n.V1-7.06 | THE GRADUATION PACKAGE
01| If a single transfer decays and a single training fails, the obvious response is to stop delivering single things. The graduation approach does exactly that: a productive asset — typically livestock — plus consumption support while the asset matures, plus training, plus regular coaching visits, plus savings encouragement, plus basic health services. Six components, delivered together, over roughly two years, to households in extreme poverty.
02| It was evaluated in six randomised trials — Ethiopia, Ghana, Honduras, India, Pakistan, Peru — covering roughly 11,000 households, and published together. Consumption gains held at least a year past the end of implementation. Benefits exceeded costs at five of the six sites.
03| That is, on the evidence available, the strongest result in this book part. It is also the model most often left out of comparisons between anti-poverty approaches, which is a tell: a comparison that omits the best-performing available case is not comparing against the field, it is comparing against a convenient version of it.
04| Why it works is the interesting part, and the trials do not answer it. Six components were delivered as a bundle and the bundle was measured. Nothing decomposes which elements carried the effect, whether they interact additively or multiplicatively, or whether a cheaper four-component version would have done the same. This is not a small gap. The programme is expensive, the expense is the main obstacle to its use, and nobody knows which parts of it are being paid for unnecessarily.
05| There is a strong candidate mechanism, and it connects this chapter to the two before it. The graduation package is the only common design that combines a transfer large enough to plausibly cross an occupational threshold with the complementary inputs that make crossing it survivable. The Bangladeshi eleven-year panel found that households above a threshold of initial assets accumulate and move into better occupations while those below it slide back. A programme that hands over livestock, feeds the family until the livestock produces, and teaches the household to run it, is a programme designed — whether or not its designers said so — to push households over that line and hold them there while they cross.
06| If that is the mechanism, the implication is blunt: the graduation package works because of the size and type of the asset, and the other five components are scaffolding around the crossing. If it is not the mechanism, then something in the coaching or the health services or the savings discipline is doing work that nobody has isolated. Either answer would change how the model is deployed, and neither has been established.
07| The limits deserve stating plainly. "Sustained a year after the programme ended" is a real finding and it is not persistence in the sense this book part cares about. The Uganda grants looked excellent at four years and had converged by nine. Nothing in the graduation trials rules out the same trajectory; they simply have not run long enough to see it. The eleven-year Bangladeshi panel is encouraging on that point but it is a different programme in a different country.
08| The second limit is the one that recurs throughout this book part. These six trials were run by organisations that designed the model, believed in it, selected and supervised the field staff, and were being watched. Nothing here tells you what happens when a ministry runs it. Given what happened to identical teaching contracts in Kenya when the carrier changed, that is not a hypothetical concern.
09| What would settle it: component-decomposition trials, a ten-year follow-up, and one government-implemented arm. All three are expensive, none is impossible, and the absence of any of them after a decade of enthusiasm for the model says something about what the sector funds.
