6 minute read

Methods That Rank People According to Welfare

Choosing among Targeting Methods | 301

BOX 5.6 (continued)

literature in particular has a good deal in common with the proxy means testing literature about which data and methods of inference will be most suitable. In general, the profiling defines the treatment track (for example, whether or how much training support will be provided), but it does not get into granular detail, over which the individual and case worker maintain some agency (for example, the specific training program in which to enroll).

Useful references include Barnes et al. (2015); Desiere, Langenbucher, and Struyven (2019); and Loxha and Morgandi (2014).

Source: van Landeghem, Desiere, and Struyven 2021. a. The analogy is not perfect, as the most intensive active labor market policies may be offered to those at medium risk of long-term unemployment, with those at low risk expected to gain work with little support, and those at highest risk of unemployment referred to other social services to handle issues such as health and rehabilitation, substance abuse, and so forth, with active labor market policies available but not the first focus.

Methods That Rank People According to Welfare

The choice among household-specific targeting methods (when they are to be used) depends on several factors. Factors related to the availability of information are well known, but there are others as well—related to the institutional landscape and social contract.

Factors Related to Formality and Information Sources

The availability of administrative data is a traditional consideration in choosing among methods for household-specific assessments of moneymetric welfare.17 The following are particularly influential: (1) the share of formal employment and formal earnings; (2) the capacity of the administration to observe and verify the income (and assets) of the country’s population through databases designed for other purposes—income tax, social security contributions, land registries, car ownership, passport use, payment of utility bills to state-owned enterprises, and so forth; (3) the quality and frequency of such data; and (4) the ease with which data from different sources can be matched.

When the administration can easily verify the main sources of income and assets of the applicant population, the means testing method can be accurate and is often viewed as the gold standard as it is the only method that does not include any inherent errors. This method is typically found in

302 | Revisiting Targeting in Social Assistance

Organisation for Economic Co-operation and Development/high-income countries, with their highly formal economies and extensive data systems that allow for verification. There are also applications in middle-income countries, such as Brazil, China, and South Africa, which started with little verification of declared income. Over time, among other actions, such as improved interoperability, these countries have built robust monitoring and evaluation systems to identify inconsistencies and effective strategies to communicate to the population the consequences of false declaration, which would lead households to suspension of benefits and penalties that would block any member of the family from receiving any type of assistance. Means tests stand alone as the only method used in half the instances in which they are used. Means testing is also commonly used to restrict categorical programs, such as child allowances or social pensions for needier families; when means testing is used with another method, 83 percent of the time it includes categorical eligibility.

The definition of income used for eligibility purposes—sometimes referred to as the administrative or program definition of income—may differ slightly from the economic definition of income. Some types of income that add welfare may not be included in the administrative definition, such as transfer income from other social protection programs or portions of earnings that may be disregarded to avoid an income trap that would discourage work or run counter to the objective of the program of focus. Other types of income—typically occasional, rare events, or those that cannot be verified—may also be disregarded. However, the two measures, administrative and economic definitions of income, will be closely correlated in terms of ranking as well as levels.

When formality is insufficient for a verified means test, an HMT that relies on the measurement of formal incomes (observable and verifiable) and the estimation of some of the informal incomes to produce an estimate of total welfare may be appropriate. An informal employment rate of over 20 percent will reduce the precision of a means test. However, if formal employment is somewhere between 50 and 80 percent, the economy has a large share of its income in the formal sector. Hence, income is “visible” to the administration, tilting the balance toward methods of observing rather than estimating the welfare of applicant households. Asset registries—of land ownership, dwellings, cars, and so forth—can be used to impute incomes, to add to the verifiable sources or as filters for affluence testing.

When a large proportion of jobs are in the informal sector, it is difficult to observe, measure, and/or verify the level of income or consumption and assets of a household. Therefore, it is difficult to identify the population using observed welfare measures, which calls for an alternative approach.

Choosing among Targeting Methods | 303

This was the motivation when the PMT was invented in Chile 40 years ago; it is also why the method caught on and spread among the many other countries with significant informality.

The PMT method consists of estimating household welfare based on a set of relatively easy-to-observe indicators such as individual demographic characteristics, aspects of the dwelling, ownership of household durable goods, or location. The parameters of PMT programs are determined using a representative household survey that collects information on household welfare and the characteristics correlated with it and statistical modeling. The precision of the PMT depends on the quality of the information, the procedure used to estimate the coefficients, and the strength of the association. However, because it uses inference rather than measurement, PMT involves inherent error.

PMT is generally combined with another targeting method. In over three-quarters of the ASPIRE observations where PMT is used, it is part of a mixed methods approach. Categorical (69 percent) and geographic (48 percent) methods are the most common partners, but PMT commonly appears with all the other methods, including community-based ones (31 percent) and means testing (23 percent).18

In any of means testing, HMT, or PMT, the calculated welfare measure is compared with an eligibility threshold, and different thresholds may be used for different programs. The threshold(s) can be estimated based on administrative data, for example, if the country has a comprehensive income tax, or from representative income surveys to get an idea of how many people would qualify at each level. For PMT, the use of multiple thresholds presents additional complexity in statistical modeling of the relationship between proxies and true welfare, a theme taken up in chapters 6 and 8.

As a country becomes more formalized and improves the interoperability of its databases, it may progress from PMT to HMT to means testing, as each later stage requires greater capacity but is more accurate. The case study of Chile in box 5.7 is a classic example of this progression. It was the first country to deploy a rudimentary PMT in 1979. As program objectives, data availability and technical capacity, evolved, it moved to a version of HMT and now uses means testing. Several countries that use PMT, such as Albania, Armenia, Costa Rica, Georgia, and Turkey, include formal sector information (on formal wages, social protection program benefit, or receipt) in their PMT formula or as exclusionary filters. As part of its regular review of the performance of its scoring systems, Armenia has examined moving from PMT to HMT, but no decision has been taken at the time of writing. Saudi Arabia is considering transitioning to an HMT.

This article is from: