A recent World Bank working paper asks “Should cash transfers be confined to the poor?”, and answers with a resounding “Yes!” The authors would vote for Manchester University economist Armando Barrientos in my debate with him at Deutsche Bank’s Global Financial Institute site. Armando argued against the proposition “Should pensions be paid out universally to all elderly?”, and I argued in favour.
This paper compares for 13 Latin American countries the poverty and inequality impacts of cash transfer programs that are given to all children and the elderly (that is, “categorical” transfers), to programs of equal budget that are confined to the poor within each population group (that is, “poverty targeted” transfers). …. Overall, the findings suggest that although in the Latin American context poverty targeting tends to deliver higher poverty impacts, there are circumstances under which categorical targeting confined to geographical regions (sometimes called “geographic targeting”) may be a valid option to consider.
Pablo Acosta, Phillipe Leite and Jamele Rigolini, “Should Cash Transfers Be Confined to the Poor?“, Policy Research Working Paper 5875, World Bank, November 2011.
The authors are economists employed by the World Bank. The paper carries a warning that “The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the … World Bank …” It is a product of the office of Mr Augusto de la Torre, Chief Economist for Latin America and the Caribbean. Mr de la Torre, Prior to joining the World Bank in 1997, headed Ecuador’s Central Bank.
The three economists praise targeted cash transfers using hypothetical assumptions that are simplistic and unrealistic. They compare poverty targeted and universal (“categorical”) transfers of a cash amount equal to 0.5% of GDP, first to children aged 0-5 years and then to adults older than 65 years. In their main simulation, they set administrative costs and targeting errors equal to zero. The implicit assumption that it is possible to identify the poor and the non-poor without cost and without error is unrealistic. Also unrealistic is the assumption of a fixed budget (0.5% of GDP), to be divided equally among recipients, which guarantees that the poor benefit more from targeted transfers. In practice, universal programmes almost always have much larger budgets than poverty targeted programmes, possibly because taxpayers resent paying for cash benefits from which they are excluded.
From simulations based on these assumptions, Acosta, Leite and Rigolini reach the following conclusion (p. 14):
In all countries under consideration, a targeted cash transfer can achieve the same poverty impact than a categorical one using from 2 to 13 times less resources. The large difference in costs suggest that in most countries some form of targeting, even if costlier and far from perfect, can lead to large efficiency gains and allow for more generous transfers to the poor.
It is depressing for me to see three highly-trained economists use the term “efficiency gains” in this way. What the researchers measure is not economic efficiency, but rather “poverty impacts” – cash transfers that reach the poor rather than the non-poor. A universal benefit is almost always more efficient than a targeted benefit in terms of true economic efficiency, in terms of incentives to work and to save.
Stephen Kidd was even more irritated than I was by the poor quality of this paper. Here is an excerpt from the conclusion to his review of the work.
The paper by Acosta, Leite and Rigolini adds no value to the debate on targeting. It uses a highly problematic methodology with no basis in reality. The conclusion that poverty targeting performs better than universal targeting cannot, in this case, be substantiated.
It is very simple to arrive at the conclusion that universal pensions and child grants will be more effective in tackling poverty than their poverty-targeted counterparts. Once the high exclusion errors associated with poverty targeting and the very low errors found in universal programmes are recognised, it is easy to understand that a universal programme will perform significantly better in reducing poverty, assuming that both programmes have similar benefits (an assumption that responds to reality). By reaching all – or almost all – of the poor, a universal benefit will necessarily be more effective than a programme that misses over half of the poor. It will, of course, be more expensive but, since it is likely to be more popular, it may well be more financially sustainable (as borne out by many universal pension programmes in developing countries).
The point of this critique is not to argue against undertaking simulations of potential social protection programmes, as they can be valuable policy tools. Rather, analysts should employ simulations sensibly and objectively, using realistic assumptions. In the case of the Acosta, Leite and Rigolini paper, at a minimum the size of benefit should have been consistent across both poverty-targeted and universal programmes and realistic targeting errors should have been used.
A final plea to the World Bank: please strengthen your peer review processes.
Stephen Kidd, “A Case Study of How Not to do Targeting Analysis“, Pathways’ Perspective #4, 27 July 2012.
Stephen Kidd is a Senior Social Policy Specialist at Development Pathways who formerly worked at HelpAge International. He’s becoming increasingly irritated by poor quality analytical work.