Notes

188 | Revisiting Targeting in Social Assistance

registration, assessment of needs and conditions, eligibility and enrollment decisions, determination of benefits and services, notification and onboarding, and provision of benefits and services. In particular, it highlights that “targeting failures” can happen at any step of the chain and discusses the roles of IDs, social registries, payment mechanisms, and adaptive social protection.

Chapter 5 relates the concerns raised in this chapter to the choice of targeting method. It first observes the evolution of targeting practice over the past two decades. It then outlines a framework for choosing method(s), accounting for fit-for-purpose with respect to program objectives and feasibility in different country contexts. Each method is assessed in the light of this framework, and quantitative simulations are introduced to help select methods.

Chapter 6 looks at notions of data and inferences and how they connect to some of the options among the methods. The chapter reviews the nature of data for inferring eligibility. The rest of the chapter provides an in-depth, technical overview of each of the methods considered—geographic, means testing, hybrid means testing, proxy means testing, and communitybased testing—considering best practices, how big data and new inference methods influence them, and how they can be adapted for shocks.

Notes

1. For example, the United Nations Household Survey Capability Program defines a household as a group of people who live together, pool their money, and eat at least one meal a day together (United Nations 1989). Eurostat defines a household in the context of surveys on social conditions or income, such as the European Union Statistics on Income and Living Conditions or Household

Budget Survey, as a housekeeping unit or, operationally, as a social unit: having common arrangements, sharing household expenses or daily needs, or in a shared common residence (https://ec.europa.eu/eurostat/statistics-explained /index.php/Glossary:Household_-_social_statistics#:~:text=A%20 household%2C%20in%20the%20context,in%20a%20shared%20 common%20residence). 2. An adult equivalence scale is an adjustment made in calculating household welfare that accounts for its demographic composition, with the underlying hypothesis that people of different ages have different needs. For example, if food is the largest item in consumption and children need fewer calories than adults, they would have a weight less than one, in proportion to established scales of caloric need by age and sex. Economies of scale attempt to capture the idea that two people together can live more cheaply than two people separately. For example, it takes less than twice the fuel to heat a larger cooking pot, housing will be less than twice as large, and consumer durables such as a television or refrigerator can be shared.

Moving from General Abstraction toward Implementable Concepts | 189

3. The use of a different unit for analysis than for policy can confuse evaluation.

In evaluations of targeting outcomes, facile use of household survey data could show errors of inclusion or exclusion when no eligibility decision process was wrongly conducted. Decisions about the appropriate assistance unit bear deep policy consideration and evaluative data analysis that matches. In some countries, care is taken so that the definitions of assistance units and households can be mapped to one another, with the social assistance program working within the nomenclature of the national statistical office. 4. Marriage rates have been declining in many countries, especially in the West, with the share of never-married women ages 40–44 rising, for example, in

Australia (from 5 percent in 1986 to 15.6 percent in 2006), Brazil (from 9 percent in 1980 to 33.8 percent in 2010), France (from 7.5 percent in 1985 to 27.9 percent in 2009), and the United States (from 4.8 percent in 1982 to 13.8 percent in 2012). In 2010, more than 40 percent of births in France,

Norway, and Sweden were to women in cohabiting relationships, compared with 25 percent in the United States (Besharov and Gilbert 2015). The share of women ages 20–24 in informal unions in Latin America rose between the 1970 and 2000 censuses to rates between 23 (Mexico) and 69 (Peru) percent (Esteve,

Lesthaeghe, and Lopez-Gay 2012). In Jamaica, according to 2014 data from the

Registrar General’s Department, 85 percent of children are born outside formal marriage (Gleaner 2016). 5. See, for example, Lambert, van de Walle, and Villar (2017); Premand, Schnitzer, and van de Walle (forthcoming); Trócaire (2017); and van de Walle (2013). 6. In Mali and Niger, eligibility for the cash transfer is based on the welfare of the compound household of husband, all wives, children, and others. The behavioral change elements that accompany the cash transfers that focus on protecting and boosting children’s human capital outcomes while empowering women are individually targeted to the mothers. In both countries, men are invited to participate in the behavioral change elements as well. In Ethiopia, in contrast, the rural Productive Safety Net Program treats each wife and her respective children as a separate household, with the husband ascribed as a member of the first wife’s household. 7. Here we focus on informal fostering where one or both parents remain prominent and trusted deciders of a child’s living arrangement as opposed to legal fostering where the state has terminated parental rights due to abuses or neglect of duty. 8. Demographic and Health survey data from 16 African countries show that the percentage of households with a foster child ranges from 15 percent in

Ghana to 37 percent in Namibia (Vandermeersch 1997). Lloyd and Desai (1992) use the same survey data to calculate the percentage of children living away from their biological parents and find rates ranging from 5 percent in

Burundi to 28 percent in Botswana. 9. The share of 18- to 29-year-olds living with their parents since the COVID-19 crisis began has reached 52 percent, surpassing the previous peak of 48 percent during the Great Depression. (https://www.pewresearch.org/fact-tank /2020/09/04/a-majority-of-young-adults-in-the-u-s-live-with-their-parents -for-the-first-time-since-the-great-depression/).

190 | Revisiting Targeting in Social Assistance

10. The threshold for block pricing can be difficult to set in ways that distinguish well the poorer from the less poor. For example, in Jordan, 85 percent of the poorest decile of households have total electricity consumption within the subsidized levels, but so do 53 percent of the richest decile (Rodriguez and

Wai-Poi 2020). 11. HelpAge, “A Resource for Promoting Dialogue on Creating a New UN

Convention on the Rights of Older Persons,” https://social.un.org/ageing -working-group/documents/Coalition%20to%20Strengthen%20the%20

Rights%20of%20Older%20People.pdf. 12. Kakwani and Subbarao (2005), writing when the HIV/AIDS epidemic was causing a surge of concern over the unprecedented strain on families, investigated poverty rates by various family structures. They found that while some of the expected generalizations about elderly poverty were true for regional aggregations, the magnitudes were markedly different. The poverty headcount in “skipgeneration” households comprised of only elderly and children was about twice as high as all poverty in Côte d’Ivoire, but only half as high in Cameroon. The

“elderly living alone” had a headcount poverty rate 8 percentage points higher than average in Uganda but 29 points lower in Nigeria. Households with both elderly and prime-age adults had slightly lower poverty headcounts in Burundi,

Madagascar, and Uganda than all households, but in Côte d’Ivoire, The Gambia,

Kenya, and Zambia, headcount poverty rates for households with elderly members were 10 or more percentage points higher than for all households. 13. Silwal et al. (2020) update the estimates with new surveys and similar methods, but the summary note contains less detail. The headline numbers for 2017 were that 17.5 percent of children in the world (or 356 million) younger than 18 years lived on less than $1.90 in purchasing power parity per day, compared with 7.9 percent of adults ages 18 and older. The poverty rates of children at the $3.20 and $5.50 lines were 41.5 and 66.7 percent, respectively. 14. See http://hdr.undp.org/en/2020-MPI on the Global MPI and World Bank (2018a) on the World Bank measure as well as an extended discussion on multidimensional measures in general. 15. See The Living Standards Measurement Study (LSMS) website (http://surveys .worldbank.org/lsms/), which presents the household survey program by the

World Bank’s Development Data Group. Since its inception in the early 1980s, the program offers technical assistance to national statistical offices in the design and implementation of multitopic household surveys and the measurement and monitoring of poverty. Other seminal publications include Beegle et al. (2012); Deaton and Grosh (2000); Smith, Dupriez, and Troubat (2014);

World Bank (2003). 16. The use of social welfare functions with continuous weights means it matters a lot less where the line is drawn in terms of measuring a program’s impact on welfare. Chapter 2 shows that use of the poverty gap means that poor households who receive benefits that are inadequate to bring them over the line are still counted as being positively affected. More generally, the distributional characteristic discussed in chapter 7 can be used to assess program performance where any household receiving a benefit is considered positively affected but with much higher weights on poorer households.

Moving from General Abstraction toward Implementable Concepts | 191

17. To facilitate comparability across countries, the World Bank has constructed a suite of international poverty lines that take into account the different costs of living in different countries ($1.90 per person per day for extreme poverty, $3.20 for lower-middle-income countries, and $5.50 for upper-middleincome countries), as well as multidimensional definitions of poverty that go beyond just the monetary dimension. Care in updating, interpretation, and looking at multifaceted pictures of poverty help to overcome the fundamental problem of definitions in poverty analytics. Moreover, as countries become wealthier and almost no households are poor when evaluated against any reasonable absolute poverty line, countries tend to move toward using a relative poverty line. For example, the European Union uses 60 percent of median income as its poverty line, which means some households will always be considered poor, regardless of how far above the absolute poverty lines their incomes are. 18. Errors of exclusion related to delivery systems (lack of outreach, difficulty in mustering required documentation, high transactions costs, and so forth) may be worst among the very poorest. 19. See López-Calva and Ortiz-Juarez (2011) for the original methodology.

Household panel survey data are used to observe income/consumption in one period and then a subsequent period. Lowess or logit regressions are then used to determine the probability of a household being in poverty in the second period given their first period income/consumption. From this probability curve, an economic security line can be set, for example, the point at which a household has less than 10 percent chance of being poor next year. See Ferreira et al. (2013) for an application to Latin America and the Caribbean, Ruggeri

Laderchi et al. (2017) for an application to East Asia and the Pacific, and World

Bank (2019) for an application to Indonesia. 20. See, for example, John Keyantash and National Center for Atmospheric

Research Staff (eds), “The Climate Data Guide: Standardized Precipitation

Index (SPI),” https://climatedataguide.ucar.edu/climate-data/standardized -precipitation-index-spi, last modified August 7, 2018. 21. The measure defines 0 meter = no risk; 0–0.15 meter = low risk; 0.15–0.5 meter = moderate risk; 0.5–1.5 meters = high risk; over 1.5 meters = very high risk (World Bank 2020f). 22. Earthquake magnitudes are commonly measured by the well-known

Richter scale. Under this scale, the magnitudes, effects, and frequency are (1) 2.5 or less, usually not felt, but can be recorded by seismograph, 900,000 per year; (2) 2.5 to 5.4, often felt, but only causes minor damage, 30,000 per year; (3) 5.5 to 6.0, slight damage to buildings and other structures, 500 per year; (4) 6.1 to 6.9, may cause a lot of damage in very populated areas, 100 per year; (5) 7.0 to 7.9, major earthquake with serious damage, 20 per year; (6) 8.0 or greater, great earthquake that can totally destroy communities near the epicenter, one every 5 to 10 years. 23. This should be done with country-specific empirical analysis. Hypotheses to test would be whether the chronic poor are among the poorest (because they have to increase their income the most to rise above the eligibility threshold) or those with barriers to increased earnings, such as low skills, poor health or

192 | Revisiting Targeting in Social Assistance

disability, residence in poor or remote areas, or membership in a historically excluded group by tribe, ethnicity, caste, religion, and so forth. 24. To counter concerns about disincentives around exit, there are various options.

For example, to encourage beneficiaries to take new jobs or formal sector jobs, workers who do so may be able to retain benefits for a period of time, there may be some sort of income disregard that allows workers to keep part of incremental income, or there could be a minimum benefit period or a right to return without wait-listing should the workers’ income again fall. To soften the transition and reduce political backlash, families found upon recertification to be above the exit threshold may receive benefits for a fixed transitional period, possibly at a lower level. 25. In the Philippines, legislation (the “4Ps Act” signed in April 2019) now mandates a regular revalidation of beneficiary targeting every three years.

Enumeration for a new Listahanan 3 started in October 2019, was disrupted and delayed by COVID-19, and as of December 2021, data encoding and validation were at the final stage. Once completed, the database should be used for targeting and recertification of beneficiaries. 26. In 2020, Pakistan adopted a recertification strategy, through approval by the

Benazir Income Support Programme (BISP) Board, for its flagship Ehsaas

Kafaalat program. BISP relied on analysis of various sources of data, including nationally representative and specialized surveys and cross-checks between the old and new National Socio-Economic Registry (NSER) data, to inform the development of the strategy. The strategy covers both NSER and programspecific elements, including (1) introducing socioeconomic status bands for the

NSER rather than a single cutoff to allow a broader range of programs to utilize the NSER as a social registry, (2) provisions for two quarters of transition benefits to all exiting cash transfer beneficiaries, and (3) continuity of benefits to exiting families with children enrolled in the education-linked conditional cash transfer program until primary school completion. The government is working on a plan that will allow multiple points of entry (physical registration desks at the subdistrict level, door-to-door surveys, and virtual services) (World Bank 2021a). 27. DNP CONPES 3877; World Bank 2021b. 28. Although the social registry was updated in 2015 (Listahanan 2), a substantial share of 4Ps households was missed, so the Department of Social Welfare and Development opted not to utilize the registry to recertify or exit families in 4Ps. 29. The government is working on a plan that will provide multiple points of entry (physical registration desks at the subdistrict level, door-to-door surveys, and virtual services) to allow data updates and program entry and exit on a more frequent basis. 30. Between periodic survey sweeps, to keep information up to date, households may request a survey for the first time or request that their household information be updated. These requests are made through the municipal SISBEN offices, which are run and funded by the municipalities themselves (World

Bank 2021b). 31. https://colaboracion.dnp.gov.co/CDT/Conpes/Econ%C3%B3micos/3877.pdf.

Moving from General Abstraction toward Implementable Concepts | 193

32. Calculations based on ASPIRE data. 33. https://www.chileatiende.gob.cl/fichas/33112-subsidio-familiar-suf. 34. Calculations based on ASPIRE data. 35. The two benefits use the same targeting and eligibility system but differ on the target groups (families with or without children) and benefit levels (MoLSA,

World Bank, and UNICEF 2021). 36. The 6 percent coverage refers to cash transfer beneficiaries, while 40 percent coverage refers to the social registry information system’s information on poor/ vulnerable households in general (nonbeneficiaries and beneficiaries of any social assistance in general). Some of the functions of Turkey’s social registry, such as case management, graduation to work, and support for informal working poor, are still nascent. Robustly developing more frequent, comprehensive case management systems and options for informal vulnerable households in

Turkey that integrate targeting and benefits between noncontributory social assistance and contributory social security systems, such as by subsidizing a unified health insurance system for all, will be especially needed for the green transition and improving the adaptability of social protection to shocks such as pandemics or earthquakes. 37. Homi Kharas, blog (2020), https://www.brookings.edu/blog/future -development/2020/10/21/the-impact-of-covid-19-on-global-extreme -poverty. 38. https://www.worldbank.org/en/topic/poverty/brief/projected-poverty -impacts-of-COVID-19. 39. The global extreme poverty rate is expected to increase for the first time since 1998, bringing it back to the level in 2017. The World Bank predicts that poverty will reach about 9.1–9.4 percent of population instead of declining to about 7.9 percent, as pre-COVID-19 estimates indicated (https://www.worldbank.org /en/topic/poverty/brief/projected-poverty-impacts-of-COVID-19). 40. There are many ways to build resilience to hazards ex ante, and resilience building has been a growing line of programming in social protection in recent years. Resilience building tends to be a “no regrets” policy as it often includes the same sorts of activities that would be important in reducing chronic poverty and building shared prosperity. At the community level, and within social protection, public works programs may help with landscape management and water retention in drought-prone areas, upgrade flood defenses in flood-prone areas, or provide landslide protection on densely settled, steep slopes. These policies can then benefit poor people who receive wages for temporary employment on the schemes and, along with their larger communities, may benefit from the protections provided in the works done. At the household level and within social protection, resilience building usually focuses on poor and vulnerable households that have slim margins between their baseline state and destitution. Activities that raise their incomes or diversify them in ways that lower risk and mechanisms that allow them to build assets or savings or join insurance pools all raise resilience. While it is very important for adaptive social protection, the topic of resilience building per se is broader than the focus of this book, which is targeting and eligibility determination.