is promoted on the reasonable ground that decisions improve when they are taken closer to the people affected.
The evidence for that proposition is strongly conditional.
Transferring a function without transferring the money produces an unfunded and a local authority that fails visibly.
Transferring money without capacity produces that nobody can evaluate.
is the variable that most programmes assume and few measure.
A district office with two accountants cannot administer a grant scheme designed for a ministry.
is the failure that attracts the most attention and is neither universal nor random.
It is strongest where local elections are uncompetitive and local media have disappeared.
Those two conditions describe a great many districts in countries that have decentralised recently.
The most reliable finding across the literature concerns what is decentralised rather than how much.
Services with visible outputs and short improve: refuse collection, local roads, primary schooling.
Services whose quality is hard to observe do not, and sometimes deteriorate.
A resident notices an uncollected bin within a day and cannot assess a water treatment standard at all.
The design implication is unglamorous and specific.
Decentralise the visible, retain the technical, and fund an that is accountable upward rather than to the authority it inspects.
Most reforms do the opposite, because the technical functions are the ones with budgets attached.
after a failure is then justified by the failure that the design produced.
The cycle has run three times in several countries and is usually described as a debate about principle.