n.V1-7.21 | CASH BENCHMARKING
01| A method, and on the argument of this book part a solution in its own right — because it is the only thing anyone has built that lets two incompatible programmes be compared without first agreeing what good means.
02| The idea is a shortcut around an unsolved theoretical problem. Rather than construct a common unit of human benefit and score everything against it, take the simplest intervention available — handing over money — cost it exactly, and ask whether the programme under examination beats it for the same spend. Cash becomes the numeraire. No weighting scheme, no philosophy, no committee.
03| USAID commissioned the test on its own programmes, which took some nerve. McIntosh and Zeitlin randomised poor, underemployed Rwandan youth across three arms: a control group, a USAID workforce readiness programme, and unconditional cash transfers costed to equal the programme's per-beneficiary cost.
04| The training programme worked. It improved productive hours, assets, savings and subjective wellbeing — a respectable result that, published alone, would have justified the programme.
05| Cost-equivalent cash moved all of those, and also moved consumption, income and wealth. In the head-to-head comparison cash proved superior across a range of economic outcomes. Training beat cash on exactly one measure: business knowledge. And there was little evidence of complementarity — combining human and physical capital did not produce more than either alone.
06| The companion study benchmarked an integrated nutrition and WASH programme the same way. Over thirteen months neither the programme nor the cost-equivalent transfer improved child growth. A substantially larger transfer of $517 did improve consumption and investment, and produced modest gains in dietary diversity and child growth.
07| Read those two together and they say different things, which is why the method is worth more than either result. The first says a professionally designed, professionally delivered programme was beaten by handing over the same money. The second says that at the tested budget nothing worked, and what mattered was the size of the transfer rather than its type — the threshold logic again, arriving from a third direction.
08| What benchmarking does to this field is disciplinary rather than analytical. It converts "our programme improved outcomes" — which almost any programme can demonstrate against no programme — into "our programme beat the money it cost", which most cannot. Every result in this book part that was measured against a control group receiving nothing is a weaker claim than it appears, because the relevant comparison was never nothing. It was the same budget in cash.
09| There is a real methodological objection and it should not be dismissed. Cash is a per-household instrument. An intervention whose benefits are collective — a road, a water system, an accountability mechanism, a functioning local institution — cannot be detected by a design that measures household-level outcomes among people who received per-household transfers. Benchmarking may systematically disadvantage anything whose returns accrue at community or system level, and that objection has not been resolved.
10| But the objection has a condition attached, and the condition is the whole point. If a programme claims benefits of a kind that per-household benchmarking cannot detect, it has to say so before the trial and specify what would detect them. Said afterwards, in explanation of a poor result, it is not an argument. It is an excuse with a methodology section.
11| Open, and the single most decision-relevant unrun study identified anywhere in this book part: what happens when a coordination or multi-service intervention — the kind that claims exactly those collective, system-level benefits — is benchmarked against cost-equivalent cash.
