The recent exchange of comments on www.wahenga.net has eloquently re-opened the debate on poverty-targeting of social transfers – in a most interesting and stimulating manner.

On the one hand, we have had Stephen Devereux making the point, clearly uncontentious at face value, that "a given resource envelope will have five times more impact on poverty if it is disbursed to the poorest 20% than if it is thinly spread over an entire population".

On the other hand we have had Charles Knox responding that such accurate targeting is impossible in the real world, and that the best way to "reach" the poor may not be to "target" them, but rather to take a more universal approach to poverty reduction, for example through social pensions and child benefits that are not means-tested. This recognises that such transfers may then also have a promotive and preventative role, as well as a purely protective one.

To choose between these two approaches, the decisive factor is therefore the feasibility of targeting accurately. This involves being able to make an accurate assessment of the degree of every individual’s (or every household’s) poverty, ideally at a national scale and often – in sub-Saharan Africa – in a context where considerably more than half of all individuals (and households) can be categorised as "poor". And the usual mechanism for attempting to do this – again limiting ourselves to sub-Saharan Africa – is through community-based targeting: in other words through asking the community members themselves to identify who are "the poorest of the poor".

Is such community-based targeting a sufficiently reliable method to be able to endorse with confidence the poverty-targeting approach so devilishly advocated by Stephen Devereux? In recent months, www.wahenga.net has reported on a study in Malawi that concluded: "the relationship between income and household selection to receive the [social cash transfer] was...effectively random". Similar suggestions have been made about the distribution of Malawi’s earlier targeted inputs programme, and – by Charles Knox in his recent comment – about the Productive Safety Nets Programme in Ethiopia, where "more than two-thirds of non-beneficiaries should have been on the programme".

So what is the answer? Is there any evidence of successful community-based poverty targeting from anywhere in the world – at national, rather than pilot, level – that can instil confidence in a poverty-targeted methodology? Or is it so totally discredited that the only option is a more universal or categorically-targeted approach? Or, finally, might there be a practical and pragmatic solution that lies somewhere in between the two poles – a "third way" for social protection in sub-Saharan Africa?

[RHVP Editorial Comment]: RHVP is currently developing a paper on targeting as one of the "hot topics" in our "Frontiers of Social Protection" research agenda. We would welcome evidence of successful (and unsuccessful) community-based poverty targeting from anywhere in the world – preferably at national, rather than pilot, level – that can inform our study.