A development project can spend a decade and tens of millions of dollars, report success, and still teach the world almost nothing. The Millennium Villages Project is the clearest recent example. Launched in 2005 under the economist Jeffrey Sachs, it set out to prove that a large, coordinated package of aid could pull rural Africa out of poverty. What it proved instead is that aid has no built-in way of finding out when it is wrong.
The idea was simple and ambitious. Instead of scattering money across unrelated programs, deliver everything at once: better farming inputs, clinics, schools, clean water and sanitation. The project was designed to break villages out of poverty traps, and its final evaluation covered ten sites in ten sub-Saharan countries over ten years. The planners believed that a big enough push, designed by experts, would work.
What the Evidence Shows
The project’s own final evaluation, published in The Lancet Global Health, found favorable effects on agriculture, nutrition, education, child health and water, but reported that only a third of its prespecified targets were met. The same evaluation found no significant impact on consumption-based measures of poverty, although asset ownership improved and results for individual nutrition and education outcomes were often inconclusive. That evaluation was produced by a team that included project staff.
A fully-independent evaluation of a separate, shorter project in northern Ghana, funded by the British government and implemented from 2012 to 2016, tells a harsher story. The project cost £11 million for 35 villages and up to 30,000 people, equivalent to roughly £370 per targeted resident using the 30,000-person upper bound. The evaluators found a significant effect on seven of 28 development indicators and no effect on poverty or hunger on all but one measure. Some of the gains were real. School attendance rose, more births were attended by skilled professionals, and toilets improved. But the evaluators concluded that the project did not achieve significant progress overall on reducing poverty or hunger.
Spending money on poor people will usually buy something. The question is whether it bought more than the same money would have bought under a system that has to answer to the people it serves.
No Profit, No Loss, No Signal
A business that spent £370 per customer and left them no better off would go bankrupt. A donor-funded project faces no such test. Its income comes from capitals, not villages. It is renewed on the strength of reports, and the reports are written by people whose work depends on the project continuing.
The early history of the Millennium Villages shows how this works in practice. In 2010, the economists Michael Clemens and Gabriel Demombynes showed that the project’s before-and-after comparisons overstated its effects. Using trends in the surrounding areas, they found much more modest results, and they pointed out that hand-picked sites and the lack of baseline data for comparison villages made a rigorous evaluation impossible. A claim about falling child mortality in an early paper then turned out to contain a mathematical error. Once it was corrected, the improvement looked similar to the national trend in the countries where the villages sat.
Notice who caught these problems. They were outsiders, working years later, using public data. Inside the system there was no price, no profit and no loss to say the project was underperforming. In a market, that signal arrives automatically and early. In aid, it arrives by accident, if at all.
Planners Cannot See What Villagers Know
Friedrich Hayek made the deeper point in 1945. The knowledge needed to run an economy never exists in one place. It is scattered among millions of people who each know their own land, their own prices and their own needs. No committee can collect it, however well-funded.
The package approach assumes the opposite. A team of experts decides that a village needs a particular mix of inputs, services and buildings, and supplies them together. But a farmer knows whether the road floods in the rains, who buys maize and at what price, and whether a bag of fertilizer will arrive in time to matter. A trader treats these questions as bets and loses money when wrong. A planner treats them as technical problems with technical answers and is paid either way.
The Ghana results fit this pattern. Planners can build what they can count: classrooms, latrines, clinic visits. Whether a household ends up richer depends on thousands of local decisions about what to plant, whom to sell to and what to risk. Those decisions cannot be designed from a distance.
Accountability to the Poor
None of this requires believing that the people behind the project were careless or cynical. Many of them clearly wanted to help. The problem is structural. When the payer is not the beneficiary, money tends to flow toward what looks good in a report.
The alternative is not complicated. Secure property rights let villagers keep what they build. Open trade lets them sell to whoever pays best. And when outside help does come, it should be tested the way every other good or service is tested, by whether the people receiving it would choose it with their own money and their own judgment.
The people of those villages did not need a decade-long experiment to find out whether the package worked for them. They could have told us in a season, if anyone had been built to listen. Feedback from the poor, through prices, purchases and the freedom to walk away, is the only reliable way to tell a good project from a bad one before ten years have gone.
