33372 World Bank Social Safety Nets Primer Notes 2005 No. 22 Designing and Implementing HouseholdTargeting Systems: Lessons from Latin America and the United States Motivation and Objectives Comparing household characteristics with A recent review of international experience (Co- pre-established eligibility criteria; and ady, Grosh and Hoddinott (2004)) finds that the mediantargetedprogramprovidesroughly25% Establishing program-specific beneficiary more resources to the poor than would random lists. allocations. Nonetheless, the study also finds significant variation in targeting outcomes both Six Country Case Studies withinandacrosstypesoftargetinginstruments. Householdtargetingsystemsinsixcountriesare Implementation details matter tremendously to examined: theVerified MeansTesting (VMT) sys- distributive outcomes. Building on these find- tems in the United States, the Unverified Means ings, this report conducts an in-depth assess- Test (UMT) system in Brazil, and Proxy Means ment of key design and implementation factors Testing (PMT) systems in Chile, Colombia, Costa and their potential impact on outcomes for Rica and Mexico. household targeting systems in six countries. Overview of Main Findings What areTargeting Systems? A number of key messages emerge from the There are a variety of targeting instruments, cross-country analysis: including household (or individual) assessment mechanisms (means testing, proxy means test- Data collection processes should be de- ing),broadcategoricaleligibility(e.g.,geograph- signed so as to ensure transparency, dynamism ic targeting), and self-targeting. Many programs (open entry into registries), outreach to the adopt a combination of these mechanisms and (potentially) poor, cost efficiency, and adminis- while the design and implementation varies trative feasibility. The strategy for conducting significantly, most household targeting systems interviews and collecting data is as important involve the following steps: to the success of household targeting systems as the type of eligibility mechanism used. Two Collecting data on specific (potentially eli- common strategies used for data collection are gible) households via interviews and/or visits described in Box 1. Other factors to be consid- using pre-designed questionnaires; ered in designing data collection strategies are: (a) the location of interviews (home vs. office Entering these data into a unified household visits,orboth);(b)thequalityofinterviews;and information registry; (c) communications. ____________________________________________________________________________________________________ Kathy Lindert prepared this note based on Castañeda,Tarsicio; Lindert, Kathy; de la Brière, Bénédicte; Fernandez, Luisa; Hubert, Celia; Larrañaga, Osvaldo; Orozco, Monica andViquez, Roxana. 2005."Designing and Implementing Household Targeting Systems: Lessons from Latin America and the United States."Social Protection Discussion Paper No. 0526.World Bank.Washington, D.C. Box 1 ­TwoTypes of Data Collection Processes for HouseholdTargeting Systems ·The quasi-exhaustive survey approach involves collecting household-level data on all households within a pre-speci- fied geographic area. This approach is cheaper (per interview) to implement and favors outreach to the poor. How- ever, the survey approach is generally static (allowing for only infrequent registration and updates) and is associated with somewhat weaker targeting accuracy in the sample cases. ·The on-demand applications approach relies on households to apply for benefits. It favors dynamic, on-going regis- tration as well as regular updating and re-certification (due to the extensive network of welfare offices usually present with this approach). It is associated with stronger targeting accuracy (lower leakage) among the cases in our sample. However, the on-demand approach can also miss the poor (lower coverage), who may be less informed or connected. Depending on the poverty density of particular areas, a mix of these two data collection approaches can be an effec- tive way to balance the goals of maximizing outreach to the poor while minimizing the costs of interviewing large numbers of likely ineligible non-poor households. Micro-area poverty maps can help guide these design choices by providing localized information on poverty prevalence and density. Information management factors affect the ·Verified means testing (VMT) produces"gold quality of household targeting systems. First, standard" results for accuracy and can also a consolidated national database is important promotetransparencyandcredibility. Howev- and can help avoid duplications and track er,VMT can be extremely costly to implement, beneficiaries, even if data are collected locally. and administratively and technically infeasible Second, proper identification of individuals is in developing countries with high degrees of crucial. A unique social identification number informality in the labor markets. should be used and software and coding sys- tems need to be designed to link individuals · Unverified means testing (UMT) can be a with particular families (or assistance units). less expensive and more feasible alternative, Third, updates and re-certifications are neces- particularly in situations in which quick deci- sary for tracking fraud, facilitating turnover sions are required. Targeting accuracy can so that other poor families may enter, and be reasonable (especially if combined with avoiding situations such as "ghost" beneficia- geographic targeting), though the outcomes ries. Fourth, database management should in the Brazil case were not as strong as those require pretesting of systems, well-designed for VMT or PMT. However, concerns about manuals, adequate user training and ensure transparency, measurement error and adverse the flexibility to respond to changing policies incentives for under-reporting make UMT less and updates. attractive from a technical and political point of view when eligibility for large or long-term Which assessment mechanism is best? The benefits is being determined. Proxy indicators choice depends on a number of factors in- (such as those used under PMT) can be used cluding (a) cost and administrative feasibility; as"consistencychecks"to"verify"self-reported (b) technical feasibility, given the degree of incomes, and improve accuracy and transpar- informality in the economy; and (c) political ency. acceptability. Household assessment mecha- nisms should seek to maximize targeting accu- · Proxy means testing (PMT) is a promis- racy at an acceptable cost and in a transparent ing alternative for targeting cash transfers in manner. Combining household assessment developing countries with high degrees of mechanisms with other instruments (such as informality in the labor market. In the cases geographic targeting) can improve redistribu- examined in this study, PMT performed well tive outcomes and poverty impact. intermsoftargetingoutcomes,costefficiency parent and accurate than UMT. PMT systems greater share of resources spent on programs in LAC have generated targeting incidence that use the household targeting systems outcomes that approximate the impressive reach the poor; (c) cost efficiency, by making record of VMT for a mere fraction of the cost efforts to minimize the cost of interviewing of interviewing and screening for eligibility. families while ensuring the integrity of intake Between 80-90% of the benefits of proxy- efforts; and (d) transparency in all aspects to means tested programs in Chile and Mexico enhance credibility and reduce fraud. The ta- are received by the poorest 40% (two quin- ble below presents a summary covering these tiles) of households. Moreover, the costs of principles for the six case studies. The systems these systems are relatively low (ranging from of Mexico and Chile perform impressively well US$2.3-8.4 per interview, or 9-34% of com- in terms of targeting outcomes, cost efficiency parable interview costs for VMT in the U.S.), and transparency. The registries in the United and administrative requirements are more States perform extremely well in terms of manageable, particularly for middle-income maximizing targeting accuracy (low leakages) developing countries. andtransparency,but thesystemisextremely costly and fails to cover about half of the poor Institutional roles should be clearly defined (in part due to the complexity of the system). and communicated. Institutional arrange- Both Colombia and Brazil are currently un- ments should be tailored to local realities. dertaking to implement significant reforms While there is no single blueprint for insti- to strengthen their registries, which should tutional roles, there are advantages in a sys- improve their performance over time. tem with centralized design and database management. In any event, arrangements Caveats and Limitations should be made to promote quality at all lev- The sample of household registries (and user els, such as federal cost-sharing and financial programs) in the six countries is too small to incentives for municipalities if they are to be establish causality between particular design/ charged with implementing data collection. implementation features and the performance Clear federal guidelines for processes should outcomes of these systems. In fact, there is becommunicated,andinstrumentsforfederal no single recommended "blueprint" recipe for oversight are needed. household targeting systems because a large number of factors are involved in design and Strong mechanisms for monitoring and implementation and these must be adapted oversight are crucial for all systems, but es- to suit each country's particular "local realities" pecially with decentralized data collection. of socio-economic and political circumstances, While no system is 100% immune to fraud or institutional arrangements, and administrative leakages, multiple mechanisms can be used capacities. The study uses the cross-country to minimize them, including: supervision of analysis to identify some common factors which interviews, verification of information, auto- require careful consideration in order to achieve matedchecks,comparingregistrieswithother successful targeting outcomes. data, random-sample quality control reviews, and citizen oversight ("social controls"). There is significant variation in the suc- cess of the six household targeting systems reviewed in this study. Four principles for judging the success of household targeting systems are used, including: (a) maximizing coverage of the poor (or alternatively, mini- mizing errors of exclusion); (b) minimizing leakages to the non-poor, to ensure that a Summary Report Card: Six HouseholdTargeting Systems ­ Principles for Judging Success Maximizing Minimizing Cost Efficiency Transparency Coverage of Poor Leakage to Non- (per interview) (see main report) Poor U.S. Registries (VMT, Fair Very strong Expensive High On-Demand Data Col- About half covered CGH: 3.3-4.0 US$25 (86) lection) Complexity is barrier Chile Ficha CAS (PMT, Low Very strong Reasonable Fairly High On-Demand Data Col- 16-27% covered CGH: 2.7-3.3 US$8.4 lection) Fiscal constraints (on-demand more costly) Mexico Registry for Fairly high Very strong Reasonable Fairly High Oportunidades (PMT, 60% covered CGH: 2.9 US$4.9-6.8 Mixed Data Collection Could cover all poor, but Processes) leakages detract Costa Rica SIPO (PMT, Data not available Data not available Reasonable Medium-High Mixed Data Collection US$4.2-7.0 Processes) Colombia SISBEN Data not available Fair Reasonable Low but improving (PMT, mostly survey CGH: 1.7 US$1.8-2.9 sweep data collec- tion) Brazil Cadastro Unico Low but expand- Fair Reasonable Low but improving (UMT, survey sweep ing CGH: 1.98 US$3.9 data collection) 27% of poor cov- ered by Bolsa Escola (2002-03); Some mis- targeting, fiscal con- straints Note: CGHreferstotheCoady-Grosh-Hoddinottindicator,whichisconstructedbasedonacomparisonofactualtargetingperformance to the outcome that would result from neutral (as opposed to progressive or regressive) targeting. A neutral targeting outcome means that each quintile receives 20% of the transfer budget or accounts for 20% of program beneficiaries. Source: Compiled by authors using information presented in this report. TheWorld Bank Social Safety Nets Primer series is intended to provide a practical resource for those engaged in the design and implementation of safety net programs around the world. Readers will find information on good practices for a variety of types of interventions, country contexts, themes and target groups, as well as current thinking on the role of social safety nets in the broader development agenda. World Bank, Human Development Network Social Protection, Social Safety Nets http://www.worldbank.org/safetynets