Civics · Law · Geopolitics: Local and Urban Governance

The key distinction is where local power comes from. In a federal system, the powers of states or provinces derive from the constitution itself and cannot be withdrawn unilaterally by the centre. In a unitary system, local power is delegated by the centre and can, in law, be amended or revoked.

But the label explains little. What actually sets the degree of decentralization is three independent channels: legislative power (who writes the rules), fiscal power (who may spend), and appointment power (to whom local officials answer). They combine freely, which is why "fiscally very decentralized, administratively very centralized" is such a common real-world form.

The general principle for dividing work is subsidiarity: what a lower tier can handle should not be moved up. The EU wrote it into the 1992 Maastricht Treaty. The principle is clear, but the judgment of what a lower tier "cannot handle" is itself the contested point.

Strongest case for decentralization: the information is local. The centre cannot know each community's preferences and constraints, so imposing uniformity allocates resources by an average; and local decision-makers sit closer to the consequences, shortening the accountability chain.

Strongest case for centralization: not every problem closes locally. Pollution, epidemics and cross-boundary transport carry externalities; redistribution requires pooling over a wide area; common standards lower transaction costs in a national market. Bardhan and Mookherjee also point to the risk of local elite capture: where power concentrates in a few local notables, devolution may hand resources to the tier that is easiest to capture.

Trade-off: decentralization buys fit and experimentation and sells uniformity and equalization; centralization does the reverse.

Equating "federal" with "powerful localities." Some federations fund localities mostly through central transfers, leaving them substantively dependent; some unitary states grant considerable local autonomy. Fiscal and appointment power tell you more than the name of the system.

Power without money is nominal power. Oates's 1972 decentralization theorem sets the benchmark: where preferences differ across places and a service has no significant externalities or economies of scale, local provision matched to local preferences beats a uniform national level. Tiebout added the mechanism in 1956: residents can vote with their feet, moving to the jurisdiction whose tax-and-service bundle suits them best.

The most common structural problem is vertical fiscal imbalance: the tax bases that are easy to collect suit central collection, while the expensive functions — education, health — suit local delivery. Spending responsibility descends faster than revenue authority, and the gap is filled by transfers. General transfers respect local preferences but are hard for the centre to steer; earmarked transfers are steerable but distort local priorities.

A further failure mode is the soft budget constraint (Kornai): if a locality believes the level above will cover a blow-up, its rational level of borrowing is higher than if it bore the consequences itself. That is an incentive problem, not a moral one.

• China's 1994 tax-sharing reform: the central share of national fiscal revenue rose from roughly 22% to about 56%. Spending responsibility, however, stayed largely local; the gap is met by transfers and other sources, and local reliance on land-related revenue has been a long-running topic of debate.
• Brazil: state debt was restructured by the federal government repeatedly — in 1989, 1993 and 1997 — a textbook soft-budget cycle. The 2000 Fiscal Responsibility Law set ceilings on debt and personnel spending benchmarked to each tier's own revenue.
• Germany: interstate fiscal equalization moves resources from richer to poorer Länder, keeping public services broadly comparable; the cost is lasting resentment among net contributors and a running argument over how much incentive for own effort survives.

Strongest case for competition: under a hard budget constraint, competition for tax base restrains over-extraction — if a government abuses its power, firms and residents leave. That is self-correction without supervision from above.

Strongest case for equalization: competition presupposes mobility, and those who most need public services are the least able to move. Making localities live off their own revenue means thin-tax-base regions deliver worse schooling and health care, hardening the gap across generations; tax competition can also become a race to the bottom.

Trade-off: competition buys efficiency at the price of inequality; equalization buys a floor at the price of weakening the incentive for local effort.

Assuming fiscal decentralization necessarily means smaller government or greater efficiency. It changes who decides, not necessarily total size; absent a hard budget constraint it can raise both total spending and total debt. A second misconception treats the volume of transfers as a measure of decentralization — the more a locality depends on transfers, the less autonomy it actually has.

Cities concentrate population, tax base and the sharpest public problems, yet their legal standing is often the weakest. Dillon's Rule in the American common-law tradition states it most bluntly: a municipal corporation holds only the powers expressly granted to it, and doubtful grants are read narrowly. In Hunter v. City of Pittsburgh (1907) the U.S. Supreme Court confirmed that a state's authority over its municipalities is near-absolute — annexation was upheld even though most voters in the absorbed city had opposed it.

The countervailing design is home rule: a sphere of local autonomy marked out in advance, inside which a city need not seek permission case by case. But even under a home-rule charter, the level above can reclaim specific subjects by preemption.

Strongest case for municipal autonomy: problem density is urban — housing, commuting and public health take a different shape in a city than in suburbs or countryside, so a uniform rule must fit one of them badly. Cities are also a natural site for policy innovation.

Strongest case for higher-level uniformity: the costs of city decisions often spill over, and a patchwork of labour, tax and building rules raises the cost of operating across boundaries. Housing matters most here: when permitting sits with the smallest unit, neighbours opposed to new building face the lowest organizing cost and the most direct benefit, so supply falls systematically short — and the cost lands on people who have not yet moved in and hold no vote locally.

Trade-off: most countries split by subject — housing and environmental standards move up, service delivery and daily management move down.

"Local government is closer to the people, therefore more democratic." Closeness is not representativeness: turnout in local elections is typically far below national elections, media scrutiny is thinner, and organized local interests carry proportionally more weight in a small constituency. Smaller scale lowers the cost of participating — and the cost of capturing.

These are two different things: participation answers "who decides," experimentation answers "how do we learn."

The characteristic form of participation is participatory budgeting: part of the budget is allocated by resident assemblies through deliberation and ranking. Its value operates on three levels — gathering dispersed information, raising legitimacy, and lowering resistance at the implementation stage.

Experimentation draws its value from parallel jurisdictions: the same problem handled differently in different places, so effects can be compared cheaply and what works spreads. Justice Brandeis's much-quoted 1932 dissent put it as a single state serving as a "laboratory," trying a new policy without risking the whole country. The structure is the same as a staged rollout.

• Porto Alegre, Brazil: participatory budgeting from 1989. Fewer than a thousand people took part in the first two years; as the first resident-chosen projects were delivered, attendance rose to roughly eight thousand by 1992 — participation is grown by promises that get kept. Gonçalves's study in World Development (2014), covering Brazilian municipalities from 1990 to 2004, finds that adopters directed a larger share of budget to sanitation and health services, accompanied by falling infant mortality.
• India: the 73rd and 74th Constitutional Amendments of 1992 gave village and urban local bodies constitutional status and reserved at least one-third of seats and chair positions for women. Kerala's 1996 People's Plan Campaign handed roughly 35–40% of the state's annual plan funds to 1,214 local governments to allocate themselves; scholarly assessments describe real progress rather than an unqualified success.
• Switzerland: referendums and initiatives are highly institutionalized at cantonal and municipal level, letting residents reject local budgets and construction projects directly. The cost is a long decision cycle; the benefit is that conflict is front-loaded.

Strongest case for participation: the real information sits with users and cannot be replaced by an agency's statistics; and the legitimacy participation produces is cashed in later as lower resistance during implementation.

Strongest case for representation and expertise: participants are a self-selected sample — those with time and organization show up most, and a majority in the room is not a majority in the community. Complex questions (network planning, debt structure) need professional judgment, and voting cannot manufacture a technical answer. The time cost of attending also falls hardest on low-wage workers.

Trade-off: the common compromise confines participation to setting priorities, leaving technical design and finance to professional bodies.

Reading "the pilot succeeded" as "this can be scaled." Pilot sites are rarely chosen at random — they tend to be places with better conditions and more attention from above, and they receive extra resources and room for error. That layer of selection bias makes pilot results systematically better than full rollout. The question is not "did it work?" but "do the conditions it relied on exist elsewhere?"