n.V1-4.03 | THE LARGEST EXPERIMENT EVER RUN

01| Whether communities can be handed control of development money is not a question anyone needs to speculate about. The experiment has been run, at a scale that dwarfs everything else in the field, by an institution that then paid its own researchers to assess it.

02| Between 1999 and 2011 the World Bank spent $54 billion on community-driven development. It did not retreat afterwards: the Bank currently supports around 400 such projects in more than 90 countries, worth close to $30 billion, somewhere between five and ten percent of its annual lending over the past decade. This is the largest structured attempt anyone has made to put communities in charge of their own development spending, and therefore the best evidence available on whether that can be done from outside.

03| The machinery is consistent enough to describe in a paragraph. Communities get training, facilitation and information, and help preparing proposals in a participatory way. They compete for block grants. Winners implement the project themselves — managing, operating, monitoring, maintaining, usually contributing part of the cost. The money goes to the community rather than through the usual government channels.

04| One design element repays attention, because it is where the approach shows its structure. Projects typically spend between three and twelve months, and around a third of the total budget, training the community to make decisions inclusively and manage resources transparently — and the grant is conditional on adopting those practices.

05| Set that against the distinction in the previous chapter. Money transfers, genuinely, along with responsibility for spending it. What does not transfer is the form of the decision-making, which is designed elsewhere, taught on a schedule, and made a condition of receiving the money. The community owns the project. It does not own the terms on which it may own the project.

06| That is not hypocrisy. It is a defensible choice, made because handing cash to existing local power structures has obvious risks. But it produces the question the field's most careful researchers ended up asking.

07| In 2013 the Bank published Localizing Development: Does Participation Work? by Ghazala Mansuri and Vijayendra Rao — a review of more than 500 studies, concentrated on large participatory programmes evaluated against representative samples of the populations they targeted. Not an outside critique. The institution examining its own flagship.

08| The headline is quoted often and deserves to be: after $54 billion, the evidence base for the approach's success in reducing poverty and giving voice to the marginalised is surprisingly weak.

09| The specifics are more useful. Most community development projects are dominated by local elites, and Mansuri and Rao do not treat this as a series of programmes going wrong:

"The poor are often excluded from the process of consensus-building."— Ghazala Mansuri & Vijayendra Rao (Localizing Development: Does Participation Work?, World Bank Policy Research Report, 2013)

Community development may be "inherently subject to elite capture because of the entrenched influence of local elites."— Mansuri & Rao (Localizing Development, 2013)

10| Read plainly, that is elite capture as the expected outcome of the method rather than a risk to be monitored on the side.

11| And the domination is not evenly spread. Targeting and project quality are markedly worse in more unequal communities. The structure of that finding is the uncomfortable part: the mechanism performs worst exactly where inequality is greatest, which is to say where a redistributive mechanism is most needed. Where a community is relatively equal and cohesive, participation works. Where it is not, participation reproduces whatever hierarchy is already there. A tool built to reach the poorest works least well where the poorest are furthest from power.

12| Mansuri put the underlying problem in the title of a companion paper: can participation be induced? It is not a rhetorical question. Participation a community generates for itself, and participation an external agency creates on a project schedule with a training budget, may be two different things wearing one name — and the second may not do what the first does.

13| Before this becomes a straightforward indictment, there is a finding that cuts the other way and should not be buried. Rao and Ibáñez studied a social fund in Jamaica and found that despite elites dominating decision-making, more than 80 percent of participants were satisfied with the outcomes. Local elites are frequently the people with the time, literacy, contacts and confidence to get something built. A project they control may still deliver what the community wanted.

14| That does not dissolve the problem. Satisfaction is not equitable targeting, and the people least likely to be surveyed are the people least likely to have benefited. But "elites dominated it" and "it failed" are separate claims, and the evidence supports the first far more strongly than the second.

15| Two further findings concern durability, and these are the damaging ones. The Bank's own evaluation work finds sustainability likely for around 43 percent of community-driven development projects, against 51 percent for Bank projects generally. The approach designed explicitly to produce local ownership, and therefore lasting results, performs below the institution's average on lasting results.

16| And the effect on local institutions may be negative rather than neutral. The concern documented in the literature is that NGO-delivered training and the experience of the project do not durably strengthen local governance and may weaken it, because the project routes money and authority around it for the duration and then stops. The subproject cycle is simply too short to build capacity where little existed — again, the case of greatest need. Facilitator quality compounds it: facilitators are frequently poorly trained and inexperienced, and this worsens the faster a programme scales, so the quality of the ingredient the whole design depends on degrades in proportion to the ambition of the rollout.

17| Put it together. Two decades, tens of billions, an evidence base the Bank's own researchers call surprisingly weak, elite domination as the norm, worse performance in unequal communities, below-average sustainability, and possible harm to local government capacity. And it continues, at scale. A 2024 study asks in its title whether the Bank's community-driven development is worthy of continued support, and calls its persistence a paradox.

18| An explanation suggests itself, and we mark it as inference rather than finding, because nothing here demonstrates it. Community-driven development solves a problem for a lending institution that has nothing to do with community ownership. It moves large sums quickly, visibly, at scale. It does not require a functioning local government to receive the money, which matters most in exactly the countries where the Bank most wants to lend. It produces countable outputs and a participation narrative at the same time. An instrument with those properties would be expected to survive weak evidence — not because anyone is deceiving anyone, but because the evidence is not what determines whether it continues.

19| What the experiment establishes with confidence is narrower, and worth saying plainly. Communities can be handed money and will spend it, largely on things they say they wanted. Whether an outside agency can manufacture the ownership that makes the spending last is the question the largest trial ever run has not answered in twenty-five years.