n.V1-7.04 | CONDITIONAL CASH TRANSFERS

01| Attach conditions to the money: keep the child in school, take her to the clinic, and the transfer continues. The theory is that poor households under-invest in the future because the present is loud, and that a small price attached to future-oriented behaviour corrects it. It is the most exported anti-poverty design of the last thirty years, and the one with the most evaluation behind it.

02| Mexico's Progresa, launched in 1997, was the first national programme of the type and the reason the type exists. It was built with an evaluation attached from the start, which was almost unheard of, and the results travelled: within a decade, conditional transfers were running across Latin America and beyond, in programmes explicitly modelled on it. Renamed Oportunidades, then Prospera, it was by any ordinary measure the best-evidenced anti-poverty programme in the world.

03| In 2019 an incoming administration terminated it, characterising it as clientelist and corrupt, and replaced it with an unconditional education scholarship that dropped the health and nutrition conditionalities entirely. The evidence base did not save it. Nothing in twenty years of published evaluation had any purchase on the decision.

04| Brazil's Bolsa Família ran the other way in the same period. Beneficiary families grew from 13.2 million in December 2019 to 21.1 million in December 2023 following the programme's 2023 relaunch. Over the same window, extreme poverty fell from 5.9 to 4.4 percent of the population — the lowest recorded level since 2012 — with the national statistics agency attributing the fall to both the expansion of social assistance and improved employment.

05| The pair is the most instructive thing in this chapter, and it is not about conditionality at all. Two large national conditional transfer programmes, similar in design, similar in evidence quality, opposite in fate. What separated them was not effectiveness. Prospera was the better-evidenced programme and it died. Bolsa Família was the larger and more politically embedded one — a programme a fifth of the country lived inside, identified with a national political project — and it grew.

06| The hypothesis this suggests is uncomfortable for everyone who works on evaluation: political durability tracks coverage breadth and political identification, not measured effectiveness. A programme that reaches a fifth of the electorate acquires a constituency that defends it. A programme that reaches the poorest decile acquires a reputation for reaching the poorest decile, which is worth nothing when a government changes.

07| It is a hypothesis, not a finding. Two cases are two cases, and both are Latin American national programmes of a single type. As far as this review found, nobody has systematically studied anti-poverty programme survival across political transitions — which is odd, since survival is a precondition for every other effect a programme claims, and the sector spends heavily on measuring effects that a change of minister erases.

08| On conditionality itself the evidence is quieter than the debate. The conditions are the part that generates the administrative cost, the exclusion errors, the monitoring apparatus and most of the moral objection — the state deciding what a poor parent must do to be fed. What they buy, over and above the money, has never been cleanly separated in a national programme, because no government has run a national transfer with conditions randomly assigned. Mexico's replacement scheme dropped them without, apparently, treating the question as one that needed an answer first.

09| So the honest position: conditional transfers work, in the sense that transfers work, and they are the vehicle through which more of the world's poorest households receive money than through any other designed instrument. Whether the conditions are doing anything worth their cost is a question the field has left open while scaling the answer. And the thing that determines whether any of it survives to the next decade is not in the evaluation literature at all.