Documents recently made public by the UK government reveal the cost of poverty reduction in the Millennium Villages Project, a self-described "solution to extreme poverty" in African villages created by Columbia University Professor Jeffrey Sachs. The project costs at least US$12,000 per household that it lifts from poverty—about 34 times the annual incomes of those households. This highlights once again the importance of independent and transparent evaluation of development projects.
The project has called this number a distortion of the truth. Such behavior is sad, since the $12,000 figure comes straight from the project’s own documents about a new Millennium Village site in Ghana. The business plan for that site, based on the project’s experience at 14 previous sites, states the full cost (page 13):
“The total direct costs of the project are expected to be $27.1 million, which is approximately £17 million.”
The following page states the “poverty reduction” goal at this site:
“The Millennium Village aims to achieve substantial poverty reduction for up to 2,250 households (around 11,000 people).”
This is perfectly clear: For each household the project says it can lift out of poverty at this site, the cost is $12,000—or more (it’s “up to” 2,250 households).
Why does the project call a verbatim quotation of its own documents a “distortion”? Because the project says it will affect, to some degree large or small, 6,000 households (about 30,000 people). Even those other 3,750 households that remain in extreme poverty at the end of the project will receive amenities like clean water. So is it a distortion to point out the extraordinary cost of poverty reduction via this project? It is not, for many reasons.
- The statement is true: The project foresees a cost of at least $12,000 per household that it believes it can lift out of poverty.
- The project is an anti-poverty project. For many years, the project has told its funders that it is a “solution to extreme poverty”, that lifts families out of poverty “in five years’ time”, and that it accomplishes this by sparking “self-sustaining economic growth” that lastingly releases villages from a “poverty trap”. It should therefore be judged by its ability to reduce poverty. If the project has now morphed into a charity to deliver clean water for a few years, leaving people drinking clean water in extreme poverty, this would constitute an admission of failure of its founding principle as a project to quickly achieve permanent poverty reduction.
- Perhaps the project believes that only part of the $12,000 spent to reduce poverty in each household was meant to reduce poverty, and other parts of the $12,000 were meant to do other things, like provide clean water. This means that the project could have caused more reduction of poverty, but chose to allocate resources in a way that caused less poverty reduction. Providing people with clean water is kind and generous, of course, and I admire the many charities that do it. But charity is not development. It would certainly be strange for a poverty reduction project to choose to cause less poverty reduction than it could have caused.
- The continued existence of the non-poverty effects of the project, after the short-term burst of external aid stops, depends critically on the degree of local economic development. Anyone spending large sums of outside money can temporarily deliver amenities. But if people do not have enough income to pay the continuing costs of clean water, it will soon be dirty again.
The project’s enormous expenditures matter because much more could have been done for poverty with other uses of the same money, if the project had never existed. Spending $12,000 per household lifted out of poverty means spending about 34 times the annual income of the people to be lifted from poverty.* The same $12,000 in a bank account at 5% interest would yield $600, every year, year after year, forever. That interest, given to the households as cash, would cause their incomes to nearly triple, permanently and certainly.** I stress: this effect on income would be permanent.
But what about the non-income effects of the project, like effects on health and education? Cash transfers have those effects too. When rural Africans have more money, they spend it on things that improve health and education. A comprehensive study by the UK’s Department for International Development documents this fact, in Africa and in poor communities around the world. And those effects are from comparatively small cash transfers. Transfers large enough to triple household income would have huge effects on health and education.
Can the Millennium Villages Project permanently triple the incomes of many people, or even any people, at the sites where it works? We can’t even say whether or not that has happened temporarily, much less permanently, because the project has never released any data about what has happened to the incomes of the people it experiments on. The project has been collecting income data for the past seven years. But it hasn’t released any data about how incomes have changed over time. It has chosen to release other data on changes in non-income social indicators, but not the income data.
The Millennium Villages Project is probably causing short-term improvements in things like access to clean water and skilled birth attendance at the sites it works in. My co-author and I showed this in a paper (available here, peer-reviewed version here), while we revealed that the project typically says those short-term effects are roughly twice as large as they really are. But causing short-term improvement of some kind with charity does not make a development project successful. A successful development project does more with the money than an alternative project can—otherwise diverting money to the former makes everyone worse off than if that project didn’t exist. And that has ethical dimensions.
The project’s public obfuscation of basic facts about its work underscores once again the importance of independent evaluation in development. It is especially salient given that the project claims to base its evaluation on “peer-reviewed science” produced by professors at Columbia University. Compliance with the standards set by the International Aid Transparency Initiative is critical to proper, independent impact evaluation and learning.
Update: The project has issued a statement that the above post contains “errors” that must be “corrected”. I refer readers to the unchanged post above, which transparently documents its sources: the project’s own documents, released by an outside agency. The project’s other claims, such as the claim that it’s wrong to include overhead costs in the cost of a project, or that a “solution to extreme poverty” should not be judged by its poverty effects, require little comment. I’ll stop at saying that such claims highlight the inability of most employees to objectively evaluate their own employers, just as all of us are unable to objectively evaluate our own employers. Internal, opaque evaluations that selectively release confidential data cannot substitute for independence and transparency.
* Most of the new intervention site in Ghana is located in the Upper East Region. In the most recent survey data available, the annual income of the average household in Upper East region of Ghana is US$347 per year—that is, 616 Ghanaian cedis per year, from page 107 here, converted at today’s exchange rate of 0.563 dollars/cedi. $12,000 per household is over 34 times household annual income of $347. Today’s exchange rate is the right one to use; it’s the one that tells you how much could be done in Ghana this year with the amount of dollars to be spent this year.
** I’ve been asked how the $12,000 figure squares with the project’s claim that it spends $160 per year per person. The answer is that the numbers reflect exactly the same large expenditure, but one focuses on spending, the other focuses on poverty impact. The $160 figure represents 1) on-site spending only, not the full project cost, 2) spending every year, during the first 5 years only, and 3) spending per man, woman, and child, no matter to what large or small degree the project helps that person. You can get from one number to the other like this: Start with $160/person/year. Add 20% to include off-site expenditures like project overhead, travel, and in-kind contributions from corporate partners (the 20% estimate comes from the business plan): That’s $192/person/year. Since the average household contains 5 people, that’s $960/household/year, and multiply by five for a total expenditure of $4,800/household over the intensive first five years of the project. Add 25% more to include expenditures outside of the intensive 5-year period, such as start-up costs and expenditures during the less-intensive, second five-year ‘consolidation’ period: total $6,000/household. That’s what the project spends to “meet the Millennium Development Goals.” The first Millennium Development Goal is to cut poverty in half, so if half of the targeted households leave poverty, that means the cost is $6,000/(0.5)=$12,000 per household that is lifted out of poverty. This assumes that the project actually can lift half of the households out of poverty, a claim for which it has published no evidence in its eight years of operation.