93719 Safety Nets Social Protection & Labor Technical Note June 2014 | Number 7 Highlight Defining Eligibility for Social As a poverty reduction tool, social pensions Pensions1 are less effective than broader social assistance A View from a Social Assistance Perspective policies to the poor Margaret Grosh and Phillippe G. Leite because the elderly do not necessarily represent Developing and low-income countries around the world have been unable to provide old-age income security to all through contributory pensions. Neither reforms to the large majority of pension systems nor general growth and development have helped countries to increase the poor population, coverage much. In frustration, many countries are considering or have implemented nor are they necessarily non-contributory social pensions, aiming to reduce poverty and vulnerability among the elderly. It is therefore important to consider how such programs fit into social poorer than average. assistance programming. This note discusses the horizontal equity of programs that are Social pensions should solely for the elderly and the option of targeting such programs differently. It then sheds be considered as part empirical light on the issue with simulations based on data from the Kyrgyz Republic, Niger, Panama and Yemen. of a social protection system, integrated with Are Special Programs Needed for the Elderly Poor? long-term government Proponents of social pensions take an axiomatic view that the elderly are poorer than strategies for poverty the rest of the population. However, the literature and our data analysis do not confirm reduction and risk this. Kakwani and Subbarao (2005), Guven and Leite (2014) and Devereux (2001) show mitigation. Harmonizing that the elderly can be better off than other population groups in Sub-Saharan Africa. Whitehouse (2000) found that in most studies of middle- and higher-income countries, and integrating social the old are proportionately or under-represented among the poor. Pal and Palacios pensions with other (2011) found similar results for India. Evans and Palacios (mimeo) found that poverty assistance will lower rates were lower for the elderly than for the non-elderly in about half of the 62 countries they studied, and that children were poorer than the elderly in many countries. This administrative costs and literature highlights the importance of conducting country-specific analysis before guard against unequal assuming that universal social pensions are the right policy tool to address poverty in a treatment of different given region. groups at the same Therefore the first step in formulating social assistance policy should be to diagnose welfare level. poverty and vulnerability before giving the elderly high priority. How many elderly persons are poor or at risk of poverty? What are their characteristics? What are the causes of their poverty and vulnerability? How poor are they vis-à -vis other groups? 1 Synthesis note based on Margaret Grosh and Phillippe Leite’s “Defining Eligibility for Social Pensions: A View from a Social Assistance Perspective.” In Closing the Coverage Gap: The Role of Social Pensions, eds. R. Holzmann, D. Robalino, and T. Takayama (Washington DC: World Bank, 2001). Technical Note June 2014 | Number 7 Figure 1. Poverty Rates of Different Age Groups 45% 40% Share of individuals below poverty line 35% youth 30% 25% 20% Source: Evans and Palacios (mimeo)—Figure 5 15% Note: Figure compares poverty rates using the 50 percent of median poverty line for youth, working age and elderly individuals in each country. The higher 10% line refers to youth poverty rates while the bottom line refers to elderly poverty rates. The horizontal 5% elderly axes represents the 62 countries ordered from the lowest to the highest poverty rate. 0% Figure 2: Share of Elderly in Sub-Saharan Africa by Quintile of Welfare Distribution 1st 2nd 3rd 4th 5th 35 30 25 20 % 15 10 5 0 a e ria o da ia a li a us da wi ny qu ng bi an Ma ib an iti ge la an m Ke bi Gh Co m Ma ur Ug Ni Za Rw am Na Ma of oz p Re M Source: Guven and Leite (2014)—Figure 5 Is the common lifetime income path of reliance on their lacking, then a social pension—especially a universal one— own or their partner’s wage earnings suddenly eliminated by violates the principle of horizontal equity by implying that retirement, with no replacement via pension? Do the elderly the elderly are somehow more worthy of support than other abruptly withdraw from productivity and earnings, as in needy groups, such as poor children, persons with disabilities, formal sector retirement, or do they continue their economic working families with low earnings, and so on. This issue will activities, possibly at diminishing levels or with diminishing be especially severe in low-income countries, where social earnings? Do they live alone or in families with other earners? assistance is most lacking, contributory pensions are least Do families pool income across members of different ages? common, the coverage gap is largest, and social pensions are The answers to such questions are not only country-specific, thus of most interest. but vary depending on whether the elderly had formal employment, the sector in which they worked, and perhaps In countries with programs for other groups, horizontal their ethnicity, if that affects household structure. inequity may be less marked but does not vanish, as separate programs are rarely fully equal. The question of efficiency also The poverty diagnostic should also include other groups arises: should income support be provided by a social pension, besides the elderly. If programs that serve other groups are or by including the elderly in other social assistance programs? Social Protection & Labor | World Bank Group A single system offers obvious administrative advantages for Much is known about the pros, cons, and requirements targeting, payments, monitoring and evaluation, and so on. of targeting systems for the general population. Below, we Furthermore, by integrating groups into a single program, examine whether such methods might work equally well for the issue of which is more worthy of support is avoided. the elderly. Integration of the elderly poor into a poverty-targeted social assistance program is thus the preferred option. Assessing the Impact of Social Pensions on Poverty Of course, the elderly are always implicitly integrated into We conducted an empirical analysis to assess the impact general needs-based programs, even in four of the best-known of social pensions on poverty using national representative conditional cash transfer programs—those in Brazil, Ecuador, household surveys that show poverty determinants and Jamaica, and Mexico—which are usually thought of as serving poverty incidence for different population groups. To show only children. Broader social assistance programs can be how poverty profiles differ across countries, four were studied: similarly modified to ensure that the elderly are well served the Kyrgyz Republic, Niger, and the Republic of Yemen, which through adjustments to eligibility rules, the determination are poor, and Panama, which is middle-income. of benefits, or other program requirements. For example, the eligibility threshold for Bulgaria’s guaranteed minimum Most elderly in the four countries live in households with income program is higher for families with elderly members. working-age adults; only 6 percent live without such support In Jamaica’s PATH program, the proxy means test was adjusted in the Republic of Yemen, and only 10 percent in Niger. to lower the weight given to housing assets, allowing significant Households anchored by the elderly are more common in numbers of elderly living alone to participate. The elderly also the Kyrgyz Republic and Panama, where almost a third of the receive their full payment even if children in the household elderly live in households without working-age adults. In the default on the conditions pertinent to them and fail to qualify three poor countries with less formal labor markets, about two- for their own benefits. In the U.S. food stamp program, the thirds of elderly members contribute income to the household. certification period is longer for elderly-headed households In Panama, only a third make such contributions. than for others, since the earnings of the elderly are less likely to change. In Romania’s guaranteed minimum-income Overall, about one in five households includes an elderly program, the elderly are exempt from the public service member, but these households are not uniformly poor. Among requirement. the 20 percent lowest-consumption households, those with elderly are somewhat poorer in the Kyrgyz Republic and Should Separate Social Pension Programs be Panama. The difference is more marked in Niger, but not as great Universal or Targeted? in the Republic of Yemen. The highest poverty rates are found in households whose elderly do not earn income, but that category Fiscal constraint is a pervasive concern in social assistance, accounts for only a quarter to a third of households with elderly especially in countries where the safety net is incomplete as members, less than 10 percent of all households. measured by coverage or the adequacy of benefit levels. The costs of universal social pensions are not trivial, particularly Missing-generation households (composed of elderly viewed against spending on other needy groups. Moreover, household heads and children, but no other adults) are not the fiscal costs of social pensions will present an increasing very numerous in countries without severe HIV epidemics. challenge with demographic change, especially in the countries In the Kyrgyz Republic and in the Republic of Yemen where the population is aging most rapidly. Countries with poverty rates for these households are about the same as large social pensions can spend more than 50 percent of their for all households. In Panama the rates are much lower, total assistance budget on them despite the fact that social about 5 percentage points below the national average, and pensions cover a small share of the population.2 In Thailand in Niger they are more than 10 points below the average. the cost of social pensions is projected to more than double The average or low level of poverty in missing-generation between 2012 and 2040 (0.44 percent to 0.98 percent of GDP) households contrasts with the popular impression that due to demographic changes,3 while the universal social these households are destitute. The poverty gap for missing- pension in Mauritius, already more than 2 percent of GDP, generation households is higher than, or close to, the average could eventually absorb almost 7 percent of GDP4 These for all households in all four countries, despite their lower- expected cost increases are raising government concerns than-average poverty headcount rates. The low or average about the sustainability of social pensions over the long run. headcounts imply that elderly persons living in households Therefore, targeting and or contributory system reforms should without working-age adults are mostly those who can afford be considered. it, but the higher poverty gap indicates that there is a subset of missing-generation households that are very poor indeed. 2 Monchuk (2014) shows that in Lesotho almost 50 percent of spending Again, although these households are not numerous, they are in safety nets is for universal social pensions, while in Mauritius, non- contributory old age pensions account for 87 percent of total government of salient policy concern—the figurative “poster child” for social assistance spending. 3 Jitsuchon et al. (2012) Guven and Leite (2014) 4 Technical Note June 2014 | Number 7 Figure 3: Poverty Levels and Household Composition in Niger HHs without elderly; 83% 65% HHs with elderly 25% 17% 6% 4% Elderly contributing to family income Elderly not contributing to family income Missing generation Elderly only Total* (head's characteristic) HHs without elderly poverty rates Elderly only poverty gap Missing generation Elderly contributing to family income Elderly not contributing to family income 0% 5% 10% 15% 20% 25% 30% 35% Source: Grosh and Leite (2009)—Figure 12.1 social pensions. pension programs, we compare two schemes. One covers all elderly (universal coverage), while the other targets only the The size and age structure of households may affect whether elderly poor. In both scenarios, the potential beneficiaries are households with the same consumption level per capita live age 65 and older. equally well. There may be economies of scale; it may cost as much to boil a pot of rice, provide a television set, or heat a For the universal social pension scheme, we estimate that house for one person as for three or six people. And not all about 20 percent of households in all four countries contain members may need the same things to be equally well off: at least one elderly person. The elderly number 321,000 in the children need fewer calories from food than adults, prime-age Kyrgyz Republic and 210,000 in Panama, 6 to 7 percent of the adults need less health care than elderly persons or children, population. They total 355,000 in Niger and 688,000 in the the elderly need no school fees or textbooks, and so on. Republic of Yemen, 2 to 3 percent of population. Simulating the Effect of Universal and Targeted For the targeted social pension scheme, household per capita Social Pensions on Poverty Rates income is used as the welfare measure. The elderly in the poorest 20 percent of households are considered eligible. To help understand how targeting affects the outcome of social Social Protection & Labor | World Bank Group We find that a universal social pension in all four countries Kyrgyz Republic, Panama, and the Republic of Yemen. Errors would be mildly progressive because households with elderly of exclusion are higher in Niger, the poorest country in the members are somewhat poorer than average. However, a sample. Inclusion errors are low, with only 5 to 20 percent of substantial share of the benefits would go to the non-poor, those households predicted to be poor not actually being poor. because most of the elderly—more than 80 percent—live in Again, performance is substantially better in Panama than households that are not poor. In the Kyrgyz Republic, 85.4 among the poorer countries and is worst in Niger, in keeping percent are non-poor; in Niger, 80.2 percent; in Panama, 88 with the general experience that it is more difficult to target percent; and in the Republic of Yemen, 83.4 percent in the narrowly in countries with very low incomes. Republic of Yemen. Moreover, the universal social pension would only partly address overall poverty, as a strong majority With a fixed budget, the generosity of simulated programs varies of poor households have no elderly: 79 percent of those in the widely according to the number of beneficiaries selected. Transfers Kyrgyz Republic, 78 percent in Niger, 79 percent in Panama, for universal social pensions represent between a tenth and a third and 80 percent in the Republic of Yemen. of the amount possible with the targeted options. Because of this smaller transfer amount, the universal social pension has less The targeted social pension scheme yields fewer participants effect on poverty. It also is less cost-effective, since most elderly are than the universal program. Thus, either the unit benefit can not poor. Separate estimation of the model for households with be increased or the overall program cost can be reduced. As and without elderly increases cost-effectiveness in every country ever, targeting induces exclusion errors and reduces inclusion except the Kyrgyz Republic. errors. As much as 40 percent of the target group may be excluded (but most often, those excluded are around the When the budget is not fixed, the cost of the universal program threshold of eligibility). Coverage in the poorest decile is quite increases to more than 0.5 percent of GDP in all countries, but has high, with only around 10 percent of this group excluded in the no impact on cost-benefit ratios compared with the fixed budget Figure 4: Simulated Results of Alternate Policies: Fixed Costs, Kyrgyz Republic, Niger, Panama and the Republic of Yemen Annual Cost Number of Transfer per Average ∆FGT(0) ∆FGT(1) ∆FGT(2) Correlationb Cost benefitc (LCU) beneficiaries beneficiary per capita (percent)a (percent)a (percent)a consumption of families with beneficiaries Kyrgyz Republic (potential beneficiaries 46,946) Universal Social Pensions 3,913.3 321,247 1,015 14,695 1.6% 3.2% 4.4% 0.14 Means test 3,913.3 33,530 9,726 8,204 3.6% 7.9% 9.4% 0.561 0.34 PMT 1 3,913.3 46,177 7,062 8,473 3.8% 10.6% 14.7% 0.686 0.46 Niger (potential beneficiaries 70,473) Universal Social Pensions 7,439.5 355,409 1,744 11,570 3.0% 4.1% 5.2% 0.18 Means test 7,439.5 55,624 11,145 8,601 2.4% 6.0% 7.8% 0.188 0.26 PMT 1 7,439.5 61,738 10,042 5,060 4.6% 9.8% 12.1% 0.461 0.43 Panamad (potential beneficiaries 24,058) Universal Social Pensions 330.7 210,679 131 2,105 2.0% 2.2% 2.7% 0.11 Means test 330.7 28,948 952 632 3.5% 7.6% 10.5% 0.779 0.38 PMT 1 330.7 20,904 1,318 528 5.1% 9.2% 12.8% 0.802 0.46 Yemen (potential beneficiaries 114,552) Universal Social Pensions 122,273.5 688,216 14,806 123,641 1.9% 3.4% 4.2% 0.15 Means test 122,273.5 205,038 49,695 136,298 1.7% 3.3% 3.9% 0.099 0.15 PMT 1 122,273.5 117,230 86,919 58,836 5.1% 8.3% 9.7% 0.331 0.37 Source: Grosh and Leite (2009)—Figure 12.6 Note: Potential beneficiaries are elderly persons (age 65 or older) living in poor households. The poverty line is equal to the first quintile (Q1) maximum household per capita consumption. Universal social pensions represent a transfer T to all the elderly population (age 65 or older). The means test is defined under household per capita income. PMT 1 is estimated jointly for households with and without elderly members. a. Poverty measures are computed for household per capita consumption, setting elasticity of consumption equal to 1. Only households with elderly members are considered here. b. For the means test, the correlation between household per capita consumption and household per capita income is used; for the proxy means test, the correlation between household per capita consumption and household predicted per capita consumption is used. c. Reduction, in U.S. dollars, in the poverty gap for each dollar spent in the program. d. Cost is set as 0.1 percent of per capita GDP. Technical Note June 2014 | Number 7 case. The cost for all of the targeted programs fall to less than 0.2 Universal pensions reach all the elderly poor, but most resources percent of GDP, and the cost-benefit ratio increases compared go to the non-poor and so the programs are not very cost- with the fixed-budget case. The universal social pension has a effective. In a budget-constrained environment, this means higher effect on headcount poverty because there are no exclusion that the benefit level is likely to be so low that the policy cannot errors. (In addition, when the benefit is fixed, the budget is provide adequate benefits to the elderly poor it does reach. higher.) Again, however, it has the lowest cost-effectiveness ratio Targeted social pensions are much more cost-effective per dollar because the great majority of elderly are not poor. spent, and with fewer beneficiaries they could, for a fixed budget, convey a higher benefit. But they entail some errors of exclusion, Despite the much higher budget for the universal social the rates of which are quite variable by country. pension, the targeted program has close to the same impact on the poverty gap as a result of its higher cost-effectiveness. Considering the findings and the need for cost-effective programs due to financial constraints, particularly in poor Conclusions countries, the universal approach is undesirable in many cases. If a targeting system is to be devised, there is no prima facie The issues surrounding the targeting of social pensions are case for a social pension system to be run separately from a not very different than for other aspects of social policy. It is general needs-based social assistance program. Integration will important to consider the situation of the target group by itself lower administrative costs and guard against unequal treatment and in relation to others, and to consider all available options. of different groups at the same welfare level. References Devereux, Stephen. 2001. Social Pensions in Namibia and South Africa. IDS Discussion Paper 379. Brighton: Institute of Development Studies, University of Sussex. Evans, Brooks and Robert Palacios. mimeo. How poor are the old relative to other age groups? World Bank. Washington, DC. Grosh, Margaret Ellen and Phillippe G. Leite. 2009. “Defining Eligibility for Social Pensions: A View from a Social Assistance Perspective” In Closing the Coverage Gap: The Role of Social Pensions, eds. R. Holzmann, D. Robalino, and T. Takayama. Washington, DC: World Bank. Guven, Melis and Phillippe G. Leite. 2014. “The Slippery Slope: Explaining the Challenges and Effectiveness of Social Pensions to Fight Poverty in Sub-Saharan Africa.” The World Bank. PowerPoint. Jitsuchon, Somchai, Emmanuel Skoufias, and Mitchell Wiener. 2012. “Reducing Elderly Poverty in Thailand: The Role of Thailand’s Pension and Social Assistance Programs.” Working paper 80527. World Bank. Washington, DC. Kakwani, Nanak, and Kalinidhi Subbarao. 2005. “Ageing and Poverty in Africa and the Role of Social Pensions.” Social Protection Discussion Paper 0521. World Bank. Washington, DC. Monchuk, Victoria. 2014. Reducing Poverty and Investing in People: The New Role of Safety Nets in Africa. Washington, DC: World Bank. Pal, Sarmistha and Robert Palacios. 2011. “Understanding Poverty among the Elderly in India: Implications for Social Pension Policy.” Journal of Development Studies 47(7): 1017-1037. Whitehouse, Edward. 2000. “How Poor are the Old? A Survey of Evidence from 44 Countries.” Social Protection Discussion Paper 0017. World Bank. Washington, DC. The findings, interpretations, and conclusions expressed herein are those of the author(s), and do not necessarily reflect the views of the International Bank for Reconstruction and Development/The World Bank and its affiliated organizations, or those of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work.