Abstract
Background papers to the "Strategy for IMF Engagement on Social Spending"
Background Paper IV. The Debate on Universal and Targeted Transfers1
This paper sets out the issues that need to be considered when providing policy advice on the targeting of transfers, and the trade-offs involved in different approaches to targeting. Broader population coverage may be desirable due to administrative constraints or social and political preferences (e.g., to build public support for social programs and structural reforms). Greater reliance on methods that result in large population coverage and fiscal cost needs to be accompanied by progressive and efficient taxation to reduce the economic cost of redistribution. Since transfers need to be financed, it is important to consider both the tax and transfer sides when designing redistributive fiscal policy to ensure that taxes do not significantly offset the redistributive impact of transfers. Achieving distributional objectives requires that the share of lower-income groups in transfers is sufficiently higher than their share in taxes. The issue is therefore how to effectively channel resources to lower-income groups given administrative, social, and political constraints.
A. Introduction
1. There is a growing debate on the relative merits of universal and targeted social assistance transfers, especially in low-income contexts.2 This was flagged both in the IEO Evaluation Report on “The IMF and Social Protection” and during the consultation process. At the heart of the debate is the argument that targeting of benefits using means tests is very imperfect, resulting in large undercoverage of the targeted poor population, and can result in strong work disincentives when benefits are withdrawn rapidly as income increases. This is particularly so in LICs with large “informal” sectors (often characterized by self-employment and multiple and volatile sources of income) and limited administrative capacity, which makes verification of income very difficult.
2. The distinction between targeted and universal benefits relates to the use of eligibility conditions for receiving benefits. A universal benefit is defined as a benefit that is available to everyone without any eligibility conditions. For example, a Universal Basic Income (UBI) is typically defined as a uniform cash transfer that every person is entitled to regardless of income or other conditions (e.g., age, gender, or location) (IMF 2017; Francese and Prady, 2018). A targeted benefit has eligibility criteria, based on income (or “means”) or on characteristics that are typically thought to be highly correlated with poverty such as the number of children or elderly in a household.
B. Means-tested Transfers
3. In theory, the case for means-tested targeting is straightforward.3 In the presence of a budget constraint, an ability to perfectly target transfers to lower-income (or “poor”) households based on their incomes will result in a greater increase in social welfare (or decrease in poverty) compared to an untargeted benefit (i.e., a UBI).4 For instance, if the transfer budget is just sufficient to eliminate poverty, then perfect targeting will result in each poor household receiving a transfer equal to the gap between its income and the poverty income line; non-poor households will not receive a transfer. Therefore, households receive a transfer equal to the gap between their “means” and their “basic needs”. However, in the absence of perfect targeting, some poor will be excluded from the program, while some non-poor will be included, so that the poverty impact will be lower. Or some poor households may be included but receive lower transfers compared to perfect targeting.
4. In practice, many countries do not have the capacity to implement perfectly targeted transfers based on sophisticated means tests. This may reflect low administrative capacity, a large “informal” sector constituting small-scale and self-employment activities, individuals having multiple and volatile sources of income (including in-kind income), and poor or non-existent bookkeeping. This makes verification of income very difficult, especially for low-income individuals. There may also be a reluctance to do such means testing for social or political reasons (e.g., beneficiary stigma or middle-class support for redistribution). Or the costs of individuals acquiring sufficient capacity to comply (or understand) may be deemed undesirable or prohibitive.
C. Categorical Targeting
5. Where effective means testing is not feasible, an alternative approach is “categorical” targeting.5 Under categorical targeting, eligibility for transfers is based on such characteristics as the presence of children (child benefit) or elderly (social pensions) in the household, the location of the household (living in poor areas) or being disabled or in ill health. Such characteristics may be highly—but imperfectly—correlated with being poor. For instance, while the poverty rate for households with children or elderly may be relatively high, not all households with children or elderly are poor, or if they are poor they are not equally so. Categorical targeting can also be used to differentiate the level of transfers (as opposed to just determining eligibility for a uniform transfer) across households based on these categorical characteristics.6
6. The imperfect nature of categorical targeting gives rise to a trade-off between poverty impact, coverage of the poor, and fiscal cost. While restricting transfers to households with children (say as opposed to a UBI) may help to channel a larger share of the poverty budget to the poor, and thus have a larger poverty impact, poor households without children will be excluded, while non-poor with children are included.7 Coverage of the poor can be increased by expanding eligibility to, say, older children or the elderly.8 Figure 1a illustrates the trade-offs involved— between coverage of the poor, leakage to the rich, the transfer levels received by the poor and fiscal cost— using simulations based on household survey data. Uniform benefits for children up to 5 years are very progressive since a high percentage of transfers go to lower-income groups; coverage of the bottom quintile is around 50 percent, falling to around 15 percent for the top quintile (Figure 1a, orange line). Expanding eligibility to children up to 10 years or to include the elderly would help increase overall household coverage, including coverage of lower-income groups. Moving to universal benefits would obviously ensure universal coverage but, under a fixed budget, also require lower transfer levels per household across all income groups. The choice between universal and categorical transfers therefore involves a trade-off in terms of poverty impact, coverage of the poor, and the size of the transfer budget (and therefore required tax levels).
Coverage Under Alternative Categorical Programs
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Note: Income (or welfare) deciles are based on household per capita income, with 1 being the poorest and 10 the richest. In the survey, 33 percent of households have children aged 0–5 years, 50 percent have children aged 0–10 years, and 62 percent have either children aged 0–10 years or elderly 65 years or above. Over 35(60) percent of children aged up to 10 years are in the bottom 2(4) income deciles, with very little variation across age levels. Compared to child and elderly transfers, the share of benefits accruing to the bottom five deciles is always higher under PMT targeting at the 50th percentile.Coverage Under Alternative Categorical Programs
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Note: Income (or welfare) deciles are based on household per capita income, with 1 being the poorest and 10 the richest. In the survey, 33 percent of households have children aged 0–5 years, 50 percent have children aged 0–10 years, and 62 percent have either children aged 0–10 years or elderly 65 years or above. Over 35(60) percent of children aged up to 10 years are in the bottom 2(4) income deciles, with very little variation across age levels. Compared to child and elderly transfers, the share of benefits accruing to the bottom five deciles is always higher under PMT targeting at the 50th percentile.Coverage Under Alternative Categorical Programs
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Note: Income (or welfare) deciles are based on household per capita income, with 1 being the poorest and 10 the richest. In the survey, 33 percent of households have children aged 0–5 years, 50 percent have children aged 0–10 years, and 62 percent have either children aged 0–10 years or elderly 65 years or above. Over 35(60) percent of children aged up to 10 years are in the bottom 2(4) income deciles, with very little variation across age levels. Compared to child and elderly transfers, the share of benefits accruing to the bottom five deciles is always higher under PMT targeting at the 50th percentile.D. Proxy-means Testing
7. Targeting eligibility based on proxy-means tests (PMTs) also results in leakage and undercoverage. This approach, which is the subject of much debate given its increasing importance in practice, attaches a continuous score to households based on various household characteristics strongly correlated with welfare, often based on the coefficients from a regression analysis of income or consumption on these characteristics. It has been argued that, by design, this approach is prone to significant leakage and undercoverage of the target poor population, especially of the poorest (Brown and others, 2016). Figure 1b illustrates the trade-offs using the same survey data as above. Each line, going from bottom to top, shows the change in coverage across income deciles as the program increases from 10 percent of the population to 100 percent based a standard form of PMT. Under all PMT schemes, coverage is substantially higher for lower-income groups than for higher-income groups. As the program expands upward from 10 percent of the population, coverage of lower-income groups increases significantly, reaching around 80–90 percent for the bottom quintile at 40 percent coverage. If the objective is to ensure almost universal coverage (say above 80 percent) of each of the bottom three deciles, then the program would need to expand to 50 percent of the population.
8. PMT targeting can be designed to outperform categorical child and elderly targeting in terms of both coverage and benefit incidence. Coverage of lower-income groups is higher under the PMT covering 50 percent of the population compared to categorical transfers (Figure 1a). Combined with the sharp drop-off in coverage over higher-income groups, this results in a higher share of the transfer budget going to lower-income groups under PMT. Therefore, for a given budget, the PMT will typically have a larger poverty reduction impact and better coverage of lower-income groups than under the categorical targeting. The poverty impact could be increased further by differentiating transfers by household size and composition (e.g., using the PMT to target child transfers). However, the random nature of exclusion and exclusion around the eligibility cut-off score, and the associated lack of transparency in defining eligibility, can generate significant community discontent as they observe that poor households are excluded while better off households are included. This issue of horizontal inequity is inherent to PMT. In addition, the structural nature of the underlying statistical approach means that the PMT scoring system needs to be regularly updated.
9. An alternative is to use the PMT only to differentiate benefit levels across the population with universal coverage. Benefit differentiation could be based on the PMT score. This would help to eliminate undercoverage of poor beneficiaries (however defined), although not all poor beneficiaries would have the same transfer. Figure 2a shows the outcome in terms of share of benefits accruing to the poorest (and richest) 30 percent of the population. Under the UBI, by design, the poorest 30 percent receive 30 percent of the fixed transfer budget. The second set of bars show the share under a “tiered PMT” where the ratio of benefits received by individuals is 4:2:1 across the lowest three PMT deciles, the next four deciles, and the highest three deciles, respectively. This increases the share of benefits accruing to the bottom three deciles to over 40 percent, while that for the richest three deciles decreases to just above 15 percent. In addition to having a bigger poverty impact, this also eliminates eligibility undercoverage. Although the PMT that targets half the population has a slightly higher share of benefits accruing to the bottom three deciles (Figure 2a, fourth set of bars), it also comes with significant undercoverage of lower welfare deciles (Figure 2b, line). Tiering benefits also eases, but does not completely eliminate, horizontal equity concerns.
Benefit Share, Benefit Level, and Coverage
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Note: In panel b, chart bars show the share of each decile receiving different benefit levels under the ratio 4:2:1 for the bottom three deciles, next four deciles, and top three deciles.Benefit Share, Benefit Level, and Coverage
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Note: In panel b, chart bars show the share of each decile receiving different benefit levels under the ratio 4:2:1 for the bottom three deciles, next four deciles, and top three deciles.Benefit Share, Benefit Level, and Coverage
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Note: In panel b, chart bars show the share of each decile receiving different benefit levels under the ratio 4:2:1 for the bottom three deciles, next four deciles, and top three deciles.E. Financing Transfers
10. Coverage expansion needs to be financed through progressive and efficient taxation.
This strategy should include:
Strengthening personal income taxes (PITs). Where administrative capacity is low, this could first focus on broadening coverage of taxes on wages and salaries and paying particular attention to taxation of higher incomes. From the perspective of fiscal redistribution, this allows the claw back of universal transfers from these income groups and can reduce reliance on other less progressive tax instruments. Strong PITs should also be reinforced by effective taxation of corporate income.
Strengthening consumption taxes. Broad-based consumption taxes play a key role in increasing tax capacity in developing economies (Coady, 2018). Efficiency requires that differentiation of consumption tax rates across goods be minimized, and the strengthening of the social safety net through expanding coverage of lower-income groups greatly dilutes the case for preferential rates on income distribution grounds. Setting the tax registration threshold at a reasonably high level can also enhance the progressivity of the consumption tax burden since smaller scale businesses typically have lower incomes and lower-income groups often buy from small-scale retailers. In a high inequality setting, i.e. where higher-income groups account for a disproportionately high share of total consumption, significant redistribution can be achieved through simple tax and transfer systems. For instance, a UBI financed by higher consumption taxes can be a feasible and efficient approach to redistributing income and protecting the poor (Figure 3).
Expanding use of efficient excises. Taxation of consumption (in addition to standard consumption taxes) that generates negative externalities can raise significant revenues in an efficient an equitable manner. For instance, increasing taxes on fossil fuel energy presents a “win-win” opportunity in terms of helping to reduce domestic and global pollution (and associated health damage) and ensuring a progressive distribution of the tax burden (Figure 3). Other candidates for excise taxes on efficiency grounds include alcohol, tobacco and possibly sugar. Note that since broadening the consumption tax base can increase the tax burden on vulnerable groups, it is important that the safety net is capable of protecting these by ensuring they are covered and transfers are increased accordingly (Lustig, Pessino and Scott, 2013). In addition, these tax policies will also often require significant investment in strengthening revenue administration systems, which (together with good tax policy) can also help to fight tax evasion and avoidance, both domestic and cross-border.
Distributional Impact of Tax and Transfer Programs
(percentage change in household per capita consumption)
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Distributional Impact of Tax and Transfer Programs
(percentage change in household per capita consumption)
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.Distributional Impact of Tax and Transfer Programs
(percentage change in household per capita consumption)
Citation: Policy Papers 2019, 017; 10.5089/9781498318907.007.A004
Source: Calculations based on India’s 2011–12 National Sample Survey.References
Atkinson, Anthony B., 1995, “On Targeting and Family Benefits,“ in Incomes and the Welfare State: Essays on Britain and Europe, ed. by Anthony Barnes Atkinson, pp. 223–261 (Cambridge: Cambridge University Press).
Besley, Timothy and Ravi Kanbur, 1993, “The Principles of Targeting,” WB Policy Research Working Paper No. WPS385 (Washington: The World Bank Group).
Brown, Caitlin, Martin Ravallion, and Dominique van de Walle, 2017, “Reaching Poor People,” Finance & Development, Vol. 54, No. 4, pp. 48–51.
Brown Caitlin, Martin Ravallion, and Dominique van de Walle, 2016, “A Poor Means Test? Econometric Targeting in Africa,” NBER Working Paper No. 22919.
Coady, David, 2018, “Creating Fiscal Space,” Finance & Development, Vol. 55, No. 4, pp. 24–27.
Coady, David, Margaret Grosh, and John Hoddinott, 2004a, “Targeting of Transfers in Developing Countries: Review of Lessons and Experience,” WB Regional and Sectoral Studies No. 30230, (Washington: The World Bank Group).
Coady, David, Margaret Grosh and John Hoddinott, 2004b, “Targeting Outcomes Redux,” The World Bank Research Observer, Vol. 19, No. 1, pp. 61–85.
Francese, Maura, and Delphine Prady, 2018, “Universal Basic Income: Debate and Impact Assessment,” IMF Working Paper No. 18/273 (Washington: International Monetary Fund).
Gentilini, Ugo, 2018, “What Lessons for Social Protection from Universal Health Coverage?,” Let’s Talk Development Blog, August 22.
Independent Evaluation Office (IEO), 2017, “The IMF and Social Protection,” 2017 Evaluation Report (Washington: International Monetary Fund).
International Monetary Fund (IMF), 2017,”Fiscal Monitor: Tackling Inequality,” (Washington: International Monetary Fund), October.
Lustig, Nora, Carolina Pessino and John Scott, 2013, “The Impact of Taxes and Social Spending on Inequality and Poverty in Argentina, Bolivia, Brazil, Mexico, Peru, and Uruguay,” Public Finance Review, Vol. 42, No.3, pp. 346–67.
Picketty, Thomas, and Emmanuel Saez, 2013, “Optimal Labor Income Taxation,” Handbook of Public Economics, Vol. 5, Ch. 7, pp. 391–474.
Rutkowski, Michal, 2018, “Reimagining Social Protection,” Finance & Development, Vol. 55, No. 4, pp. 10–13.
van de Walle, Dominique, 1998, “Targeting Revisited,” The World Bank Research Observer, Vol. 13, No. 2, pp. 231–48.
Prepared by David Coady and Nghia Piotr Le (FAD).
For more detailed discussion, see Atkinson (1995); Besley and Kanbur (1993); Coady, Grosh, and Hoddinott (2004a,b); and van de Walle (1998).
This ignores the issue of work disincentives related to the withdrawal of benefits as income increases, which is a concern in all countries regardless of level of development. The theory of optimal income distribution emphasizes the important role of gradual withdrawal of means-tested benefits with income levels to efficiently manage the trade-off between work disincentives (efficiency) and redistribution (equity) (Picketty and Saez, 2013).
This is often referred to as “tagging” or “statistical targeting.” For a broader discussion of targeting alternatives, see Coady, Grosh, and Hoddinott (2004a,b).
Note also that categorical targeting can reduce efficiency costs when the categories (or “tags”) used are linked to household or individual characteristics that cannot be easily changed or hidden.
Although child benefits are often described as being “universal,” from a redistributive perspective they are just a different form of targeting based on the demographic composition of a household.
The approach of expanding coverage across groups over time is similar to the concept of “progressive universalism”, which is generating broad support (Rutkowski, 2018; Gentilini, 2018).
