…how is cost-effectiveness achieved?
It is been recognized that the major problems concerning targeting for social intervention programs include the often ignored factor of under-coverage, the problem of leakage to the rich, and the high risk that targeted programs could end up being regressive. However, if the programs that have shown good results are critically examined, it can be seen that many of these programs require complex methods and advanced institutional capacity for the identification of the poor and the transfer of resources, which translates into big administrative budgets. As a consequence, although targeting is presented as a more effective method to combat poverty within a fixed budget, there are, in practice, limiting factors that minimize the poverty-reducing effect of a targeting program.
Although the cost of identifying the poor is not neglected in the theoretical literature on targeting, it is often not deemed relevant in evaluations of targeting programs. Data on the administrative costs of targeting are often incomplete and seldom possible to compare systematically because of a lack of key information. Experts raised this issue when international financial institutions started to promote targeting of social programs.
However, there is still no coherent or standardized method to measure the costs of targeting (including administrative, identification, and transfer costs), leakage, exclusion, and overall efficiency. The costs of targeting are sometimes only reported in terms of the cost of identification of the beneficiaries, although there is clear evidence that the continuous administration of a targeted system requires more administrative resources than a universal program. Literature has also discussed the difficulty associated with “the imprecision in calculating administrative costs”. Empirical works of some experts put forward the argument that even if only trying to count cost-effectiveness with a simple approach of cost per unit of the benefit received by the poor, it is still difficult to find acceptable studies on cost-effectiveness on targeting social projects.
The ongoing need to identify the poor, using different methods, and more often a combination of methods, is costly, time-consuming, and requires continuous institutional capacity. Accurate identification of the poor, minimizing the under-coverage among the poor and controlling leakage (while maintaining high rates of under-coverage), and implementing well-developed fraud control systems often ends up being a very expensive process.
An empirical work by Grosh calculated the average cost of administrating individual targeting schemes, which have the best success rates in terms of targeting, at 9 percent, varying between 0.4 percent and 29 percent, of total program costs (see also Gwatkin 2000). Other less efficient models (such as geographic targeting) tend to be cheaper to administrate at 6 percent to 7 percent of program costs according to Gwatkin and 7 percent according to Grosh, but they are also less effective due to higher levels of under-coverage and leakage. The median costs of self-targeting schemes are 6 percent of total program costs; and in a study of social funds, Rawlings et al. in their empirical work in 2004, reported that social funds expenses among countries vary between 7 percent and 13 percent of total program costs. Also, there is controversy regarding what should be defined as targeting costs. Grosh’s definition of targeting costs, as stated above, is quite narrow since it relates only to the initial screening cost. Once the poor have been identified, there are other costs involved in the process such as for the delivery of services to the eligible and exclusion of the non-eligible, and for fraud control, given that the cost of fraud in targeting systems may constitute a very high cost for the program. When calculating costs for targeting schemes in Latin America for instance, Coady et al. in their empirical exercise in 2004 considered that “in most cases, it appears that corruption and theft contribute more to total program expenses than legitimate administrative expenses”. If this were correct, it would mean that at least 20 percent of the budget of an average targeted program disappears in legal administrative costs and corrupt practices. Regardless of program outcomes, this raises several questions about cost-effectiveness. Undertaking simulations of safety net transfer in low-income countries, Smith and Subbarao (2003) calculated total administrative costs for targeting programs at 30 percent (compared with 15 percent for universal programs).
In many cases, targeted programs cost even more, as several studies have indicated. An Asian Development Bank Institute paper on the cost-effectiveness of various targeted antipoverty programs in India includes several studies that focused on the cost-effectiveness of alternative targeting programs and provides examples of several large targeting programs costing more than the actual benefit reaching the poor. Approximate estimates suggest that the cost of transferring a rupee to the poor in the Maharashtra Employment Guarantee Scheme, which was introduced in 1972, was 1.85 rupees per rupee transferred in targeting administration costs in its early years, while the later national employment scheme, Jawahar Rogzar Yojana, cost 2.28 rupees and the targeted food subsidy program cost 6.68 rupees. A separate evaluation of employment guarantee schemes in three Indian states by Weiss in 2004 found the cost per daily job to be between 200 and 300 rupees, while the benefit itself was just between 35 and 50 rupees. Despite being one of the most famous self-targeting programs due to its high administrative costs, it was widely concluded that the Maharashtra Employment Guarantee Scheme is less equitable and efficient in reducing poverty than similar universal schemes.
Thus, an important issue is the cost of effective targeting. The Temporary Assistance for Needy Families (TANF) scheme, which was referred to by Lindert in his empirical work in 2005 as the gold standard for targeting, is the most effective verified means-testing scheme in the United States, with extremely low levels of leakage but with quite high rates of under-coverage. The identification cost is $86 per screening process based on interviews, which in a Latin American context would be equivalent to $25 per application process. To this should be added the costs of administrating the program, including money transfers, fraud control, and so on. At the time of his empirical work, there are no numbers on total administrative costs of these programs in proportion to the total cost of the programs; however, an earlier study indicated screening administrative costs of 15.6 percent.
Another question, however, is whether there is a difference in administrative costs between targeted and universal programs. Although administrative costs often consume an important part of allocated funds, there is no clear and standardized way to report and evaluate costs. Earlier comparisons between universal and targeted programs in the United Kingdom indicated administrative costs of 3.5 percent for universal programs and between 5 percent and 15 percent for means-tested programs, while studies from the United States found 2.5 percent versus 13 percent for universal programs as compared to means-tested programs. Simulation studies that included “standardized” administrative costs calculated 30 percent administrative costs for targeting programs and 15 percent for universal programs. Although clear conclusions cannot be drawn on the exact differences in costs between means-tested and universal provision, it could be safely assumed that targeting always signifies a substantially higher cost.
Concerning effectiveness (or lack of effectiveness), there is a greater need for more complete reporting on the costs of targeting programs. The problem is that most studies and evaluations lack an overview of all costs involved. A coherent and standardized methodology in this field would facilitate comparative cross-country studies of targeted schemes (as well as universal schemes) and would clarify the differences in costs between different models. By only applying a cost-effectiveness analysis, as promoted by some development agencies for evaluating the impact of program interventions, there is no doubt that most of these programs would fail due to their high direct and indirect costs and high rates of exclusion. Given that such approaches cannot also evaluate the direct costs (for society and the poor) of those excluded from the interventions, the level of failure could be even higher.
The high rates of the poor that are excluded from these programs raise even more concerns. Similar figures about costs and effectiveness in a private company evaluation would certainly draw attention. Furthermore, focusing on poverty indices does not take into account the various non-economic costs of targeting, of which the issue of sustainability of the system is probably the most important. Including such costs would probably change the conclusions of many studies since non-economic costs could result in serious real economic losses for the poor.