n.V1-7.12 | COMMUNITY-DRIVEN DEVELOPMENT: THE PARTICIPATION CLAIM

01| The previous chapter examined the machine. This one examines the claim the machine is built on: that convening people to decide about money durably changes who holds power locally. Not that the road gets built — that part is easy to verify — but that the village governs itself differently afterwards.

02| Casey, Glennerster and Miguel tested it in Sierra Leone with a pre-analysis plan, filed in advance, specifically to stop the researchers cherry-picking from the dozens of indicators a claim like this generates.

"positive short-run effects on local public goods and economic outcomes, but no evidence for sustained impacts on collective action, decision making, or the involvement of marginalized groups, suggesting that the intervention did not durably reshape local institutions."— Casey, Glennerster & Miguel (Reshaping Institutions: Evidence on Aid Impacts Using a Preanalysis Plan, Quarterly Journal of Economics, 2012, Vol. 127, No. 4, pp. 1755–1812)

03| The infrastructure appeared. The institutional change did not. And the pre-analysis plan is doing heavy lifting in that sentence, because a study of the same programme without one would have had thirty outcome measures to choose from and would almost certainly have found something significant to report. How much of the wider positive literature on community-driven development would survive the same discipline is a question nobody has wanted to answer.

04| Mansuri and Rao's review reaches a compatible conclusion across the whole literature, and adds two things that are harder to absorb. The first is that elite capture of induced participatory initiatives is widespread and worse in more unequal communities — that is, worst exactly where the participation was supposed to help most. Capture is the expected outcome, not the risk case.

05| The second inverts the story the sector tells itself. Participatory projects, they conclude, are not a substitute for weak states; they depend on strong central support to function at all. The bottom-up model does not route around a failing state. It requires a working one to convene it, fund it, audit it and enforce against it — which is why PNPM, running inside a functioning Indonesian administration with a real enforcement apparatus, produced more than most, and why the same design dropped into a country without that produces meetings and little else.

06| That finding should change the axis of the argument. Bottom-up versus top-down is not the useful distinction, because the bottom-up cases that work are sitting on top-down infrastructure. The useful distinction is which decisions are genuinely transferred and which only appear to be.

07| On that, the PNPM material is precise. What the design transfers is real: which projects get funded, in a room where villages rank each other's proposals and the money runs out visibly. What the design cannot transfer is the form of the deciding — the grant size, the eligible categories, the procedure, the deadline, the document format. The funder has to specify all of that in advance in order to release the money at all. So a community receives authority over the contents of a decision whose shape was fully determined before anyone arrived.

08| And procedure has a hard ceiling. A rule can compel attendance, composition and record — six delegates, at least three of them women, a signed minute. Every one of those is verifiable from a document, which is why every one of those is what gets measured. None of them is evidence that anyone spoke. The Indonesian study found existing power structures and elite control in PNPM villages largely unchanged, in a programme that ran the meetings properly.

09| What follows is not that participation is worthless. It is that participation delivers what it can deliver — better-chosen infrastructure, visible rationing, a local institutional structure that outlasts the grant — and does not deliver the thing it is usually sold on, which is a durable redistribution of local power. Sold on the first, it is a decent buy. Sold on the second, it has been tested and it does not do it.

10| The open question is the one the literature has never taken up: what design feature, if any, has been shown to prevent capture. Mansuri and Rao establish that capture is the default. Nothing in the reviewed evidence identifies an intervention that reverses it. Every programme in this family manages the problem with rules about who must be in the room, and the Sierra Leone and Indonesian results both indicate that rules about who is in the room do not reach it.