
6 minute read
Key Elements for Means Tests
How to Harness the Power of Data and Inference | 355
Key Elements for Means Tests
Means-tested programs are typically found in high-income countries with highly formal economies, such as the traditional Organisation for Economic Co-operation and Development (OECD) members. Some countries, notably Australia, Canada, Japan, New Zealand, the United Kingdom, and the United States, employ means tests for most of their social assistance programs. European Union (EU) countries use means-tested programs as a last resort, intended to catch individuals or families whose income is still low after the support provided by the other social protection programs, including generous categorical social assistance programs. Among the income support social assistance programs of 35 European countries, Coady et al. (2021) find that means-tested programs represent about one-third of the programs, or about 1.1 percent of gross domestic product.10 Within the European Union, southern countries like Greece did not have a meanstested program a decade ago, given the high level of informality. A minimum income program called the Social Solidarity Income, meant to support the poorest households, was rolled out nationally in February 2017 in Greece, bringing the country in line with other EU and most OECD countries.11
The landscape of means-tested programs is varied. For example, in the United States, means-tested programs are used extensively to support lowincome individuals and families. At the beginning of 2020, before the COVID-19 crisis, they included 79 federal programs that covered about 19 percent of the population.12 These programs include the Medicaid program, which provides free medical care to low-income adults and children, the elderly and disabled, and long-term care; the Earned Income Tax Credit, which provides a tax credit to families and individuals with relatively low levels of earnings; the Supplemental Security Income program, which provides cash benefits to low-income aged, blind, and disabled individuals; housing subsidy programs, which provide housing vouchers to low-income families, subsidized rent in public housing projects, and support for construction of low-income housing; the Supplemental Nutrition Assistance Program (SNAP), formerly called food stamps, which provides an allotment of funds for food expenditure for low-income families and individuals; the Temporary Assistance to Needy Families (TANF) program, which provides cash assistance for general consumption to low-income families (mostly single mothers and children); school food programs (subsidized breakfasts and lunches for children from low-income families); the Head Start program (providing early education and childcare for children of low-income families); and the Women, Infants, and Children (WIC) program (providing nutritional assistance to mothers, infants, and children at nutritional risk).13 The extreme fragmentation of the means-tested programs in the
356 | Revisiting Targeting in Social Assistance
United States reflects the historical development of safety net as well as voter preferences for in-kind programs and concerns about supporting working-age people who might not be working (Moffitt 2015).14 Other countries, such Australia, have centralized the administration of the programs, and the United Kingdom has consolidated the means-tested programs for working families under a single program, the Universal Credit. Most countries in the European Union have one or few last-resort meanstested program(s); they may use means testing selectively for social pensions, disability support, or labor market programs.15
Means-tested programs can be designed for different assistance units: households, families, or individuals. If the assistance unit is the household (or family), total household (family) income is assessed based on the total current income of all members, as well as the incomes that are not attributable to specific members (for example, incomes from agriculture, rental income from jointly owned assets, and social protection family benefits). If the assistance unit is the individual, for example for a means-tested social pension or disability benefit, only the incomes of that individual are considered.
Verified means testing is often considered the gold standard of targeting methods because unlike the other methods, it directly measures the desired welfare concept: the income and sometimes some of the assets of the applicant. Welfare can be measured as the sum of the incomes and assets of the members of the assistance unit. These can be observed and verified by the administration, which has records of such incomes and assets. Assuming full take-up and no underreporting or other types of errors in the administrative databases,16 a verified means test will have no inclusion or exclusion error. In contrast, HMT, PMT, and geographical targeting cannot observe or verify the full welfare aggregate of the assistance unit—they can only estimate it with some known modeling error. With means testing, there is the potential that by pulling data largely from other existing systems, good targeting accuracy can be achieved.
In practice, means-testing methods rely on measuring a pragmatic subset of incomes and assets rather than trying to quantify every possible source. Thus, the program’s definition of income and assets differs from its comprehensive economic definition. Typically, means-tested programs account for formal wages from the main job and sometimes secondary jobs; retirement and other pensions received from pension institutions; regular monthly income from other social protection programs, such as labor market or social assistance programs; and unearned income, such as rents, dividends, court-ordered alimony or child support, and so forth. All these income sources are regularly recorded in administrative or private databases, which are often used to verify the income declared by applicants.
How to Harness the Power of Data and Inference | 357
Certain incomes, such as small, irregular, or informal/unrecorded incomes, are typically excluded from the program definition of income. Whether to try to include informal incomes in the program’s definition of household income, which represent a small share of total income in the countries where verified means testing is typically used, is more debatable. For example, a low-wage worker may have easily reported and verified wages but pick up money off the books from occasional side gigs like babysitting for a neighbor. Not counting such income does undercount welfare. To count it is very difficult, certainly raising transaction costs to participants and administrative costs. Moreover, if such informal income is small, trying to count it may not improve targeting accuracy by much; instead, it could push the low-wage worker into the category of fraudster or discourage work effort, neither of which is desirable. Such sources of income that represent a small proportion of the average total income are irregular or not typically reported; in practice, they are disregarded. Another source that is disregarded, given the difficulty of estimating and verifying it, is consumption from a household’s own production. However, if such informal income is larger, more structural changes in method may be needed, such as moving to an HMT or a PMT.
The program administration‘s ability to observe (the largest proportion of) applicant income and assets implies that it can also observe the change in the level of that income or assets; thus, it can protect those who are vulnerable after a shock. One of the advantages of means testing is that it should be able to detect changes in income and allow households to apply for benefits any time their income falls and receive assistance fairly quickly thereafter. For this to happen, the income sources should be paid and reported at frequent intervals, such as once or twice a month. Thus, a welloff person who becomes unemployed will qualify for assistance from a means-tested program in the next period. This is also straightforward for assets, which are a stock concept (box 6.2). A shock that reduces the stock of a particular asset could be quickly considered, directly (if the program also uses an eligibility line for assets) or indirectly (if the reduction in the stock of a productive asset reduces its income-generating capacity). When used in this way, programs can prevent adjustments from which it is difficult for families to recover—for example, a family losing its housing or selling the car or motorcycle it needs to get the breadwinner to work to make rent—or an outcome with long-term consequences—especially a child raised in hardship long enough to affect the child’s growth or development.
Apart from the list of income sources that are included in the definition of income used by the program, the recall period during which incomes should be reported and counted should also be specified. Shorter recall periods, such as a month or a quarter, will allow households that fall on