Social Assistance Programs and Household Welfare in Eswatini STUDY BRIEF Picture: World Food Programme, Eswatini This brief summarizes findings and recommendations from a study that assesses the performance of Eswatini’s largest social assistance programs, based on recent national household survey data and program administrative information. The study examines the population coverage rates of these programs, as well as their incidence and effectiveness in reducing poverty and inequality. The study also examines the association between program participation and negative shocks reported by households, in particular drought and food price shocks associated with the 2015/16 El Niño event. STUDY BRIEF 1 Eswatini B AC KG R O U N D Eswatini has notably high levels of poverty and As part of its efforts to address poverty and Overall inequality compared to other lower-middle-income vulnerability and its adverse consequences on poverty rate countries, as well as developing countries more households, the Eswatini government administers a mix generally. In 2016/17, the overall poverty rate in of social insurance programs, social assistance programs Eswatini was 58.9 percent, based on the country’s upper and social welfare services. The HIV/AIDS epidemic 58.9% poverty line (or 28.3 percent, based on the international and its negative effects on household welfare is the poverty line of $1.90 per day per person in 2011 raison d’être for several social assistance programs and purchasing power parity U.S. dollars). The country’s Gini social welfare services. Measured in 2016/17, Eswatini index, a measure of inequality, was 49.3 percent. had the highest HIV prevalence in the world, with 27 percent of those age 15 and older living with HIV. Compared to other countries worldwide, Eswatini is assessed to be at a medium risk of humanitarian crises Primary administrative responsibility for many of the Gini Index and natural or human-caused disasters, which reflect main government social assistance programs lies with the relatively high socioeconomic vulnerability of its the Department of Social Welfare (DSW), under the population, and its weak institutional and infrastructural Deputy Prime Minister’s Office (DPMO). International 49.3% capacity to respond to negative shocks. These dynamics donors have supported some of the government’s social further challenge government plans and actions aimed assistance programs and social welfare services through at tackling structural poverty and inequality. official development aid, and have also supplemented these programs and services with direct humanitarian aid. 2 SOCIAL ASSISTANCE PROGRAMS AND HOUSEHOLD WELFARE IN ESWATINI Aims, scope, and approach This study brief on “Social Assistance Programs and survey. Of particular relevance for the analysis of shocks Household Welfare in Eswatini”1 examines the perfor- and social assistance programs is that the survey data mance of Eswatini’s main social assistance programs. span the El Niño-induced drought of 2015/16, which These programs employ varying combinations of had large and widespread negative effects on crop and categorical and self-targeting to reach the poor and livestock production and food security. vulnerable. The study examines the five social assistance programs Specifically, the study asks two primary questions. with the largest numbers of beneficiaries, as reported in The first question is: How successful are the main the SHIES 2016/17: social assistance programs in reducing poverty and inequality?2 This question is answered through standard 1. neighborhood care points (NCPs), which are benefit incidence analysis. To draw further insights for community-run centers that provide free informing public policy, the study also asks an ancillary cooked meals (and basic early childhood care counterfactual question, namely: Could the social and development services) to needy, young assistance programs achieve greater gains in reducing orphaned and vulnerable children (OVC); poverty and inequality if benefits were targeted to 2. school feeding program, through which the poor households through a proxy means test (PMT) government subsidizes cooked lunches for model which the study constructs for this hypothetical students in government primary, secondary and exercise?3 The question is answered by conducting high schools; microsimulations. 3. OVC education grants, through which the The second question is: Do the main social assistance government subsidizes tuition and exam fees for programs tend to cover households that experience OVC attending government secondary and high adverse shocks? This coverage may be an ex-ante schools; result obtained through program eligibility criteria 4. emergency food aid, through which the government not directly linked to shock exposure, or it may be an distributes food commodities to households ex-post result if program eligibility criteria, timing or facing acute food insecurity; and level of program benefits are adjusted in response to the 5. elderly grants, through which the government shock. The question is answered by examining the types provides regular cash benefits to those aged 60 of adverse shocks reported by households, as well as the and above. association between commonly reported shocks and poverty status, food insecurity status, area of residence, Neighborhood care points are fully financed through receipt of social assistance program benefits, and the international aid, with the Eswatini government types of coping responses reported by households. involved in a coordinating capacity. Emergency food aid is financed by the government, and is often The analysis is primarily based on data from the supplemented by international aid. The other three Swaziland Household Income and Expenditure programs are fully financed by the government. While Survey (SHIES) 2016/17, which is the latest available there are other government social assistance programs, national household sample survey with relevant data. most of them fall under the responsibility of the The SHIES contains dedicated modules on social Department of Social Welfare, they each cover less than protection interventions and shocks. Measures of 1 percent of the national population. poverty and inequality are based on the survey’s household consumption data. The Eswatini government The study examines the performance of each individual administers and uses the SHIES for its official estimates program, as well as the collective performance of all of poverty and inequality in the country. five programs and of the four government-financed programs, excluding NCPs. The exclusion of NCPs The reference period for the SHIES 2016/17 data on makes little difference to the overall results as the social protection interventions is the year preceding the outlay and coverage of NCPs are small. 1 Raju, Dhushyanth, and Stephen D. Younger. 2021. "Social Assistance Programs and Household Welfare in Eswatini." Social Protection and Jobs Discussion Paper 2106, World Bank, Washington, DC. 2 Observational data are used in this study. Given this, potential selection bias would impair estimates of the role of social assistance programs in influencing outcomes, such as education, health, nutrition and labor, among others. Other appropriate data designs and empirical strategies are needed for such an investigation. 3 A PMT model uses a few easily identifiable characteristics of people or households to proxy household consumption and hence poverty status. STUDY BRIEF 3 FIGURE 1 Government Spending on Social Assistance Program Benefits, 2014/15–2017/18 a. Absolute spending across programs 800 600 Emalangeni (in millions) 340.4 185.0 400 168.4 165.0 137.3 161.2 141.6 140.6 200 135.5 126.8 126.9 138.5 47.9 22.8 35.4 0 FY2014/15 FY2015/16 FY2016/17 FY2017/18 b. Relative spending across programs 100 36.7 35.7 35.2 80 49.6 60 Percent 26.5 30.6 30.2 40 23.5 26.2 20 27.7 29.5 18.5 9.2 5.2 0 FY2014/15 FY2015/16 FY2016/17 FY2017/18 Other grants and benefits Food aid School feeding OVC education grants Elderly grants Source: Official budget books of the Government of the Kingdom of Eswatini; Eswatini Ministry of Education and Training. Note: OVC = orphaned and vulnerable children. Other grants and benefits include disability grants, military pensions, benefits for foster children, and transfers to CARITAS Swaziland and Baphalali Eswatini Red Cross Society. 4 SOCIAL ASSISTANCE PROGRAMS AND HOUSEHOLD WELFARE IN ESWATINI Main findings The Eswatini government spent between E458 In recent years, annual government spending on social million and E687 million annually on social assistance assistance programs has accounted for 2.5 to 3.5 programs between 2014/15 and 2017/18 (figure 1a). percent of total annual government spending, and has Over the reference period of the SHIES 2016/17, the been roughly equivalent to 1 percent of gross domestic government spent mostly on elderly grants, followed product (GDP) (figure 2). This level of government by OVC education grants and school feeding (figure spending on social assistance programs as a share of 1b). Elderly grants accounted for about one-third of GDP is lower than the corresponding average levels in total government social assistance program spending. Sub-Saharan Africa and developing countries in general. Its position as the program with the largest outlay has In each of these groups of countries, annual spending become more pronounced following large increases in on social assistance programs averages about 1.6 grant benefit values in 2017 and 2020. OVC education percent of GDP. grants and school feeding each accounted for about one- quarter of total government social assistance program The level of spending on social assistance programs spending. Government spending on emergency food by Eswatini is also lower than those of its neighbors, aid tended to be small (less than 5 percent of total specifically other member countries of the Southern government spending on social assistance programs). African Customs Union (SACU), where, for example, Likewise, government spending on other social recent annual spending on social assistance programs assistance programs under the Department of Social relative to GDP has been more than 3 percent in Welfare was also small (less than 5 percent of total Lesotho and South Africa. government spending on social assistance programs). Coverage by the five main social assistance programs is reasonably extensive (figure 3). An estimated 41.3 percent of the population is covered by at least one FIGURE 2 Annual Government Spending on Social Assistance program. School feeding and elderly grants cover large Program Benefits Relative to Annual Total Government shares of their presumed eligible populations (79.3 Spending and GDP, 2014/15–2017/18 percent of government primary, secondary and high 4 school students, and 81.1 percent of those aged 60 3.5 years or older, respectively), but smaller shares of the 3.0 general population (22.8 and 5.8 percent, respectively). 2.9 3 Emergency food aid provided by government covers 2.5 16.3 percent of the population. NCPs and OVC education grants have the lowest coverage rates, at 4.0 2 percent each. 1.2 All of the programs are pro-poor and pro-food insecure. 1.0 1.0 1 0.9 Collectively, the five programs cover 54.6 percent of the extreme poor and 50.8 percent of the moderate poor.4 In contrast, 25.9 percent of the nonpoor are covered. 0 Likewise, taking all five programs together, 48.7 FY2014/15 FY2015/16 FY2016/17 FY2017/18 percent of those that are severely food insecure are Percentage of total government spending Percentage of GDP covered, while 25.7 percent of the those that are food Source: Official budget books of the Government of the Kingdom of Eswatini; Ministry of Education secure are covered.5 and Training; Ministry of Economic Planning and Development. 4 The extreme poor are those with consumption below the country’s lower poverty line of E463.4 per month per adult equivalent, in January 2017 prices. The moderate poor are those with consumption above the lower poverty line and below the country’s upper poverty line of E975.3 per month per adult equivalent, in January 2017 prices. Nonpoor are those with consumption above the upper poverty line. 5 Food insecurity is measured by a categorical variable (severely food insecure, moderately food insecure, mildly food insecure, and food secure) constructed based on the Household Food Insecurity Access Scale (HFIAS) and Household Food Insecurity Access Prevalence (HFIAP). STUDY BRIEF 5 FIGURE 3 Program Coverage Rates The pro-poor nature of the programs is further reflected by program concentration coefficients, which a. Individuals vary between –0.255 and –0.152 across the various Percent programs (figure 4).6 These values indicate that the 0 20 40 60 80 100 programs are progressive. However, these values are less progressive than those of many social assistance NCPs 4.0 programs in other developing countries. For a point of reference, the average concentration coefficient value across direct transfers of any type in 32 developing School countries is –0.30, and the lowest value is –0.63.7 22.8 feeding Despite the pro-poor nature of the programs, their Food aid 16.3 marginal effects on poverty and inequality are small: None of them reduce either the Gini index, or the poverty rate, by more than one percentage point.8 OVC education 4.0 Collectively, the five programs reduce the Gini index by grants 2.0 percentage points (compared to a baseline value of 49.3 percent), the overall poverty rate by 1.5 percentage Elderly grants 5.8 points (compared to a baseline value of 58.9 percent), and the overall poverty gap by 3.0 percentage points All programs (compared to a baseline value of 24.9 percent).9 (except NCPs) 37.4 While small in absolute terms, the marginal effects compare favorably to the recent changes in inequality or All programs 41.3 poverty in Eswatini. The collective marginal effect of the five programs on inequality is roughly equivalent to the decline in inequality in Eswatini between 2000/01 and 2016/17. Likewise, the collective marginal effects of the five programs on the overall poverty rate and poverty b. Presumed eligible individuals gap are equivalent to 17 and 54 percent of the decline in the overall poverty rate and poverty gap, respectively, Percent between 2009/10 and 2016/17. 0 20 40 60 80 100 Each one of the five programs is quite effective at reducing the poverty gap, and between 73 and 81 NCPs Orphans, younger 12.4 than age 8 percent as effective as transfers of equivalent outlays that are perfectly targeted to the poor would be. But Govt. primary effectiveness at reducing poverty severity is much less 79.3 for each program: They are between 48 and 54 percent school students as effective as transfers of equivalent outlays that are School feeding perfectly targeted to the poor would be.10 This indicates Govt. secondary 60.8 that while the programs perform well at reaching the school students poor, they perform much worse in reaching the poorest of the poor. The programs are also less effective in Govt. primary and reducing inequality than reducing the poverty gap. secondary school 73.4 students Food aid Severely food 23.1 insecure 6 The concentration coefficient is a measure of program incidence. A education Orphans in govt. 54.7 grants coefficient value closer to minus one indicates that the benefits from OVC secondary a program are more concentrated among poorer households, while a schools value closer to one indicates that they are more concentrated among richer households. 7 See the Commitment to Equity (CEQ) Institute database at http:// Elderly 81.1 grants Age 60 or older commitmentoequity.org/datacenter. 8 The marginal effect is a measure of program incidence. It provides the change in the estimate of poverty or inequality when a program benefit is added or removed from consumption. 9 Overall poverty is measured with respect to the country’s upper Source: Estimates based on data from the Swaziland Household Income and Expenditure Survey poverty line. The poverty gap measures the average shortfall in 2016/17. consumption of the population from the poverty line, and reflects Note: NCP = neighborhood care point; OVC = orphaned and vulnerable children; govt. = the depth of poverty. government. Orphans comprise those who have lost one or both parents to any cause of death. 10 Poverty severity is the square of the poverty gap. The measure Government secondary schools comprise both government secondary and high schools. assigns greater weight to the poorest. 6 SOCIAL ASSISTANCE PROGRAMS AND HOUSEHOLD WELFARE IN ESWATINI FIGURE 4 Concentration Coefficients Given sizeable, recurrent fiscal deficits in the recent past, the Eswatini government is constrained in Concentration coe cient its ability to increase its spending in general. The -0.3 -0.2 -0.1 0 government aims to undertake a fiscal consolidation, although the COVID-19 pandemic has derailed this plan. Given the expected tight fiscal position over -0.255 the medium term, the study conducts simulations to NCPs examine whether the performance of social assistance OVC programs can be improved by better targeting them -0.242 education grants to the poor, and by reallocating savings from the targeting toward larger program benefits for existing -0.235 Food aid beneficiaries. The study performs this analysis for OVC education grants and elderly grants, constructing a PMT model for Eswatini based on the SHIES 2016/17 -0.195 School feeding and splitting the population into “PMT-poor” and “PMT-nonpoor.” -0.193 All programs Targeting these programs to the PMT-poor produces All programs significant savings in spending (totaling about E30 -0.191 (except NCPs) million for OVC education grants and about E60 million for elderly grants), but large proportions (30 percent or -0.152 Elderly grants more) of beneficiaries would lose their benefits. Simply eliminating benefits for the PMT-nonpoor would lead to slight increases in poverty and inequality because of Source: Estimates based on data from the Swaziland Household Income and Expenditure errors of exclusion inherent in PMT-based targeting. Survey 2016/17. However, reallocating savings from targeting towards Note: NCP = neighborhood care point; OVC = orphaned and vulnerable children. higher benefits for PMT-poor beneficiaries would contribute to further meaningful declines in poverty and inequality (figure 5). FIGURE 5 PMT-based Targeting of OVC Education Grants and With respect to shocks, on average, households report a Elderly Grants, Reform Simulation Results little less than one negative shock in the year preceding Percentage point the SHIES 2016/17, and a little more than two in the -1.5 -1.0 -0.5 0.0 0.5 five years preceding the survey. “Drought or floods” and “large rise in price of food” are the dominant shocks, 0.1 each representing one-third of the shocks reported in Benefits targeted to 0.2 the year preceding the survey. These large shares are the PMT-poor 0.1 presumably driven by the El Niño-induced drought of 0.1 2015/16. Change in Gini index Benefits targeted to -0.1 the PMT-poor and 0.2 Change in program savings -0.1 overall reallocated toward -0.1 poverty rate higher benefits -0.2 Change in overall 0.1 poverty gap 0.4 Benefits targeted to Change in 0.3 the PMT-poor overall 0.2 poverty rate 0.1 Change in -0.3 extreme Benefits targeted to poverty rate 0.1 the PMT-poor and -0.3 program savings -1.0 reallocated toward higher benefits -0.4 Source: Estimates based on data from the Swaziland Household Income and Expenditure Survey 2016/17. Note: PMT-poor = those below the 59th percentile of the proxy means test (PMT) score distribution, in line with the percentile cutoff for the overall poverty rate (58.9 percent). STUDY BRIEF 7 The vast majority of households report receiving no Nevertheless, reporting a drought shock is positively assistance from others (figure 6). Only 4 percent report associated with the receipt of benefits from each of the receiving assistance from the Eswatini government. In five programs. The association between these shocks light of this, it appears that the government and other and emergency food aid is presumably a result of an institutions are not responding to shocks in ways that ex-post response, while the associations for the other make their interventions the main coping strategies for programs are presumably an incidental result of ex-ante an appreciable share of affected households. coverage. Given the data used in the study, the extent to which the receipt of benefits from the social assistance programs moderate the negative welfare effects of shocks cannot be credibly estimated. FIGURE 6 Distribution of Assistance Received, Conditional on While our analysis is based on standard benefit Reporting a Given Shock Type incidence and microsimulation methods, these methods have well-known analytical limitations. We estimate Percent average benefit incidence and not benefit incidence at the margin. The latter may be more relevant when 0 20 40 60 80 100 simulating the effects of reforms that adjust the extent of program coverage. In performing our simulations of reforms to social assistance programs, we generally ignore potential behavioral responses in household Drought 87.2 7.7 labor supply, income generation, or consumption. We also ignore potential general equilibrium, lifecycle, or intergenerational effects, as well as political economy responses in our simulations. Further, in terms of survey data quality, the Food price correspondence between SHIES 2016/17 estimates and 91.6 6.8 Shock type administrative information on beneficiary numbers and outlay is reasonably good for all programs, with the exception for cases of emergency food aid. The shortfalls in beneficiaries and outlay for this program in the survey data are substantial when compared to administrative information. Consequently, estimates of coverage, Other 70.7 23.8 incidence and effectiveness for emergency food aid are likely to have a large downward bias. However, given that the program is small, these underestimates would have little bearing on the overall results. None NGO or religious organization Family or friend Government Source: Estimates based on data from the Swaziland Household Income and Expenditure Survey 2016/17. Note: Information on shocks refers to the 12 months before the survey. “Other” shock includes all shocks except “drought” and “food price” shocks. Pictures: World Food Programme, Eswatini 8 SOCIAL ASSISTANCE PROGRAMS AND HOUSEHOLD WELFARE IN ESWATINI Directions for future research While in line with the state of the art, our analysis of beyond household consumption and food security (for program performance based on the SHIES 2016/17 is example, individual education, health, and nutrition correlational and partial. Critical knowledge gaps remain. outcomes); (ii) the behaviors and outcomes of specific To address them, future research should include rigorous subpopulations (for example, the poor or vulnerable, evaluations to assess the potential impacts of social and young or adolescent children); and (iii) behaviors assistance programs on welfare, including the programs’ and outcomes in the near term as well as over a potential welfare-protective impacts in the face of shocks. longer time frame. In light of this approach, future Causal impacts can be uncovered through creative evaluations should seek to examine program impacts in empirical strategies applied to available observational data. a comprehensive manner. However, they will likely require the collection of new data following appropriate research designs for causal inference. Future research should also include program operational reviews to assess shortcomings in design parameters and Eswatini's social assistance programs are explicitly or design-implementation gaps, covering the foundational implicitly designed to influence: (i) specific outcomes elements and links in the entire program delivery chain. Directions for public policy The findings from the program performance analysis, its public spending on social assistance programs is coupled with the information on program design and low. The government’s fiscal policy is considered to implementation discussed in the study, suggest that have a limited bearing on the country’s poverty and program intensity (rather than scale or reach) is a inequality challenges. Of course, the government’s primary issue. This issue impairs the performance of overall spending is not solely aimed at reducing current programs in addressing inequality, poverty and the consumption poverty and inequality in the most negative effects of shocks. In some cases, actual benefit direct way possible (such as through cash or in-kind levels are lower than intended benefit levels. For example, transfers to poorer households); the government the amount of food procured by the government for spends to achieve a host of other goals. As a first level the school feeding program is inadequate. Similarly, of evaluation, the government needs to carefully weigh shortfalls in food procured and distributed to NCPs the objectives of different lines of spending, and any have a disruptive impact on their operations. potential trade-offs that would arise from adjusting the composition of overall spending. As a second level of In other cases, intended benefit levels are inconsistent evaluation, the government should carefully assess the with the aims of the program. For example, the OVC effectiveness and efficiency of a given line of spending education grant amount provided by the government against the goal that the spending seeks to achieve. does not fully defray the tuition costs of the beneficiary. Likewise, the assigned portion of the per-student With respect to both these levels of evaluation, grant under the government’s Free Primary Education there appears to be reason for concern. The Eswatini initiative does not fully defray the operational costs of government allocates relatively high levels of its budget school feeding in government primary schools. to compensation for public employees and transfers to extrabudgetary entities and public enterprises, Public policy should seek to correct these disconnects. and suffers from weaknesses in its public financial If additional financing is not forthcoming for these management and procurement systems. Effectively programs, correcting these disconnects may require addressing these concerns is critical for improving the (further) efforts to ensure that they are targeted to the government’s overall fiscal position and creating the neediest. Greater program intensity may need to be fiscal space for vital social and core capital spending. traded off against reduced program coverage. The administrative system of the government’s social The option of increased financing of social assistance assistance programs requires a major upgrade. The programs nonetheless merits serious consideration. needed improvements span the entire delivery chain and Compared to many other countries, Eswatini’s levels relate to communication and outreach, targeting and of poverty and inequality are much higher, while enrollment, benefit transfer processes, case and grievance STUDY BRIEF 9 management, and monitoring. A top priority should be the One axis for rethinking the combination of social development and implementation of a single integrated assistance programs is the balance between cash social registry and an integrated management information and noncash transfers. Presently, among the five main system for implementing and monitoring these programs. social assistance programs examined in the study, four offer food or other in-kind benefits. Only the The government should also reconsider the mix of social elderly grants program offers cash benefits. Another assistance programs. They should seek to ensure that axis for rethinking the combination of programs is the aims and designs of their programs are consistent the balance between nonlabor income transfers and with an up-to-date profile of the main welfare risks support for labor income. Programs such as productive (including adverse shocks) that are being experienced inclusion and labor-intensive public works fall under by the poor and other disadvantaged sociodemographic the latter type. groups. A lifecycle framework for identifying the welfare risks faced by individuals is one approach that the In view of the current state of Eswatini’s social assistance government could adopt to ascertain the combination system, including the constraints faced by the system of social assistance programs. from various sources (political, financial, administrative), a gradual stage-by-stage approach toward advancing In determining the combination of social assistance the development of the social assistance system may be programs, the government should also account for the most practical. The first stage would be to address the whole landscape, current and expected, of social benefit-related disconnects in the existing social welfare services, social insurance programs, labor assistance programs. The second stage, staggered with market programs, and other development interventions the first, would be to strengthen the social assistance in Eswatini. In so doing, the government should aim administrative system. The third stage would be to to address gaps or weaknesses (i.e., uncovered or rethink the mix of social assistance programs. An weakly covered adverse conditions or risks) through its integral part of advancing the development of the social social assistance programs and to forge stronger links assistance system would be to make it more resilient and and enhance synergies between its social assistance more responsive to adverse shocks.11 programs and other programs and services. 11 Productive inclusion programs are also referred to as economic inclusion or graduation programs. Labor-intensive public works programs are also referred to as workfare. Picture: Adobe Stock © silvye 10 SOCIAL ASSISTANCE PROGRAMS AND HOUSEHOLD WELFARE IN ESWATINI Epilogue: The COVID-19 crisis The COVID-19 pandemic presents a substantial setback to the government’s efforts to address structural poverty and inequality through its social assistance programs and other development interventions. The International Monetary Fund estimated in April 2021 that Eswatini’s real GDP declined by 3.3 percent.12 To address the economic fallout from the pandemic, the government approved a supplementary budget, and provided food and cash assistance to needy and affected households, tax refunds to enterprises, and cash assistance to laid-off or unpaid employees of enterprises, among other actions. Eswatini's fiscal deficit increased in 2020/21, as government revenues from taxes and SACU receipts declined, and expenditures increased. The economic recession is thought to have increased the extent and depth of poverty in the country in 2020. Such fallout from the COVID-19 pandemic underscores the need for the Eswatini government to have a robust, flexible, and effective social protection system. The recommendations made in this study for future data and research initiatives as well as for public policy are consistent with this imperative. Indeed, the country’s social assistance and social security policies and associated implementation plans, when finalized and approved by the Eswatini government, can serve as a basis for building such a system. 12 Information obtained from the IMF’s April 2021 Regional Economic Outlook for Sub-Saharan Africa. STUDY BRIEF 11 Picture: World Food Programme, Eswatini