Old Age Security and Social Pensions

Transcription

1 Old Age Security and Social Pensions Anita M. Schwarz Social Protection World Bank May 2003 I would like to thank Margaret Grosh, Kalinidhi Subbarao, Robert Palacios, Richard Hinz, Stefano Scarpetta, Asta Zviniene, Aniruddha Bonnerjee, and Montserrat Pallares for helpful comments. The errors and omissions are all mine.

2 2 Old Age Security and Social Pensions Currently, the UN estimates that there are 606 million people in the world over the age of 60, roughly 10% of the population, with the number expected to more than double to 1.6 billion by 2050, reaching 19% of the world s future population. Of this elderly population, already 62% live in developing countries, but by 2050, 80% of the world s elderly will live in developing countries. Given the fragile nature of old age security arrangements in developing countries, many of these elderly face the risk of poverty in their old age. As individuals age, their capacity for work diminishes, forcing them to seek other means of support. Traditionally, older individuals could rely on their families for support, 1 but as societies move away from tradition-bound multi-generational families to more modern nuclear families in the process of development, the elderly are often left with little support. But even under traditional structures, the safety net was not perfect as there were always people without children to support them or whose children were too poor to support them. The development process exacerbates the imperfections of the safety net as children move away to urban areas, leaving the elderly and their obligations to them behind. At the same time, as the family s source of income shifts from traditional sources to the market, the older individual s ability to contribute to family income declines, resulting in a decline in status for the elderly and a decline in the willingness of other family members to care for the elderly. An alternative source of old age support is the savings that an individual generated during years of working, whether the savings occur on one s own or through some type of savings/pension mechanisms offered through place of employment. This safety net is also imperfect in that the lifetime poor are unlikely to have savings or to have participated in enough formal sector labor activities to have employment related old age support. Safe and reliable long term savings vehicles often do not exist and even where they do, rarely are there opportunities to convert lump sum savings into annuities or other instruments which provide a stream of payments throughout one s retirement. Even reliance on non-financial assets such as livestock or land entails risks since the livestock can die and the land can be barren due to weather conditions. A final source of old age support comes from public old age programs, whether limited to those who have contributed or more widely available to those who have not contributed as well. Public contributory programs in most developing countries cover civil servants and in the majority of the cases also cover workers in public enterprises, although sometimes through separate programs. In many countries formal economy workers in the private sector are also covered, albeit through yet separate programs. In some countries, informal, self-employed, and rural workers are also covered, whether on 1 A relatively large body of literature refutes the notion that traditional families did in fact always care for the elderly, however prevalent this idea may be. Examples include Burman (1996), p. 590 where she suggests that 80% of rural South African widows had no source of income and turned to neighbors and friends for help before turning to children.

3 3 a voluntary or mandatory basis, sometimes through separate programs and sometimes through the main scheme. While almost all OECD countries have provisions for noncontributory old age pensions, not all developing countries have such programs. In yet other countries, a universal pension is publicly provided, where the main criterion for receiving the pension is age. Coverage Under Contributory Programs. Accumulating information on coverage under the contributory program is more difficult than it initially appears. Rarely are the appropriate data available. In a few countries where the World Bank has actively collected pension data, disaggregated pension data by age of pensioner are available. These are shown in Figure 1-3 and in Appendix Table % % 80.00% 60.00% 40.00% 20.00% Elderly Receiving Pensions Eastern Europe 0.00% Croatia Estonia Hungary Kyrgyz Lithuania Macedonia Malta Moldova Poland Portugal Romania Slovakia Slovenia Turkey Ukraine Figure 1 shows the case of Eastern Europe where it generally appears that a large number of elderly, both male and female are currently receiving pensions. While these numbers accurately reflect the status quo in the 1990 s, there is reason for concern in even this region as the transition to a market economy has led to drops in formal labor force participation, particularly among women, as well as more informalization of the economies, with increasing numbers of workers not contributing to the social security system. As a result, the same graph in 20 years and certainly in 40 years is destined to show far fewer coverage. But even within this generally high rate of coverage, there are clearly pockets of people who are not receiving pensions. Furthermore, this graph does not address the issue of whether the pensions being received are sufficient to avoid poverty. In some cases such as Kyrgyz Republic, the pensions are frequently below the poverty line.

4 4 Figure 1 Elderly Receiving Pensions Latin America 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Argentina Bolivia Brazil Chile Colombia Costa Rica Dom Republic Ecuador El Salvador Mexico Nicaragua Paraguay Peru Uruguay Figure 2 shows the case of Latin America. Aside from the largest and highest income countries, few countries provide pensions to the majority of the elderly. Furthermore, this information from the 1990 s may be optimistic for the future, but for different reasons than in the case of Eastern Europe. Many of the Latin American countries have undertaken substantial reforms which impose much higher eligibility requirements than existed in the past. For example, in Argentina, the minimum number of years of service have risen to 30 years from 25 in the past; similarly in Uruguay to 35 years from 30 in the past; in Brazil, even the years required for a quasi-contributory pension are rising from 5 in 1994 to 15 in the future. As a result, many of the individuals who qualify today for pensions would not qualify under the rules which apply to future retirees, unless there is a marked change in labor market behavior. Record-keeping has also improved markedly so that while previously the social security agencies could not disprove statements by individuals claiming long work histories, in the future, fraudulent claims will be more easily detected. In other regions of the world, the picture is more dismal, with less than 20% of the population covered on average and in many cases less than 5%. Coverage frequently is limited to civil servants and workers in public enterprises with few others receiving any type of coverage. As shown in Figure 3, there are a few African countries like Mauritius which have a universal pension for all those above a certain age. The southern African countries, like South Africa, Namibia, and Botswana have such universal systems as well as a couple of other countries like Nepal and Hong Kong. While there has been a concerted effort to increase coverage in the non-universal countries which should be

5 5 reflected in broader pension coverage of the elderly in the future, the efforts have been limited and far from successful enough to allay the fears of substantial old age poverty Figure 2 Elderly Receiving Pensions Latin America 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00% Argentina Bolivia Brazil Chile Colombia Costa Rica Dom Republic Ecuador El Salvador Mexico Nicaragua Paraguay Peru Uruguay for the future. The low coverage, for example, in Korea shown below reflects the start up of a new system in 1988, which will result in more pension coverage in the future. Figure % % 80.00% 60.00% 40.00% 20.00% Elderly Receiving Pensions Asia and Africa 0.00% Cape Verde Korea Mauritius Morocco Philippines Senegal Tunisia

6 6 Appendix Table 2 shows the coverage rates among the working age population in a number of countries. These coverage rates calculate the percentage of the active labor force contributing to the pension system at a particular point in time or the number of contributors compared to the working age population, typically defined as ages While the data for this measure provide a better indication of future coverage rates among the elderly than the previous figures, there are still complications. When comparing the contributors to the working age population, the applicable ages for both starting work and ending work vary by country. But more importantly, since the number of years of contributions required to collect a pension are usually below the total number of working years possible, low contribution coverage rates can still result in high coverage among the elderly. For example, until 1996, Brazilians were only required to make 5 years of contributions in order to retire at age 65 for men, and age 60 for women and receive a pension worth 70% of their previous wages. Given that the possible number of working years could span from age 20 or earlier to age 65 for men, at least 45 years, contribution coverage rates of 11.1% could be fully consistent with full coverage of the elderly. Of course, in reality, not everyone in Brazil retired under these provisions, but it does illustrate the large differences possible between contribution coverage and coverage of the elderly. In general, we can estimate that across the world, about 10% of the labor force in South Asia is covered and will receive pensions in the future, with coverage slowly expanding. In Africa, less than 10% of the work force is covered and the coverage seems to be stagnant or even declining in some countries, with younger workers joining informal sector labor markets with no access to pension arrangements. In East Asia, the coverage varies dramatically, from countries like Vietnam with low coverage to countries like Singapore, with coverage in the range of 75%. In Latin America, about a third of the labor force is covered, and there have been few increases over the last two decades. In Eastern Europe, currently about 75% of the labor force is covered, down from 100% in the former socialist days, but coverage continues to decline as workers leave formal sector state enterprises for self-employment or for the informal sector. In the countries where we have information in the Middle East and North Africa, coverage is about 40% and stagnant, given the high rates of unemployment. Are the Elderly Poor? Given the programs that exist and the family safety nets that exist, the question is whether the elderly are in fact poor. Those coming from OECD backgrounds usually do not regard the elderly as particularly poor since so much of the elderly population in these countries is generally covered by relatively generous pension programs. While the precise answer to whether the elderly are in fact poor or poor relative to individuals in other age groups will of course vary from country to country, the evidence suggests that the elderly are poor and the extreme elderly, those over the age of 70, face the greatest risk of poverty of any age group. Figure 4 shows the poverty by age groups for Brazil with and without the pension system. The elderly above the age of 70 are clearly poorer than all other age groups. This result is in line with expectations since the primary mechanism to escape poverty among the poor is labor income and the elderly are less and less able to earn labor income with age.

7 7 Figure 4 Poverty Among The Elderly Bulgaria Nepal Nicaragua Panama Peru Tajikistan Single Man Single Woman Couple HH with elderly HH with no elderly Figure 5 Household survey data from a sample of six countries shows virtually the same results. Figure 5 shows data from LSMS surveys for Bulgaria, Nepal, Tajikistan, Panama, Nicaragua, and Peru. With the exception of Panama, households with no elderly

8 8 tend to be less poor than households with elderly. The poverty rates for single elderly men range from 20% in Tajikistan to 63% in Bulgaria. The poverty rates for single elderly women range from 18% in Tajikistan to 69% in Peru. For elderly couples, the lowest poverty is in Panama with 14%, with the highest in Peru with 53%. The elderly who live in larger households are less likely to be poor than elder-only households, ranging from 19% in Nepal to 32% in Nicaragua. The comparison between multigenerational households with elderly and households with no elderly shows that with the exception of Panama, household with elderly are generally poorer than nonelderly households, with the poverty rates being anywhere from 1% to 29% higher in households with elderly. The elderly get even poorer when whatever pension income they have has been eliminated. Other studies have shown similar results. In Malaysia, while the elderly constitute only 5.9% of the population, they constitute 32% of the poor, suggesting that the elderly are more often poor than other segments of the population. However, a major flaw with most of the household level studies with regards to the elderly is the level of analysis. The usual analysis takes household income and constructs per capita levels of income and then compares these to the poverty line to determine whether an individual is poor or not. Thus, poverty is established for the household as a whole rather than for particular individuals. In more sophisticated versions, the composition of the household is sometimes taken into account, converting all household members into adult equivalents. 2 However, anecdotal evidence suggests that intrahousehold allocation of resources is most certainly not uniform. Furthermore, the elderly often retain control over some portion of their own resources. The Pakistan Aging Study shows that almost 90% of the elderly feel that they control their own resources. Thus, it is possible that households measured as poor contain individuals who are not poor and other individuals who are extremely poor while even households not measured as poor may contain individuals who would be considered poor. Within the household, it is likely that the elderly, particularly if they have access to few resources of their own, are among the most poor. Public Policy Response. All evidence, regardless of the data used, suggests that a large number of the elderly in developing countries have no access to pensions and few other resources so are either poor or at major risk of poverty. As the percentage of the world s elderly in developing countries increases, the risk of poverty increases as well. The concurrent impact of modernization, with its weakening of family bonds, combined with the increased numbers of elderly suggest an urgent need to address the issue of old age security in developing countries. Finding an appropriate response is complicated by the high degree of coresidence, of elderly living with other generations. Figure 6 shows the degree of coresidence among the elderly in the six LSMS countries shown in Figure 5. In all but Bulgaria, among the sample, the vast majority, of the elderly live in multigenerational households. In Nepal, Nicaragua, and Tajikistan, over 90% of the elderly live in these households, while in Panama around 75% live in multigenerational households and in Peru 85%. 2 Similar issues addressed in Deaton and Paxson (1995).

9 9 Degree of Coresidence Bulgaria Nepal Nicaragua Panama Peru Tajikistan Single Living with Partner Living with Others Clearly, when there is such a degree of coresidence, providing a supplement to the elderly person which is designed to raise the individual out of poverty will not be sufficient to do so particularly if the resources are shared among the larger household. For this reason and several other reasons, it might make more sense to offer a general social assistance scheme which applies to all families rather than to just the elderly individual. Since households with elderly tend to be poorer than households without elderly, the assistance will end up subsidizing the elderly more often than other age groups, but this result would fall out of a general means-tested scheme rather than being pursued as a policy objective. Furthermore, the whole issue of what income to attribute to the elderly if the old age scheme is means-tested, the household income or the individual income of the elderly person, becomes highly complicated when the norm is coresidence. The cases of Brazil and South Africa 3 where pensioners are made better off after retirement than before retirement and little social assistance is available for any other age group clearly demonstrate the disadvantages of separate old age assistance from overall social assistance. Returning to Figure 4, while the elderly would be poorer than even children as they age without a pension, the pension results in poverty rates among the elderly falling to well below that of all other age groups. General social assistance programs have also already developed some type of targeting methods and criteria which could then be applied fairly readily to the elderly. However, there are some issues which are more unique to the elderly poor and which might require a special program. 3 See for example Case and Deaton (1998) for South Africa and Schwarzer and Querino for Brazil.

10 10 First, since the elderly generally have a reduced capacity to work, issues of work disincentives which are often central to the design of general social assistance are less applicable to the elderly. Second, while general social assistance often requires frequent requalification as people move in and out of poverty, the elderly, particularly those living as a couple or alone, once in poverty are likely to remain there. Thus, frequent requalification is less necessary in the case of the elderly than in the case of younger age groups. 4 Third, the elderly poor are often people who were not necessarily poor when younger. As such, the stigma attached to applying for social assistance may be a larger obstacle for them than it would for younger people who see such assistance as potentially temporary or expected. Combining the program with the contributory social security program at least administratively could eliminate this stigma and result in some efficiencies arising from one administrative unit maintaining records on all the country s elderly. Fourth, experience around the world seems to suggest that social assistance spending is considered more discretionary than pension spending and as such is more often subject to budget cuts. As a result, some of the most successful poverty reduction programs are those targeted at the elderly precisely because they are less discretionary and continue to function even in difficult budget environments. The U.S. and Brazil are both examples of poverty reduction functioning well through specific old age programs while other more general social assistance programs have not been well funded. Fifth, the receipt of a pension may help to empower the elderly and prevent some amount of elder abuse. The elderly generally have some control over their resources even in a multigenerational household which may provide for a more equitable distribution of resources within the family. The resources that the elderly person brings to the household may also result in the individual no longer being viewed as a burden, but as a valuable member of the family which would elevate both the status of the elderly and the care provided. 5 As many as 11.8% of households not headed by an elderly person in Namibia rely on a pension as the main source of household income. 6 Evidence from Namibia also suggests that the pension of the elderly is used as a household coping device. When income for the household as a whole is low, the whole family exists on the pension of the elderly. It forms a stable source of income in circumstances where other income is highly variable. The stability of the income also allows the family access to credit which would otherwise not be available. 7 Social assistance often has difficulty in dealing with sudden shocks to income, since a 4 However, in the case of multigenerational households, the family income would have to be retested regularly if the pension were based on household income. 5 See Burman (1996), p. 591, for example. 6 Adamchak (1995), p Devereux (2001), p

11 11 scaling up or scaling down of the program often takes time. The same is true in Uruguay where in the 1990 s almost 50% of urban households report some income from pensioners and use this income to reduce the household poverty level and generate some stability in income flows. 8 The evidence also seems to suggest that the elderly do not prefer to live in multigenerational households. A body of literature from the OECD clearly shows a preference for independent living. 9 In the sample of data from the countries in Figures 5 and 6, despite the high rate of coresidence, elderly individuals in the fourth and fifth quintiles have a much lower rate of coresidence than poorer individuals, suggesting that the elderly when they can afford to, prefer to live independently. If the only assistance to the elderly comes from general social assistance, the elderly might be forced into undesirable living circumstances. Providing some income to the elderly themselves may provide the freedom to choose more favorable living circumstances. A related body of literature suggests that even within a multigenerational family context, pension payments to the elderly allow middle aged and younger adults to move out of the family home in search of better employment, leaving the elderly able to care for themselves and for the grandchildren. Thus, the family becomes better off not just by the amount of the transfer but by the increased opportunity to take advantage of better employment. 10 Finally, recognizing the differences between the elderly and other age groups, most countries do have separate programs for the elderly. 11 Direct Support to the Elderly Four basic approaches exist among numerous combinations of these approaches to address the issue of old age poverty First, substantial effort could be put toward increasing coverage under the existing contributory schemes. Second, a noncontributory or social pension could be incorporated within an existing contributory pension framework, with people being given pensions for very few years of work. Third, universal pensions could be offered to all elderly regardless of income level. Fourth, means tested pensions could be offered to all indigent elderly where the measure of need was measured either the same as for other social assistance programs, in which case this measure becomes similar to integration with overall social assistance. Even if the measurement of need is the same, if the program is administered by a different agency or if the source of funds is budgeted differently, the outcome could be significantly different. 8 Saldain and Lorenzelli (2002), p For example, Iacovu (2000), McGarry and Schoeni (1998), and Costa (1998). 10 See Adamchak (1995) for evidence supporting this hypothesis for Namibia. Conversely, in Brazil, there appears to be a reverse migration from urban to rural areas as young adults move back to take advantage of their parents or grandparents pensions. 11 Exceptions include Germany, Luxembourg, Norway, Canada, and Switzerland, Turner et.al, ILO (2000).

12 12 Increase in coverage in contributory system. Even if coverage could be easily increased in the contributory system, it might not be sufficient to prevent old age poverty. Some people will always be too poor to save or contribute. Others might save but might have more immediate needs for that saving earlier in life due to job loss, illness, or natural disasters. People also might be subject to interrupted careers, either by choice such as women opting to drop out of the labor force to raise children, or by force due to unemployment, which would perhaps make them ineligible for pensions or reduce the value of the pensions they would eventually get. Finally, accurate record-keeping and collection may simply be too difficult in rural areas and with self-employed workers. Even in the U.S. which is noted for its collection and record-keeping, while collection and compliance have been measured to be around 96% for employees, compliance among the self-employed is estimated to be less than 50%, with even less for household employees who are not included in either of the other groups. More importantly, no workable strategy for increasing coverage has been developed. Obstacles to coverage, even when mandated, arise from the formality of employment usually covered under pension systems. Those thus employed often must also purchase health insurance, unemployment insurance, family allowances, and a host of other benefits. They also have to pay income taxes. Finally and perhaps most importantly, they also become covered under the labor code, with potentially costly regulations and expensive hiring and firing costs. The experience of Latin America shows that despite mandating coverage, coverage has in fact not expanded significantly. The single most important factor in determining coverage in contributory systems appears to be per capita income 12. Until the developing world sees substantial rises in income, substantial numbers of elderly may remain uncovered under contributory systems. Integration of Non-Contributory with Contributory. In some countries, a noncontributory system or quasi-contributory system is fully integrated with the contributory pension system. Brazil is a perfect example of such an integration. Brazil had until the 1999 reforms, a pension system which provided a benefit of 100% of the last 3 years salary after 35 years of contributions for men and 30 years for women and a benefit of 70% of the salary base after 30 years of contributions for men and 25 for women or after only 5 years of contributions if the contributor was male and aged 65 or if the contributior was female and aged 60. Following the mandate in the 1988 Constitution to provide pensions to rural elderly, Brazil began to offer two further concessions to rural elderly. Rural workers could get the 70% pensions if aged 60 for men and if aged 55 for women. Even though the years of contributions required to be eligible for this pension are rising from 5 in 1994 to 15 in 2012, the decision was made to count the rural worker s years of service not years of contributions for eligibility determination. Furthermore, while the urban worker and his employer contribute between 28 and 33% to become eligible for this pension, the rural worker pays nothing. Instead, a tax of 2% is assessed on all rural products, and this revenue helps to finance the pension for rural workers. While reforms were made to the benefit formula for the pure contributory pension, the option for the age 12 Palacios and Pallarés-Miralles, p. 24.

13 13 pension with all its rural provisions has remained intact with the 1999 reforms. Furthermore, most rural workers receive the minimum pension which is equal to minimum wage. During their working years, most earned less than half minimum wage, resulting in an enormous subsidy to these workers. While the Brazilian program is notably more generous than most, many other countries in Latin America have similar quasi-contributory provisions where years of contribution required for eligibility go down dramatically as age goes up, from 35 years at age 60 in Uruguay, for example, to only 15 years at age 70. While there is a slight reduction in benefits for those with fewer years of service, in fact the minimum benefit received at age 70 is higher than that at age 60. Similarly, Argentina offers a flat benefit after 30 years of service, but 2/3 of that benefit after only 10 years of service, although the bulk of this service must take place in the years just prior to retirement. What are the advantages of this integration? To the extent that there is a stigma attached to receiving social assistance as compared to a benefit that has been earned by contribution, there would be little stigma attached to those receiving benefits through this type of integrated system. Provision of cash benefits with eligibility conditions also frequently involves economies of scale. Integrating the systems allows the quasicontributory system to take advantage of the institutional structure of the contributory system. Furthermore, unlike pure poverty-related programs which often suffer from reduced political support and funding over time, integrated programs are often politically protected. In the case of Brazil, all rural pensioners automatically get a small percentage of their pension deducted and transferred as dues to a rural workers trade union. The rural workers trade union and the various urban trade unions present a united front and oppose any cuts to the contributory pension system, thus resulting in strong political support for the poverty prevention function and continued funding. What are the disadvantages of this approach? First, it requires either crosssubsidization between the contributory system and the quasi-contributory system or direct subsidies from the government. Despite the government s perhaps best intentions, if the contributory system is running a surplus, rarely does the government actually transfer funds, as can be seen in the cases of Morocco and Costa Rica. Ultimately, the cross-subsidization undermines the fiscal position of the pure contributory system. In the case of contributory systems already in deficit, governments often attribute the deficit to the quasi-contributory part and unnecessarily delay the restructuring of the true contributory part. And even in countries like Brazil, urban unions are beginning to be leery of subsidizing rural workers so heavily. Second, these systems undermine the contributory system by providing workers with the possibility of receiving reasonable pensions without a full contributory record. For example, urban workers in Brazil could evade the high contribution rates in the contributory system, return to the rural areas for a few years prior to retirement, and retire with a reasonable pension. Similarly, in Uruguay, the contributory system would require that workers contribute continually from age 25 to age 60 at a combined employeremployee contribution rate of 27.5% of wage each year in order to collect a pension. The

14 14 same worker could contribute only from age 55 to age 70 and collect a higher pension if the worker was likely to be eligible only for the minimum pension. Since the minimum pension is higher than minimum wage and the contributory pension only pays 50% of the average of the last 10 years salary, a large number of workers have little incentive to contribute for the majority of their working careers. Pay as you go (PAYG) contributory pension systems are based on the concept that contributions from current workers finance pensions for current pensioners. If workers do not contribute full careers as expected and instead choose options which may be cheaper for them, the contributory system becomes unfinanceable. In fact, countries often respond to the loss in revenues by raising contribution rates, which only increases the incentives to choose the cheaper options, resulting in the opposite of the desired effect, a further loss in revenue. Third, while Brazil represents a somewhat extreme case of generosity, most countries require some minimal level of contribution within these contributory systems, such as the 15 years required in Uruguay or the 10 required in Argentina. There will be individuals who will not have satisfied these requirements. Thus, these countries require yet a third system of old age security provision in the economy, one which is purely noncontributory, but means tested. Frequently, as in Uruguay, the pure noncontributory pension and 15 year, age 70 quasi-contributory pension pay the same minimum benefits. Only the individual who expects to get a benefit higher than the minimum or one who could not possibly qualify under the means test given the household finances even after juggling of assets would contribute the first 15 years, further decimating the resources of the contributory system. Universal Pensions. Some countries as noted above, offer universal pensions to all residents above a certain age regardless of income or assets. Even non-citizens who have fulfilled a minimum residency requirement in the country are often eligible. Universal pensions are clearly the best way to drastically reduce old age poverty since all the elderly, regardless of work or contribution experience, are eligible to receive some benefit. However, as with any universal benefit, many people who will be receiving it will not be poor. As a result, in terms of poverty reduction achieved per dollar spent, the expenditure may not be wholly effective. Furthermore, in the countries where such programs have been adopted, such as Namibia 13 and Mauritius, cost considerations are forcing the governments to move toward means-testing and other methods of providing old age security. Table 1 provides some estimates of the cost of providing a universal pension of $1 per day to those above age 60, 65, 70 and 75 in 40 African countries. The numbers range from a low of.01% of GDP in Seychelles to a high of 17% of GDP in Ethiopia. The high numbers are driven by the percentage of elderly in a country, with the older countries costing more, and by the relative generosity of $1 per day when compared to TABLE 1 Country Pen/GDP per capita Angola 2.4% 1.5% 0.9% 0.4% 77% 13 Devereux (2001), p. 25.

15 Benin 4.3% 2.8% 1.7% 0.9% 99% Botswana 0.4% 0.2% 0.1% 0.1% 11% Burkina Faso 7.3% 4.5% 2.4% 1.1% 155% Burundi 14.3% 9.5% 5.9% 3.1% 225% Cameroon 3.6% 2.3% 1.4% 0.7% 60% Cape Verde 2.1% 1.4% 0.8% 0.4% 28% Central African Republic 7.4% 4.7% 2.7% 1.3% 107% Chad 8.5% 6.1% 3.2% 1.9% 170% Comoros 4.2% 2.3% 0.7% 0.7% 84% Congo 1.8% 1.2% 0.7% 0.3% 37% Cote d'ivoire 2.5% 1.5% 0.8% 0.4% 51% Eritrea 10.8% 6.5% 3.4% 1.5% 227% Ethiopia 16.7% 10.6% 6.0% 3.0% 357% Gabon 0.8% 0.5% 0.3% 0.2% 8% Ghana 8.8% 6.0% 4.1% 2.4% 97% Guinea 3.7% 2.3% 1.3% 0.6% 65% Guinea-Bissau 12.5% 7.8% 4.4% 2.0% 156% Kenya 4.3% 2.8% 1.7% 0.8% 109% Lesotho 5.6% 3.6% 2.1% 1.1% 75% Madagascar 6.7% 4.2% 2.4% 1.2% 154% Malawi 9.8% 5.9% 3.1% 1.4% 249% Mali 8.3% 5.4% 3.2% 1.7% 142% Mauritania 5.0% 3.4% 1.9% 0.9% 80% Mauritius 0.9% 0.6% 0.4% 0.2% 10% Mozambique 9.2% 6.3% 3.8% 1.8% 250% Namibia 1.0% 0.7% 0.4% 0.2% 16% Niger 8.2% 5.1% 2.8% 1.3% 178% Nigeria 4.7% 2.9% 1.6% 0.7% 144% Rwanda 5.4% 3.2% 1.7% 0.7% 182% Senegal 3.5% 2.1% 1.1% 0.5% 68% Seychelles 0.1% 0.1% 0.1% 0.1% 5% Sierra Leone 12.2% 7.2% 3.6% 1.6% 190% Sudan 5.0% 3.1% 1.7% 0.8% Swaziland 1.4% 0.8% 0.5% 0.3% 26% Tanzania 5.1% 3.1% 1.7% 0.8% 206% Country Pen/GDP per capita The Gambia 6.0% 3.7% 1.8% 0.9% 106% Togo 6.3% 4.2% 2.6% 1.5% 115% 15

16 16 Uganda 3.9% 2.5% 1.3% 0.6% 122% Zambia 4.5% 2.7% 1.4% 0.7% 94% Zimbabwe 2.5% 1.6% 0.9% 0.5% 56% the per capita GDP in many countries. The last column shows the value of a dollar a day with respect to per capita GDP in each country. As is clearly seen, in many countries one dollar a day far exceeds per capita GDP. Countries have to decide whether in an environment of scarce resources, so much of GDP should be spent on the elderly in comparison with other age groups, particularly since some of this expenditure is not even going toward achieving poverty relief. Alternatively, Table 2 shows the cost of providing a pension that is 40% of GDP per capita now and in the future only to those above the age of 75. Altering the amount of the pension and raising the age of eligibility may be sufficient to make the universal pension option affordable in most countries at least for the medium term. In the long term, as the cost rises due to population aging, the amount of the pension could be held constant in real terms while GDP continues rising, which allows the system to continue reducing absolute poverty though poverty among the elderly relative to other age groups would begin to increase. The issue of whether a nontargeted universal pension is worth the cost would have to depend on a country by country assessment. Universal pensions systems involve less administrative apparatus and administrative costs than targeted systems, but the actual benefits, since they are received by the nonpoor as well as the poor, cost more. In countries like Canada and New Zealand, a progressive income tax system taxes the pensions for higher income individuals, making even the universal pension system relatively targeted. However, most of the lower income countries which might institute a universal pension system may not have a progressive income tax structure and may find it difficult to use the tax system to make the universal pension more targeted. Depending on the income distribution within the country, making the pension small enough may be sufficient for higher income individuals to not bother with collecting the pension. Such is the case in Nepal where higher income Nepalese often do not collect the very low universal benefits. A second set of issues arise regarding the service delivery for a universal pension. In the case of Namibia, despite a mandate for a universal pension, many rural elderly were not receiving the pension. The reasons ranged from lack of access to the pension distribution network and difficulties with verifying age and eligibility to different types of fraud. When a fraud-proof biometric recognition system with cash vehicles which traveled from village to village was put in place, coverage was raised from 50-60% of eligible recipients to 88%. However, the administrative cost of the system was raised TABLE 2 Country Angola 0.3% 0.4% 0.4% Benin 0.4% 0.3% 0.7% Botswana 0.2% 0.3% 0.5%

17 17 Botswana 0.2% 0.3% 0.5% Burkina Faso 0.3% 0.4% 0.5% Burundi 0.3% 0.3% 0.5% Cameroon 0.4% 0.5% 0.9% Cape Verde 0.5% 0.5% 1.3% CAR 0.4% 0.4% 0.6% Chad 0.4% 0.3% 0.4% Comoros 0.3% 0.5% 1.3% Congo 0.3% 0.3% 0.6% Cote d'ivoire 0.3% 0.4% 0.7% Eritrea 0.2% 0.3% 0.7% Ethiopia 0.3% 0.3% 0.5% Gabon 0.7% 0.7% 0.9% Ghana 0.7% 0.6% 1.2% Guinea 0.3% 0.2% 0.6% Guinea-Bissau 0.4% 0.4% 0.6% Kenya 0.3% 0.4% 0.7% Lesotho 0.5% 0.7% 0.9% Madagascar 0.3% 0.4% 0.8% Malawi 0.2% 0.4% 0.4% Mali 0.4% 0.2% 0.3% Mauritania 0.4% 0.4% 0.9% Mauritius 0.9% 1.7% 3.7% Mozambique 0.4% 0.5% 0.6% Namibia 0.4% 0.6% 0.7% Niger 0.2% 0.2% 0.3% Nigeria 0.3% 0.4% 0.6% Rwanda 0.2% 0.3% 0.6% Senegal 0.2% 0.3% 0.7% Seychelles 1.0% 1.1% 3.4% Sierra Leone 0.2% 0.3% 0.6% Sudan 0.3% 0.4% 0.9% Swaziland 0.4% 0.4% 0.8% Tanzania 0.2% 0.4% 0.6%

18 18 Gambia 0.3% 0.4% 0.8% Togo 0.5% 0.4% 0.7% Uganda 0.2% 0.2% 0.4% Zambia 0.2% 0.4% 0.4% Zimbabwe 0.3% 0.4% 0.7% from 7% of the pension s value to 14%. 14 So clearly despite the mandate for universality, achieving true universality can be costly. Since poorer regions are harder to reach, the poorest may still not be reachable with a universal pension without a sizable investment in administrative capacity. This service delivery issue applies to all types of social pensions, not just the universal type. Means-Tested Social Pensions. The last approach to providing social pensions is to provide a means-tested noncontributory pension, with or without a contributory pension. The targeting can use the same mechanisms used in other social assistance programs or can use distinctly different targeting mechanisms. Some issues related to the elderly are somewhat different and may result in different and perhaps easier and cheaper methods of targeting where only the elderly are included. For example, the elderly are more likely to own assets than younger age groups. Means-testing, thus, more often should include assets in the case of the elderly than absolutely necessary for younger age groups. Conversely, since women live longer than women and have participated generally less in formal labor markets, the likelihood of women over the age of 75 in poverty is much higher than the likelihood of poverty for women overall. Thus, the targeting mechanism could be very cheap and almost akin to a universal mechanism, but still result in a fairly cost effective measure to relieve old age poverty. Another issue when means testing is whose income should be included in the means test, particularly when high rates of coresidence prevail. Among OECD countries, typically only the income of the elderly individual and usually the spouse are considered. This has also been the case in South Africa. Typically in developing countries, the household income is considered. However, the incentive under this type of means-testing would be for the elderly to be separated from other household members so that the maximum subsidy could be obtained by the family. A few OECD countries actually require the elderly to attempt getting support from family members prior to becoming eligible for assistance. In the developing country context, countries like Thailand ask that village elders decide who are the deserving elderly, taken into account potential help from children. In France, social assistance for the elderly can actually be reclaimed upon death from any assets bequeathed to their heirs. 15 The advantage of means testing the social pensions directly is of course the cost savings in terms of overall benefits as well as the improved targeting of the fiscal expenditure directly toward poverty relief. However, in the South African case, Case and 14 Devereux (2001), pp Turner, et.al, ILO (2000).

19 19 Deaton did not find much difference between means tested and universal pensions among elderly African, although there would have been significant differences among non- African recipients had the program been universal. This situation may be unique to South Africa. 16 Moreover, the process of means-testing can be complicated and expensive. Given the problems even universal pensions face in service delivery in low income countries, adding the dimension of means testing may make them completely unwieldy from an administrative point of view. In addition, the requirements to comply with the means-testing may be expensive for the individual and thus may result in excluding many of the elderly poor that the program is in fact intended to help. In many cases, as in Costa Rica, the requirements are not considered so onerous that individuals cannot comply with them, but there are other cases where applying becomes too costly particularly where bureaucrats demand bribes in order to process the paperwork, as in India. In developed countries, the elderly frequently do not seem to take advantage of the benefits offered to them, particularly when means-tested. 17 The experience from other countries seems to suggest that when the pension is broadly means-tested such that a large percentage of the population receive it, take-up is not an issue, but when a small percentage of people get it, then take-up is a significant issue. 18 Another issue that authorities need to guard against is the possibility of fraud. If people are able to receive pensions through fraud, it undermines the fairness of the system. Thus, in addition to the costs of means testing, the public agency also needs to expend resources which could have been spent paying benefits in fraud control. Finally, individuals will alter their behavior in order to make themselves eligible for the pension whatever the criteria. The behavioral impact could include evasion of the contributory system, juggling of assets within the family, as well as decisions on labor force participation. The issue of contribution avoidance, rather than evasion, is also relevant when the level of the noncontributory pension is so high that it encourages people to avoid contributions. A good example is that of Uruguay where the minimum contributory pension at age 65 after 35 years of service is below that of the noncontributory pension available at age 70. Why would anyone want to contribute? Consequently, contribution compliance is relatively low in Uruguay. 19 However, the experience of several countries with means-tested programs for the elderly is that the political support for these programs is limited and budget either disappears over time or becomes diverted to other political purposes. Examples include that of Turkey where a means-tested benefit of 300 Turkish lira per month for the elderly has been in place for many years, despite several years of 100% inflation resulting in monthly pensions worth far less than a penny. In another part of the world, in Argentina, a limited budget for noncontributory pensions has increasingly been used for political pensions, so that by 2000, less than half of the noncontributory pensions were based on poverty and the value of the pensions for poverty relief was below that for the political 16 Case and Deaton (1998), p See for example, McGarry (1995). 18 See ILO for OECD examples where take-up is not an issue in countries like Australia; similarly take-up does not appear to be an issue in South Africa, Case and Deaton (1998), p Saldain and Lorenzelli (2002), p. 246.

20 20 pensions, resulting in well more than half the budget allocated for other purposes. In 1989, 59% of noncontributory pensioners were poor elderly or disabled. By 2000, only 32% of the noncontributory pensioners fell within this category. As a result, the program results in a reduction in poverty of only 0.5%. 20 Policy Recommendations Having reviewed the options, for most low income countries, the optimal strategy may be to begin with a universal pension, offered at an advanced age, but providing a minimal level of benefits. The size of the benefit and the age at which it is offered clearly have to take into account competing needs for the fiscal resources, such as education, health, other social protection, and infrastructure building. A risk and vulnerability assessment which would highlight whether the elderly are in fact a group which should be especially targeted is essential before beginning the process. However, a small benefit may be affordable and will result in substantial poverty relief among the most elderly, who might be among the most vulnerable in the society. Efforts can be put in place right away to take advantage of any progressivity in the tax code to return the benefits of the less needy elderly to the fiscal treasury, which further reduces the cost. Over time, however, it needs to be realized that this benefit will probably become more expensive, perhaps overly so. At that point, several options could be available. The real value of the benefit could be held constant, reducing the amount and cost over time, making it attractive only to the most poor. The age at which it becomes available could be continually raised, perhaps with life expectancy. The benefit could be made indirectly more progressive by more aggressive tax measures. Finally, the benefit could be made explicitly more progressive by explicit means testing. The law providing the benefit thus needs to be written in such a way that the benefit structure can later be changed without undue legislative hassle. 20 Bertranou and Grushka (2002).

21 21 Selected References Adamchak, D.J. Pensions and Household Structure of Older Persons in Namibia, Southern African Journal of Gerontology, 4 (1995), pp Ardington, Elisabeth and Lund, Frances. "Pensions and Development: Social Security as complementary to programmes of reconstruction and development," Development Southern Africa 12 (August 1995), pp Bertranou, Fabio and Grushka, Carlos O., The Non-Contributory Pension Programme in Argentina: Assessing the Impact on Poverty Reduction, ESS, Paper No. 5, International Labour Office, Burman, Sandra. Intergenerational Family Care: Legacy of the Past, Implications for the Future, Journal of Southern African Studies, 22 (1996), pp Case, Anne and Deaton, Angus. Large Cash Transfers to the Elderly in South Africa, The Economic Journal, 108 (1998), pp Charlton, Roger and McKinnon, Roddy. Beyond Mandatory Privatization: Pensions Policy Options for Developing Countries. Journal of International Development 12 (2000), pp Costa, Dora L. Displacing the Family: Union Army Pensioners and Elderly Living Arrangements, NBER Working Paper Series, No. 5429, January Costa, Dora L. A House of Her Own: Old Age Assistance and the Living Arrangements of Older Nonmarried Women. NBER Working Paper Series, No. 6217, October Costa, Dora L. The Evolution of Retirement: An American Economic History, Chicago: University of Chicago Press, Cox, Donald and Jakubson, George. The Connection between Public Transfers and Private Interfamily Transfers, Journal of Public Economics 57 (1995), Daly, Mary C. and Burkhauser, Richard V. The Supplemental Security Income Program. September Deaton, Angus and Paxson, Christina. Measuring Poverty Among the Elderly. NBER Working Paper, no. 5296, October Devereux, Stephen. Social Pensions in Namibia and South Africa. IDS Discussion Paper 379. Institute of Development Studies, University of Sussex, 2001.