisation as a single policy variable.
Two countries can both be described as decentralised while assigning taxation, education, transport and planning responsibilities very differently.
The label tells us little without the architecture beneath it.
A practical scale-assignment test
For any public function, ask:
Who has the most relevant information?
Who is affected by the decision?
Do important effects cross local boundaries?
Does the function benefit from a common standard?
Does it require scarce expertise that is efficient to pool?
Can local variation improve performance?
What minimum rights or protections should remain uniform?
Who has the money, staff and legal authority to perform the task?
Who measures performance?
Who resolves disagreements between levels?
What happens when one level fails?
Do not move the box; move the capability
Institutional reform often redraws an organisation chart.
The real question is whether capability moved with it.
If a local body receives responsibility, did staff, data, systems and funding move too?
If a central body takes over a fragmented function, did it also acquire the local knowledge needed to operate it?
A reform that changes the label on the responsible institution without changing the capability chain may produce administrative movement without practical improvement.
The best level can differ by stage
Policy design, financing, implementation, monitoring and appeal do not all need to sit at the same level.
A national government can fund a programme, local organisations can deliver it, an independent body can evaluate it and courts can resolve rights disputes.
This is not automatically inefficient complexity.
It can be a deliberate separation of roles.
The question is whether the interfaces remain clear enough that the citizen does not have to understand the whole architecture to receive the service.
Final answer
How do civilisations balance central coordination and local autonomy?
They centralise what requires common standards, pooled capability, system-wide visibility or management of cross-boundary effects.
They localise what benefits from context, rapid feedback, adaptation and direct knowledge of people and place.
They align responsibility with money, staff, information and authority.
They create common floors without forcing every locality into identical operation.
And they build routes for information and accountability to travel both upward and downward.
The strongest civilisation is neither perfectly centralised nor perfectly decentralised. It places each decision at the scale where the relevant knowledge and consequences can actually be governed.
Sources and scope
[1] OECD, Dorothée Allain-Dupré, Assigning Responsibilities Across Levels of Government: Trends, Challenges and Guidelines for Policy-Makers, 2018. Used for the general proposition that decentralisation outcomes depend on the assignment and design of responsibilities across levels of government.
[2] OECD, Multi-level Governance Reforms: Overview of OECD Country Experiences, 2017. Used as an institutional reference for reforms involving powers, responsibilities, resources and territorial structures. The fictional examples and scale-assignment test are original teaching applications, not OECD prescriptions for a particular country.
Continue through the Civilisation Atlas
Return to the Civilisation Atlas, connect scale to How Civilisations Govern Shared Resources, or follow implementation through How Civilisations Turn Decisions into Public Services.