n.V1-7.11 | COMMUNITY-DRIVEN DEVELOPMENT: THE BLOCK GRANT

01| Give the money to the community and let it decide what to build. Indonesia's PNPM-Rural is the largest test of that proposition ever run, and it deserves to be examined on its own terms rather than through the sector's usual argument about participation, which is the subject of the next chapter. This one is about the machine.

02| Money arrives as a block grant to a kecamatan, a sub-district of perhaps twenty villages. Under the predecessor programme the grant ran from roughly $56,000 to $165,000 a year, set by population and poverty rate — calculated elsewhere, from census data, by nobody the village would recognise. What it may be spent on is an open menu: roads, bridges, water, schools, clinics, irrigation. There is also a negative list of what will not be financed, and a rule reserving up to a quarter of the grant for women's savings and loan groups. The village chooses freely inside a space whose edges were drawn in Jakarta and Washington.

03| The economics of the programme are the strongest thing about it and are almost never quoted. Across PNPM-Rural's first two phases, block grants were $1.63 billion out of a total project cost of roughly $1.86 billion. Eighty-five percent of the money went out as grants to communities. Facilitation and training took $137.7 million, under eight percent. Implementation support and technical assistance, $71.4 million. National project management, $22.6 million. Monitoring and evaluation, about three and a half percent. Whatever else is true of this programme, it was not an institution spending money on itself.

04| One design decision inside that deserves more attention than it gets. Alongside the investment grant, each sub-district received a separate planning grant — $74.5 million across the programme — whose only purpose was to pay for the process of deciding: the meetings, the volunteer training, the printing, the travel, the cross-village audits. Deliberation was funded as a line item rather than taken out of the thing being deliberated about. That is why PNPM's meetings actually happened, and it is the single most copyable feature of the design.

05| The largest dependency is a contractor with a motorbike. Each sub-district gets a facilitator team — a social facilitator, a technical facilitator, sometimes a third for finance — hired as consultants, not civil servants. In 2009, the year of full national scale-up, 11,378 were in post against a quota of 12,752. Sub-district facilitators ran at 89 percent of quota, district facilitators at 87 percent, while provincial and national consultant posts were fully staffed. Read those in order: the further from the village, the closer to full staffing. The completion report records that vacancies were worst among the financial and technical specialists — the two roles that check whether the money and the concrete are real.

06| Then the incentive problem underneath the staffing problem. A study of marginalised groups in PNPM villages found that the programme's incentives push facilitators to complete each stage of the process correctly, with far less attention to the quality of participation inside it. The facilitator is measured on whether the meeting happened, the minutes were signed and the schedule was met. Nobody measures whether the poorest person in the room said anything. Given a shortage of staff, a heavy caseload and a deadline, a rational facilitator does the measurable thing.

07| Allocation is genuinely devolved, and this is the part of PNPM most worth taking seriously. Villages do not receive grants; they compete for them. Proposals pass a verification team, then delegates from every village in the sub-district rank all the proposals against each other, and a second assembly confirms which are funded. Delegates from twenty villages sit in a room with a fixed sum and a ranked list, and the line is drawn where the money runs out. A village that ranks low goes home with nothing, having done all the same work.

08| Two consequences follow. The process rewards proposal-writing — a document with background, objectives, benefits, local resources, an implementation plan, a maintenance plan, a technical design and a full bill of quantities, requiring literacy, numeracy, technical drawing and the confidence to defend it in public. Villages with someone who can produce that are advantaged over villages without, and that person is rarely the poorest person in the village. And competition between villages transfers the rationing decision downward without transferring the constraint: the total was fixed elsewhere, so what the assembly actually decides is who goes without.

09| The anti-corruption machinery was built from an experiment rather than an assumption, which is rare enough to note. Under the predecessor programme, a controlled trial tested what reduced losses. Raising the probability of an audit worked. Reading the audit findings aloud in a village meeting, immediately, worked considerably better. Not prosecution — publicity, delivered fast, in front of the people whose money it was. PNPM built on that by tripling audit sampling, adding complaints handling, adding financial management facilitators at district level and training local government auditors.

10| The reported result: 0.3 percent of block grants known to be affected by fraud and corruption, 57 percent of it recovered, 262 court cases, around eighty prison sentences including facilitators and officials, and about three hundred facilitators — three percent of the workforce — dismissed. Sanctions ran upward too: two provinces and twelve of some 340 districts had disbursements suspended. Most programmes of this kind cannot name a single person who went to prison.

11| How much to believe it is a separate question. The 0.3 percent is the programme measuring itself, and the phrase is "known to be affected". The same completion report notes that corruption cases were not always handled promptly, that the management information system performed poorly, and that the national consultant's field oversight was inadequate. A detection system with those weaknesses reports a low number partly because it is a detection system with those weaknesses. The honest reading is that PNPM caught 0.3 percent, prosecuted it hard and published the result — which is more than almost anyone else in this field can say, and is not the same claim.

12| The weakest step is the last one. When construction finishes, the facilitator convenes a maintenance meeting, the village forms an operations and maintenance committee, and the asset transfers to the village. The completion report records, plainly, that the maintenance committee was often composed of the same people as the implementation committee. The body that built the thing hands it to itself. Where the asset generates revenue — water, a market, irrigation — there is money to maintain it. Roads were the most popular investment in the entire programme and are the most expensive thing to maintain from community resources, and no formal institutional arrangement existed for maintaining them. The Bank's own evaluation work puts sustainability at around 43 percent for community-driven development projects against 51 percent for its projects generally: the method built specifically to produce local ownership, and therefore durability, performs below the institution's average on durability.

13| The pattern across all twelve of those paragraphs is one thing. What can be specified in advance — the grant formula, the menu, the meeting sequence, the delegate quotas, the document format, the audit rule — was specified well, and mostly worked. What cannot be specified in advance — whether the poorest person spoke, whether the road gets repaired in year four — was left to arrangements that the programme's own reports show did not hold.