SOCIAL PROTECTION & JOBS DISCUSSION PAPER No. 2303 | APRIL 2023 Social Protection Program Spending and Household Welfare in Ghana Dhushyanth Raju, Stephen D. Younger and Christabel Dadzie © 2023 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: +1 (202) 473 1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. RIGHTS AND PERMISSIONS The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: +1 (202) 522 2625; e-mail: pubrights@worldbank.org. Abstract retro geometric background: © iStock.com/marigold_88 Project 41595 Social Protection Program Spending and Household Welfare in Ghana Dhushyanth Raju Stephen D. Younger Christabel Dadzie Abstract: Ghana administers multiple social protection programs. One, pensions provided by the Social Security and National Insurance Trust, has a long history, but the rest of the programs have been introduced and expanded over the past two decades. This study assesses the performance of the government of Ghana’s main social assistance and social insurance programs. It discusses the main design and implementation parameters of the programs and summarizes existing evaluative and operational research. The study also examines patterns and trends in program benefit spending, based on government administrative data, and the coverage rates of the programs, their incidence, and their effectiveness in reducing poverty and inequality, based on recent national household sample survey data. Further, the study examines the relationship between household participation in social assistance programs and exposure to adverse covariate shocks, specifically, possible weather-related shocks, based on high-resolution climate risk maps for the country. JEL codes: H53, H55, I3, O12, O55 Key words: public finance, coverage, incidence, effectiveness, social protection, social assistance, safety nets, social insurance, shocks, Ghana, Africa Table of Contents ACKNOWLEDGMENTS........................................................................................................vi SUMMARY ........................................................................................................................... 1 1. INTRODUCTION............................................................................................................ 2 2. BACKGROUND ........................................................................................................... 12 2.1. Population ............................................................................................................. 12 2.2. Economic growth .................................................................................................. 14 2.3. Consumption and income ..................................................................................... 14 2.4. Poverty and food insecurity ................................................................................... 16 2.5. Inequality .............................................................................................................. 24 2.6. Shocks .................................................................................................................. 26 2.7. Economic growth and poverty movements spanning the coronavirus pandemic ... 26 2.8. Fiscal position ....................................................................................................... 27 2.9. Social protection ................................................................................................... 28 3. PROGRAMS ................................................................................................................ 31 3.1. Livelihood Empowerment Against Poverty Program .............................................. 31 3.2. Labor-Intensive Public Works Program ................................................................. 41 3.3. Ghana School Feeding Programme ...................................................................... 48 3.4. Social Security and National Insurance Trust Pension Scheme ............................ 52 3.5. National Health Insurance Scheme ....................................................................... 58 4. PERFORMANCE ANALYSIS........................................................................................ 66 4.1. Spending .............................................................................................................. 66 4.2. Coverage .............................................................................................................. 77 4.3. Incidence .............................................................................................................. 85 4.4. Effectiveness ........................................................................................................ 94 4.5. Simulated reforms to the LEAP program ............................................................... 96 4.6. Shocks and social assistance program participation ........................................... 101 5. CONCLUSION ........................................................................................................... 105 REFERENCES ................................................................................................................. 109 APPENDIX A: DATA AND VARIABLE CONSTRUCTION .................................................. 114 Welfare, poverty, and inequality ..................................................................................... 114 Social protection programs ............................................................................................ 115 APPENDIX B: DESCRIPTION OF MEASURES................................................................ 121 Program coverage ......................................................................................................... 121 Inequality measure: Gini coefficient ............................................................................... 121 Poverty measures: poverty rate and poverty gap ........................................................... 123 i Program benefit incidence ............................................................................................. 124 Concentration curve and coefficient ........................................................................... 124 Marginal effect............................................................................................................ 126 Program effectiveness ................................................................................................... 126 APPENDIX C: SUPPLEMENTAL FIGURES AND TABLES ............................................... 128 APPENDIX D: ESTIMATING HOUSEHOLD CONSUMPTION IN THE ABSENCE OF SSNIT PENSIONS ....................................................................................................................... 135 APPENDIX E: SIMULATING LEAP+ HOUSEHOLDS ....................................................... 137 ii Figures Figure 1. Population Distribution, 2021 ............................................................................... 13 Figure 2. National Income Trends ....................................................................................... 15 Figure 3. Poverty Rate Trends, Based on International Poverty Line ($1.90-Per-Day Line) . 17 Figure 4. Extreme and Overall Poverty Rates, by Region, 2016/17 ..................................... 18 Figure 5. Distribution of Poverty across Districts, 2010 ....................................................... 19 Figure 6. Poverty Rates, by Selected Subgroups, 2016/17 ................................................. 20 Figure 7. Poverty Rate Trends, National, by Area and by Region, 2005/06–2016/17 .......... 22 Figure 8. Growth and Redistribution Decomposition of the Change in Poverty Rate between 2012/13 and 2016/17 .......................................................................................................... 23 Figure 9. Food Insecurity Rates, 2020 ................................................................................ 24 Figure 10. Inequality, National, by Area, and by Region, 2016/17........................................ 25 Figure 11. Inequality Trends, National, by Area, and by Region, 2005/06–2016/17 ............. 25 Figure 12. Fiscal Position .................................................................................................... 28 Figure 13. LEAP Program Household Numbers .................................................................. 33 Figure 14. Distribution of LEAP Program Households, by District, 2016–21 ........................ 34 Figure 15. Nominal and Real Changes in Stipulated LEAP Program Benefit Levels............ 37 Figure 16. Periodicity of LEAP Payment Rounds and Number of Monthly Benefits per Payment Round, 2016–22 ................................................................................................................. 38 Figure 17. LIPW Program Beneficiary and Subproject Numbers ......................................... 42 Figure 18. Distribution of LIPW Program Individuals, by District, 2016–21 .......................... 43 Figure 19. Stipulated Minimum Wage and Average LIPW Program Wage Payment Levels. 46 Figure 20. Civil and Environmental Works Results from LIPW Program Subprojects .......... 48 Figure 21. Distribution of GSFP Beneficiaries, by Region, 2016/17–2018/19 ...................... 50 Figure 22. Active Contributor Levels, SSNIT ....................................................................... 53 Figure 23. SSNIT Beneficiary (Pensioner) Numbers and Average Benefit Levels................ 56 Figure 24. SSNIT Portfolio Performance, Actual Real Return .............................................. 57 Figure 25. NHIS Beneficiary Levels..................................................................................... 59 Figure 26. Nominal and Real Average NHIS Benefits.......................................................... 60 Figure 27. Reasons for Never or No Longer Being an NHIS Member ................................. 63 Figure 28. Spending on the LEAP Program ........................................................................ 68 Figure 29. Spending on GSFP ............................................................................................ 69 Figure 30. Spending on the LIPW Program ......................................................................... 70 Figure 31. Spending on SSNIT Pensions ............................................................................ 72 Figure 32. Spending on NHIS ............................................................................................. 73 Figure 33. Spending Across Social Assistance Programs ................................................... 75 iii Figure 34. Social Assistance Program Spending: Ghana in International Comparison ........ 76 Figure 35. Social Protection Spending, Social Assistance versus Social Insurance ............ 77 Figure 36. Coverage Rates of Individuals, by Program ....................................................... 79 Figure 37. Coverage Rates of Individuals in Target Categories for GSFP and SSNIT ......... 80 Figure 38. NHIS Coverage Rates of Individuals, by Region ................................................ 81 Figure 39. SSNIT Pensions Coverage Rates, by Region .................................................... 82 Figure 40. Coverage Rates of Households, by Program ..................................................... 83 Figure 41. Individual- and Household-Level Coverage Overlaps across Programs ............. 84 Figure 42. Relative Benefit Levels, Program Households, by Program ............................... 86 Figure 43. Distribution of Program Beneficiaries, by Poverty Status .................................... 88 Figure 44. Distribution of Program Benefits, by Poverty Status ........................................... 89 Figure 45. Concentration Coefficients ................................................................................. 91 Figure 46. Poverty Effects of Simulated Percent Increases in LEAP Program Benefits, LEAP+ Households ....................................................................................................................... 100 Figure 47. District Drought and Flood Risks, by Region, 2010 .......................................... 102 Figure 48. District-Level Pairwise Correlations between Flood and Drought Risks and LEAP Program Participation Measures ....................................................................................... 103 Figure 49. District-Level Pairwise Correlations between Flood and Drought Risks and LIPW Program Participation Measures ....................................................................................... 104 iv Tables Table 1. Marginal Effects ..................................................................................................... 93 Table 2. Program Effectiveness ........................................................................................... 95 Table 3. Simulated Reforms to the LEAP Program, LEAP+ Households ............................. 97 v ACKNOWLEDGMENTS The authors express their gratitude to a number of individuals and organizations, including Shanta Devarajan, Jeffrey Hammer, and Ritva Reinekka for useful discussions on the scope, structure, and substance of the study, including on data and methods; and to Samik Adhikari, Oti Enoch Agyekum, Cynthia Nimo Ampredu, Christopher Burningham, Katherine Chen, Enrico Calvanese, Franklin Kuma Kwasi Gawu, Jed Friedman, Mawuko Fumey, Ugo Gentilini, Antonio Giuffrida, Sakshi Hallan, Vincent Van Halsema, Kwabena Gyan Kwakye, Lucy Liu, Patrick Mullen, Alex Nartey, Mpumelelo Nxumalo, Paul Rodas, Anita Schwarz, Iffath Sharif, Karishma Talitha Silva, Stephen Tettevie, Jennifer Yablonski, and Jonathan Nasonaa Zakaria for valuable discussions, feedback, and assistance at various stages during the production of this study. We also thank several government officials for an excellent partnership, including (listing organizations and officials’ last names in alphabetical order) Franklin Ashiadey, Ghana Jobs and Skills Project Coordinating Unit; Peter Peprah, Ghana Statistical Service; Alhassan Iddrisu and Yvonne Quansah, Ministry of Finance; Alhaji Ibrahim, Richard Nartey, Myles Ongoh, Chris Siaw-Darko, and Johnson Wegba, Ministry of Gender, Children, and Social Protection; Patrick Appiah, Adwoa Boakyewaa Asotia-Boakye, Henry Bosompem, Albert Dadze Dennis, Desmond Duametu, George Kwadwo Osei-Ababio, and Stephen Tekpertey, Ministry of Local Government, Decentralization, and Rural Development; Charlotte Norman, National Disaster Management Organization; Vivian Addo-Cobbiah, National Health Insurance Authority; and Evelyn Adjei and Jones Kennedy Luri, Social Security and National Insurance Trust. We apologize to any individuals or organizations inadvertently omitted from the above list and express our gratitude to all who contributed to the study. The study was funded by the World Bank. vi SUMMARY Ghana administers multiple social protection programs. One, pensions provided by the Social Security and National Insurance Trust, has a long history, but the rest of the programs have been introduced and expanded over the past two decades. This study assesses the performance of the government of Ghana’s main social assistance and social insurance programs. It discusses the main design and implementation parameters of the programs and summarizes existing evaluative and operational research. The study also examines patterns and trends in program benefit spending, based on government administrative data, and the coverage rates of the programs, their incidence, and their effectiveness in reducing poverty and inequality, based on recent national household sample survey data. Further, the study examines the relationship between household participation in social assistance programs and exposure to adverse covariate shocks, specifically, possible weather- related shocks, based on high-resolution climate risk maps for the country. Ghana spends little on public social protection programs, equivalent to about 1.4 percent of GDP. This spending is predominately through social insurance programs. Spending on social assistance programs is equivalent to about 0.2 percent of GDP, far below international averages. Until recently, social insurance pensions were limited to retirees who worked in the formal sector. National health insurance is meant to be universal. Except for these two programs, social assistance programs in Ghana are well-targeted to the poor, and the programs tend to cover households that are exposed to high drought risk (poor areas and high drought risk areas are correlated). Yet, the effects of the social assistance programs on poverty and inequality are at best modest because their outlays are small. Population coverage rates are limited for most of the programs, and average benefit amounts are low by international standards, relative to the average consumption levels of beneficiary households, and relative to the country’s extreme and overall poverty lines per adult equivalent. While expansion of the coverage of the programs and more generous benefit amounts would have a salutary effect, the current macroeconomic and fiscal situation makes such changes difficult. Social Security and National Insurance Trust pension benefits are much more generous, but are basically received by a population (formal sector workers) that is generally well-off. While we find much that is positive about social protection programs in Ghana, an important and frequent criticism is that the National Health Insurance Scheme is slow to pay healthcare providers for the services they provide to the scheme’s members. This discourages providers from treating members or leads them to charge them fees even when healthcare services are supposed to be free, thus undermining the purpose of national health insurance. 1 1. INTRODUCTION Ghana has relatively low levels of poverty compared to the averages for Sub- Saharan Africa and lower-middle-income countries. After steady and substantial declines from 1991/92 to 2012/13, poverty reduction stagnated between 2012/13 and 2016/2017, the most recent year for which we have an actual poverty measurement (based on the Ghana Living Standard Survey 2016/17), this despite significant economic growth in the period. Inequality rose over the decade from 2005/06 and 2016/17. In 2016/17, the poverty rate stood at 12.7 percent based on the international poverty line of $1.90 in 2011 purchasing power parity dollars, 23.4 percent based on the country’s overall poverty line, and 8.2 percent based on the country’s extreme poverty rate. The Gini index of consumption per capita was 43.0 percent. Enabled by expanded fiscal space due to debt reduction and the discovery and exploitation of significant oil and gas resources, the government of Ghana has introduced or expanded multiple social protection programs over the last two decades. This study examines benefit spending performance in the following main social protection programs of the government: a. Livelihood Empowerment Against Poverty (LEAP) program, an unconditional cash transfer to households with members in certain vulnerable groups (orphans, the elderly, pregnant women, mothers of infants, and the disabled) and low proxy means test scores in geographic areas with high poverty rates; b. Labor Intensive Public Works (LIPW) program, a public employment scheme that takes place in the agricultural off-season in rural areas and uses a combination of geographic targeting to poor areas, self-selection, and community selection of the poorest candidates; c. Ghana School Feeding Programme (GSFP), a program to provide cooked lunches to primary and junior secondary students in public schools in geographic areas with high poverty rates; and d. National Health Insurance Scheme (NHIS), a highly subsidized health insurance that covers 95 percent of conditions commonly found in Ghana, with no premia payments for those under 18 years old and those aged 70 years old or older, as well as the “indigent,” Social Security and National Insurance Trust (SSNIT) contributors and pensioners, pregnant women, and women with infants under 3 months (and the infants). The indigent, pregnant women, and women with young infants also do not have to pay the registration or renewal fee; and e. The Social Security and National Income Trust (SSNIT), which offers retirement, disability, and survivor pensions to its contributors, the vast majority of whom are or were in the formal sector. The social assistance programs use a variety of targeting mechanisms to reach the poor, including certain categories of vulnerable people (orphans, the elderly, pregnant women, mothers of infants, and the disabled), a proxy means test (PMT), and geographic targeting of areas with high poverty rates. SSNIT, on the other hand, almost exclusively benefits formal sector workers, very few of 2 whom are poor. NHIS is meant to be universal. LIPW program benefits are fully financed by donors. LEAP program benefits are jointly financed by the government (general revenues) and donors. The donor financing is explicitly specified for these programs and channeled through government. GSFP, NHIS, and SSNIT pension benefits do not receive donor financing. GSFP benefits are financed through general government revenues. NHIS is funded primarily through an earmark of 2.5 percent of VAT revenues (the National Health Insurance Levy) with additional funding from SSNIT, which pays the premia and fees for its contributors and pensioners. A small amount of NHIS funding comes from other members’ premia and fees. SSNIT benefits are funded from contributions from members (salaried employees) and their employers. In this study, we describe the main design and implementation parameters of each of the five programs and selectively review the available descriptive, diagnostic, and evaluative research on the programs. The study’s main contributions are to describe the patterns and trends of spending on each of the programs and to assess each program’s performance in terms of coverage, incidence, and effectiveness at reducing poverty and inequality, including through the use of microsimulation of selected reforms. The study also examines the relationship between social assistance program coverage and adverse covariate shocks, specifically weather-related shocks. The analysis of incidence and effectiveness is based on standard methods (Bourguignon and Pereira da Silva 2003; Lustig 2018; Yemtsov et al. 2018), including standard microsimulation methods (Figari, Paulus, and Sutherland 2015). 1 The descriptions of program design and implementation parameters and the analysis of program spending patterns and trends are based primarily on information shared with the authors by the various program administrators. The analysis of incidence and effectiveness relies on the Ghana Living Standards Survey 2016/17, which is the latest available national household sample survey with relevant data for our performance review. The GLSS 2016/17 questionnaire includes questions about participation in, and benefits from, the LEAP program, GSFP, NHIS, and SSNIT pensions (but not the LIPW program). Measures of poverty and inequality are based on household consumption data in the survey. The government administers and uses the GLSS for its official estimates of poverty and inequality in the country. While we examine the important relationship between social protection programs, poverty, and inequality, 2 we recognize that the primary aims of these programs may be different. For example, NHIS is mainly meant to improve access to healthcare, and GSFP is meant to improve school enrollment, attendance, and academic achievement. We summarize existing evidence on the 1 For SSNIT pensions, the only program that offers substantial future benefits, an actuarial analysis would be appropriate, and the International Labor Organization has conducted such analyses for SSNIT, but, to the best of our knowledge, they are not published or publicly available. 2 Given the sharp rise in world food prices, considering the association between program participation and food security status is also of interest, but the GLSS 2016/17 questions on food security are nonstandard and sufficiently limited that we cannot examine this association. 3 relationship between social protection program participation and other outcomes. This study follows a series of high-quality studies on the performance of social protection programs in Ghana, based on analyses of household sample survey and government administrative data on programs and through reviews of existing research. See, for example, Wodon (2012), ILO (2015), World Bank (2016), Younger, Osei-Assibey, and Oppong (2017), and World Bank (2017). Some of these studies have covered a larger set of interventions than our review, including for example, various subsidies and active labor market interventions. Some of the studies have conducted benefit incidence analyses and microsimulations of hypothetical program reforms, using previous rounds of the GLSS (from 2005/06 and 2012/13)—just as our study does. Many of our findings and recommendations match those of these past studies. Prior to the coronavirus pandemic, in 2019, total social protection program benefits amounted to GH¢4.5 billion. SSNIT pensions accounted for 65 percent of this spending, NHIS for 22 percent, GSFP for 10 percent, and the LEAP program for 3 percent. The LIPW program was inactive in 2019. Total benefit spending in all five social protection programs across recent years was equivalent to 6.7 percent of overall government spending and to 1.4 percent of gross domestic product (GDP). 3 Total spending on social assistance program benefits across recent years was equivalent to 0.9 percent of overall government spending and to 0.2 percent of GDP. This level of social assistance program benefit spending is substantially lower than the average for Sub-Saharan Africa and for developing countries more generally, at 1.5 percent for both groups of countries. The pandemic did not result in a marked increase in spending on these programs, as the government opted to mainly respond with other instruments (discussed below). None of the social protection programs we review except NHIS is meant to cover all Ghanaians. NHIS’s coverage of the national population has ranged between 30 percent and 40 percent over the past decade, based on government administrative data on end-of-year active membership numbers. Even though they are very low, available evidence suggests that the costs of registration or renewal fees and premia are the main reason that people do not register. The other programs are restricted to poor areas (the LEAP program, the LIPW program, and GSFP) and to people with certain characteristics—infants, (school) children, the elderly, the disabled, pregnant women, mothers of infants (the LEAP program, GSFP, and SSNIT pensions). The LEAP and LIPW programs also target poorer households in poor areas. As a result, national coverage rates are low. Coverage rates jump when we restrict our analysis to population groups targeted by the programs, but remain far from universal. Based on survey data, the LEAP program covers only 1.5 percent of Ghanaians and 6.7 percent of the extreme poor nationally. The coverage rate increases to 3 The noted percentage of overall government spending should be interpreted with caution given important nongovernment sources of financing of NHIS and SSNIT pensions, discussed later in the study. 4 22.7 percent of the overall population in LEAP program areas, and to 27.8 percent of the extreme poor in these areas. However, LEAP program benefit receipt reported by GLSS 2016/17 respondents, the basis for our analysis of coverage, is underreported by 67 percent. When we predict a set of additional LEAP program households to address underreporting in the survey data and add it to the set of actual LEAP program households (a collective set that we refer to as “LEAP+” households), coverage rates increase to 4.1 percent of all Ghanaians and 17.9 percent of the extreme poor. In LEAP+ program areas, the coverage rate increases to 28.8 percent of all residents and to 34.8 percent of the extreme poor. GSFP’s coverage of the national population rose from 6 percent in 2016 to 8 percent in 2019, based on administrative data. The program covers 10 percent of the extreme poor, based on survey data. But among public preprimary and primary school students in targeted areas, the coverage rate is 61 percent overall and 64 percent for the extreme poor. Over the last decade, SSNIT pensions’ coverage of the national population has ranged between 0.4 percent and 0.7 percent, based on administrative data. The program covers 5.2 percent of those 60 years old or older, based on survey data. SSNIT pensions do not reach the poor. The main limitation here is that until recently SSNIT was available only to formal sector workers, which is a small share of the overall labor force. Aside from SSNIT pensions, the levels of total benefits received by households from the various programs are low. As a percentage of total household consumption among program households, the LEAP program averages 12.8 percent, GSFP 5.2 percent, and NHIS 3.6 percent. For SSNIT pensions, the percentage averages 60.3 percent. The average percentage increases for all programs when the sample is restricted to poor households; however, it continues to remain low for all programs except SSNIT pensions. The results are qualitatively similar when we examine total benefits as a percentage of the extreme and overall poverty lines (set at the household level) averaged across program households. LEAP program and GSFP benefits are highly concentrated among the poor with concentration coefficients of –0.53 and –0.22, respectively. These compare favorably with similar social assistance programs in a large sample of low- and middle-income countries. The concentration coefficient for NHIS benefits is 0.19, not pro-poor, but still more concentrated among the poor than household consumption in general. (The Gini coefficient of consumption per adult equivalent is 0.42.) Of course, if NHIS were truly universal as intended, it would have a concentration coefficient of zero. The concentration coefficient for SSNIT pensions is 0.74, which is highly regressive. This is not surprising as SSNIT pensions are relatively generous and, until recently, available only to those with at least 15 years of employment in the formal sector. Despite the strong targeting performance of the LEAP program and GSFP, they have only small effects on poverty (based on the country’s extreme and extreme poverty lines) and inequality (the Gini index), at about one-tenth of a percentage 5 point or smaller. The effects are somewhat larger for LEAP+, GSFP in GSFP areas, and NHIS, at about one-fourth of a percentage point or smaller. The effects are largest for the LEAP program in LEAP areas, at about four-fifths of a percentage point. Finally, SSNIT pensions have almost no effect on poverty and inequality. We measure effectiveness at reducing poverty (based on the country’s extreme and overall poverty lines) and inequality (Gini index) by comparing the actual changes induced by each program to a perfectly targeted transfer with the same outlay (“impact effectiveness”), or a perfectly targeted transfer that achieves the same impact but with a smaller outlay (“spending effectiveness”). The LEAP program is 69 percent as effective at reducing the Gini index as a perfectly targeted program would be. It is 65 percent as effective as a perfect program at reducing the poverty gap, but only 33 percent to 35 percent as effective at reducing poverty severity (both measures based on the overall poverty line). The effectiveness results for LEAP+ are similar. GSFP is somewhat less effective than the LEAP program at reducing inequality and poverty: It is 57 percent as effective as a perfect program at reducing inequality, 48 percent as effective at reducing the poverty gap (based on the overall poverty line), and only 24 percent to 28 percent as effective at reducing poverty severity (based on the overall poverty line). The LEAP program in LEAP program areas and GSFP in GSFP areas are much less effective at reducing inequality in these areas than they are nationally. NHIS is 28 percent as effective as a perfectly targeted transfer at reducing the Gini index, 22 percent as effective at reducing the poverty gap (based on the overall poverty line), and 11 percent to 16 percent as effective at reducing poverty severity (based on the overall poverty line). SSNIT pensions actually increase inequality and are completely ineffective at reducing poverty. Because of LEAP’s prominence in the government’s antipoverty strategy, we simulate different reforms to the program’s coverage and benefit levels, based on the LEAP+ sample. We find that using the government’s PMT to target the extreme poor would improve the program’s already impressive concentration coefficient from –0.49 to about –0.73, even if the LEAP program were to abandon its current multilevel, multimethod targeting strategy. Further, we estimate that scaling up the LEAP program to cover all areas nationally while restricting benefits to the PMT-extreme poor only would reduce the program’s outlay by 65 percent. Notwithstanding, the program’s effect on poverty remains small across various simulations of coverage expansions (based on PMT targeting) because of its very low benefit level. Simulations of benefit increases also show limited poverty effects because the existing benefit level is low. In terms of survey data quality, the correspondence between the GLSS 2016/17 estimates and government administrative information on beneficiary numbers and benefit spending is not always good. Total beneficiaries and benefits for the LEAP program are both underreported by 67 percent in the survey data, by 33 percent and 30 percent for GSFP, by 12 percent and 22 percent for NHIS, and 6 by 43 percent and 64 percent for SSNIT pensions. It is especially important to keep this in mind when considering our estimates of coverage and the effects of each program on poverty and inequality, which will be biased downwards. If the underreporting is distributed randomly across the income distribution, then it will not bias estimates of incidence or effectiveness. But if the underreporting is more heavily concentrated among richer (poorer) households, concentration coefficients and effectiveness estimates will be biased downward (upward) as well. The GLSS 2016/17 questionnaire does not include questions on shocks that respondents have experienced, and thus does not allow for an examination of the correlation between shocks and social assistance program participation. We draw on Nxumalo and Raju (2022) who link data on drought and flood risks with data on household program participation at the district level. District program participation rates in the LEAP and LIPW programs are correlated with drought risk, while the results for flood risk are less clear. The association between program participation and drought risk is an indirect result, partly mediated through poverty. That is, social assistance programs target poor areas, and poor areas tend to be drought prone. The link between household participation in the LEAP and LIPW programs and drought risk stem from pre- shock program coverage of drought-prone areas, achieved indirectly through the intentional coverage of poor areas. However, in the case of the LIPW program, there might be an element of ex post response at work. The location, type, and scale (including workforce size) of the public works subproject may be influenced by local drought-risk considerations. Unlike social assistance programs in other low- and lower-middle-income countries, those in Ghana have been the subject of much academic research, including studies with careful designs that allow identification of causal effects. We selectively review this research in this study. In contrast, our own results are descriptive only. But the descriptions are nationally representative, which the causal studies are not. As such, they provide a useful complement to the existing literature on these programs’ effects. One positive aspect of the implementation of social protection in Ghana is that, unlike many other lower-middle-income countries, researchers and practitioners review the programs with some care and, at least for the LEAP program and NHIS, have made important adjustments to the programs in response to this information. For the LEAP program, careful evaluations (Handa et al. 2014; Handa et al. 2017) find significant problems with the distribution of payments, which were irregular and sometimes difficult to obtain. The LEAP program changed payment mechanisms to a more modern electronic distribution system which greatly improved the regularity of payments. Handa et al. (2017) find large increases in consumption, fertilizer use, and agricultural assets (including livestock) among households who gained access to the LEAP program, but are unable to distinguish these gains from a control group that had similar gains. 7 Another evaluation of the LEAP 1000 intervention, which targeted LEAP program transfers and free NHIS membership to pregnant women and young infants and their mothers finds modest and mixed impact on a variety of measures of wellbeing, but this was to be expected as the study compared households that were just below and just above the PMT cutoff value for participation. The rate of NHIS membership did increase substantially. Osei-Akoto et al. (2016) evaluate the LIPW program based on a field experiment. They find that the program increased individual employment and earnings outcomes, measured variously. The authors also conducted focus group discussions to complement the quantitative data. Respondents mostly expressed satisfaction with the program, with the main complaints being long delays in receiving their payments, and conflicts between the demands of LIPW program work and work on their own farms (which should not occur since LIPW subprojects are meant to take place only in the off season). For GSFP, a field experiment of a significant expansion of the program found that it increased height-for-age, a standard nutritional measure, for girls and for poorer students, and increased scores on standardized tests, with larger increases for poor students (Gelli et al. 2019; Aurino et al. 2023). Operational reviews reveal that the government is often slow to pay the caterers who prepare the school meals, and that the amount paid has been eroded by inflation to such an extent that the caterers can no longer make a profit if they provide the stipulated nutritionally adequate meal. For NHIS, there is a much larger literature. Blanchet and Acheampong (2013), Okoroh et al. (2018), and Degroote, Ridde, and Del Allegri (2020) provide useful literature reviews. There is a clear consensus across studies based on a variety of methods that: a. NHIS membership increases utilization of healthcare services; b. NHIS members have lower out-of-pocket healthcare expenditures than nonmembers; c. NHIS members have lower “catastrophic” healthcare expenditures, variously defined; and d. NHIS members usually do not have better health outcomes than non- members. There also have been many operational reviews and adjustments to NHIS operations, including a change from fee-for-service reimbursements to standardized diagnostic related groupings (DRGs) and an experiment with capitation payments for outpatient services. An important theme in the operational reviews is that healthcare providers struggle with delays in claims reimbursement or outright rejection of their claims. This has caused some to refuse to see NHIS patients, others to charge “copays” that should not, in fact, be charged to members, and still others to send their patients to pharmacies to buy medicines that should be provided under NHIS. (See Blanchet and Acheampong 2013 and Christmals and Aidem 2020 for reviews.) In terms of the response to the pandemic, as noted earlier, the government did 8 not resort to a large response through its main social protection programs. It did not make any major adjustments in coverage or benefits under GSFP, the LIPW program, SSNIT pensions, and NHIS. GSFP operations were disrupted when schools closed, and LIPW program operations were interrupted for a brief period around the three-week lockdown of Accra and Kumasi metropolitan areas in April 2020. Following the lockdown, the LEAP program intended to offer a special additional two months of benefits (that is, 14 months of benefits in 2020 instead of the stipulated 12 months of benefits every year). However, due to subsequent delays in budgetary releases to the program, it could only provide 10 months of benefits to program households by the end of 2020. Thus, instead of a bump up in total benefits in 2020, it transformed into a dip in the year. Notwithstanding, financed by the World Food Programme (WFP), the government provided small cash transfers (in two installments) over the period spanning late 2020 and early 2021 to daily wage earners in Accra and smallholder farmers in the Western and Ashanti regions (about 74,000 individuals in total). Financed by the World Bank, in late 2020, the government also started providing small, one-time cash transfers individuals from Accra and Kumasi and from other parts of the country. identified as poor, vulnerable, or pandemic-affected based on data gathered under the Ghana National Household Registry. The intervention aimed to reach 124,000 individuals. The main responses to the pandemic by the government included cooked meals and dry rations to poor and vulnerable individuals in Accra and Kumasi during the lockdown in April 2020 (2.7 million cooked meals; dry rations to 470,000 families). They also included free water to 10 million customers (9 months for regular customers; 12 months for low-consuming customers) of the Ghana Water Company Limited, and free or subsidized electricity to 4.7 million customers (9 months for regular customers; 12 months for low-consuming customers) of the Electricity Company of Ghana and the Northern Electricity Department of the Volta River Authority. Both interventions were initiated in April 2020. Through the Coronavirus Alleviation Programme Business Support Scheme launched in May 2020, the government also provided soft loans to over 300,000 micro, small, and medium-sized enterprises. In terms of scale and the volume of spending, these responses dwarfed the various cash transfer-based responses. All these interventions were concentrated in urban areas, particularly Accra and Kumasi. See Hallan and Raju (2023) for a detailed description of the social protection responses to the pandemic in Ghana. In setting social protection policy, the government benefits from an unusually large evaluative and operational research literature which it appears to use to good effect. The targeting of the LEAP program and GSFP, in particular, is both technocratic and highly effective at reaching the poor. The impact of these programs on poverty and inequality, though, is limited due to their limited outlays and the stinginess of their benefits which inflation has eroded in real terms, and which never have been generous. An obvious recommendation, then, is to increase outlays for these programs. Unfortunately, that runs head on into a difficult macroeconomic and budgetary environment that, because of high debt burdens, seems unlikely to improve in the medium term. General 9 government debt is estimated at 89 percent of GDP in 2022. 4 Deficits have been persistent for over a decade, with double-digit deficits since the onset of the pandemic. Thus, the government is left with a frustrating trade-off: it has proven tools to reduce poverty and inequality effectively, but must balance those with the exigencies of fiscal deficit reduction. The recent government announcements of large increases in the benefit amounts of the LEAP program and GSFP do not in any way imply that this trade-off is less severe. The LEAP program, GSFP, and presumably the LIPW program have made good use of geographic targeting. Both the LEAP program and GSFP have better targeting nationally than they do within the geographic areas where they are active. That is, both programs have chosen the areas in which to operate well given their intention to reach the poor. As these programs continue to expand their national coverage, as both have done over the past decade, they will lose their effective geographic targeting which may worsen their targeting overall. On the other hand, the government uses a PMT to target the LEAP program to households within selected communities, and our simulations show that its sole use can be as effective or more so than the prior multilevel, multimethod targeting strategy used by the program, which included targeting individuals with specific sociodemographic characteristics as well as those living in poor areas. So, there is every reason to believe that the LEAP program will continue to have highly pro-poor targeting. GSFP, however, cannot target individual children; it is a school-level intervention. Expansion to nationwide coverage, then, will certainly dilute its targeting effectiveness. This suggests a policy adjustment: The government could concentrate its social assistance program spending meant to alter the distribution of income directly in the LEAP program and perhaps the LIPW program where the PMT can be used to target the poor effectively. GSFP, on the other hand, would cease to be a transfer program. Instead, it could be assigned to schools where it is most likely to improve children’s health and education outcomes, based on the evaluation results in the academic literature. Of course, these schools would have many poor children, but the redistributional effects would be a side benefit, not the main purpose, which would be nutrition status, school enrollment and attendance, and academic achievement. NHIS presents the most daunting problems we find in this study. Unlike the other social protection programs, NHIS is meant to be universal, i.e., to provide health insurance for all. Yet over the past fifteen years, enrolment has been 30 percent to 40 percent of the population. Yet even for this reduced population coverage, NHIS has difficulty covering its expenses. By far the most common criticism we found of any social protection program in Ghana is that NHIS pays healthcare providers late or not at all for the services they provide to NHIS members. If NHIS is to be truly universal, it will need substantially more resources, more than twice its current outlay. Once again, this collides with the macroeconomic and fiscal situation. It seems unlikely that the government can allocate the additional resources need to make NHIS universal. 4 Based on data from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. 10 What other options are there? Raising fees and premia might seem to be the obvious answer. They are very low, covering less than five percent of NHIS expenditures. But it is also true that operational evaluations and survey data suggest that fees and premia, however small, are the main impediment to people registering or reregistering for NHIS. Lacking this, NHIS must either accept that it will not cover everyone, or reduce the scope of the coverage it offers, or gain an implausibly large allocation from the central budget, perhaps by increasing the national health insurance levy. SSNIT is also far from universal. In the past few years, it has tried to draw in more informal sector workers, but the response has been minimal. And there is a lesson from the NHIS experience: The government probably cannot and will not fund a universal pension, so any expansion of membership will require actuarially fair premia payments from new members, something that will probably keep them from joining. An important observation from many operational reviews and the press is that the government has a pervasive and chronic problem: it often accumulates substantial arrears to its service providers. Over the previous decade, the government failed to pay more than GH¢3 billion it owed to SSNIT for its employees’ premia. NHIS is so slow to reimburse healthcare providers that many now refuse to accept NHIS patients, or charge them illegal copays. GSFP caterers and LIPW program participants have sometimes suffered long delays in payments as well. The convenience of these arrears for a cash-strapped government is understandable: They constitute an interest-free loan and also hide the true size of the deficit, for a while. But arrears have real costs in terms of program effectiveness. The arrears to SSNIT reduce its investment earnings, jeopardizing its already doubtful ability to pay future pensions. The arrears to healthcare providers cause them to deny NHIS members the free healthcare that NHIS is supposed to guarantee. Unpaid GSFP caterers must cut corners in the school meals they provide to children. The rest of the paper is organized as follows. To help contextualize the findings, section 2 presents background information on Ghana. Section 3 describes key design and implementation parameters of the social protection programs. Section 4 discusses results from the analysis of the performance of the social protection programs. Section 5 concludes with thoughts on the future of social protection policy in Ghana. The paper also includes five appendices. These discuss the construction of key variables for the program performance analysis (appendix A), analytical concepts for the analysis (appendix B), supplemental results (appendix C), our strategy for estimating counterfactual household consumption for households with SSNIT pensions in the absence of those pensions (appendix D), and our strategy for predicting additional LEAP program households to correct for underreporting of LEAP program households in the GLSS 2016/17 data (appendix E). 11 2. BACKGROUND This section presents background information on Ghana to help contextualize the findings of our analysis on the performance of the government’s main social protection programs, based on GLSS 2016/17 data and program administrative data. This discussion covers population; national income and economic growth; consumption and income, poverty, and food insecurity; inequality; shocks; economic growth and poverty trends during the coronavirus pandemic period; overall government revenues and expenditures; and the overall social protection sector. 2.1. Population As of 2021, Ghana had a total population of 30.83 million people, with, 97.6 percent residing in households and 2.4 percent in institutional arrangements. 5 The average size of a household was 3.6 persons, a decline of about 1 person on average since 2010. Declines in average household size have been particularly large in the Northern, Savannah, North East, and Upper West regions, all in the north of the country. Average household size is smaller in urban areas (3.3 members) than in rural areas (4.0 members) (GSS 2021a). In 2016/17, the average household size was 3.8 persons, with the average size decreasing with consumption quintile, from 6.2 persons in the poorest quintile to 2.4 persons in the richest (GSS 2019). Four of the country’s 16 regions account for more than half of the national population, namely Greater Accra (17.7 percent), Ashanti (17.6 percent), Eastern (9.5 percent), and Central (9.3 percent), all in the south (figure 1). Most of the national population is urban, at 56.7 percent, an increase from 50.9 percent in 2010. The urban share of the population ranges from a low of 25.4 percent in the Upper East region to a high of 91.7 percent in the Greater Accra region. Most regions in the south of the country are majority urban, while the regions in the north are majority rural (GSS 2021a). In terms of age structure, in 2021, 35.3 percent of the national population was between ages 0–14 years, 60.4 percent was between ages 15–64 (standard working age), and 4.3 percent was above age 65. The age structure varies markedly across regions. At the top end, 66.5 percent of the population in the Greater Accra region are between ages 15–64, while, at the bottom end, 51.4 percent are in this age group in the North East region (GSS 2021b). These patterns are indicative of strong rural-to-urban migration and north-to-south migration, importantly for work opportunities. 5 Households include the homeless. 12 Figure 1. Population Distribution, 2021 a. Distribution across regions b. Urban-rural distribution, by region Source: Statistics obtained from GSS (2021). Note: Figure shows the distribution of Ghana’s population across regions (panel a) and the population distribution across urban and rural areas in each region (panel b). 13 2.2. Economic growth Ghana is classified as a lower-middle-income country by the World Bank. In 2021, the country’s GDP per capita totaled $5,435 (in constant 2017 purchasing power parity international dollars). Growth in GDP per capita averaged 4.3 percent annually between 2010–19, buoyed by the discovery and exploitation of offshore oil and gas reserves (figure 2). Ghana began the 2010s with GDP per capita similar to the average for Sub-Saharan Africa but grew much faster than the region as a whole. The rapid growth has been nevertheless erratic, with large swings driven by the country’s dependence on mining, oil, and gas for national income, a serious drought that affected power supply, and perhaps a political business cycle. 6 Between 2010 and 2019, GDP growth rates ranged from a low of –0.3 percent (in 2015) to a high of 11.3 percent (in 2011). GDP per capita growth fell to –1.5 percent in 2020, due to the coronavirus pandemic. It rebounded to 3.3 percent in 2021. Decomposition of Ghana’s economic growth between 1970 and 2016 shows the main contributions of change—total factor productivity, capital accumulation, or labor accumulation (including labor quality as reflected by education)— during different subperiods. The analysis suggests that the change in labor accumulation has contributed meaningfully to the country’s economic growth in all subperiods. The contribution to growth from the change in total factor productivity became positive and sizeable roughly between 1990 and 2010. Meanwhile, the contribution from the change in capital accumulation became positive and sizeable roughly between 2005 and 2015 (Nxumalo and Raju 2020). 2.3. Consumption and income In 2016/17, Ghana’s annual household consumption averaged GH¢12,900. Greater Accra ranked as the wealthiest region, with an annual household consumption level of GH¢21,300, while Upper West was the poorest, with an annual level of GH¢6,100—more than a three-fold difference. In general, the regions in the north have the lowest consumption levels. The household consumption level is much higher in urban areas (at GH¢15,600) than in rural areas (GH¢9,400) (GSS 2019). 6 Ghana holds national elections every four years, with the most recent in 2020. See Younger (2016) for a discussion of macroeconomic developments in Ghana, including political business cycles. 14 Figure 2. National Income Trends a. GDP per capita b. GDP per capita growth Source: Statistics obtained from the World Bank’s World Development Indicators databank. Note: The figure shows trends in annual GDP per capita in 2017 purchasing power parity (PPP) international dollars and annual GDP per capita growth for Ghana. It also shows averages for these measures for Sub- Saharan Africa and lower-middle-income countries. SSA = Sub-Saharan Africa. LMICs = lower-middle-income countries. 15 In terms of categories of goods and services consumed, food constituted 42.9 percent of total consumption for households nationally, housing 15.8 percent, and other nonfood consumption 41.3 percent. The food share in total consumption is higher in rural areas than in urban areas (50.6 percent versus 39.2 percent); the share also declines with consumption quintiles (from 49.2 percent in the poorest quintile to 38.2 percent in the richest) (GSS 2019). Annual gross household income averaged GH¢33,900 for Ghana in 2016/17, with the income level for urban areas about 2.5 times higher than for rural areas. In terms of quintiles, the income level was about 7 times higher for the richest income quintile than the poorest. On average, across the country, earnings from nonfarm self-income employment constituted three-quarters of household income, followed by wage employment earnings at 14.1 percent and agricultural earnings at 5.0 percent, rental income at 3.6 percent, and remittances at 1.4 percent. Transfers were a negligible source of household income. Compared to richer households, agricultural earnings are a much more important source of income for poorer households, while nonfarm self-employment earnings are a much less important source. There is no clear pattern with respect to the contribution of wage employment earnings to household income across income quintiles (GSS 2019). 2.4. Poverty and food insecurity Poverty levels in Ghana declined steadily beginning in the 1990s through 2012/13, when they stopped falling despite significant economic growth between 2012/13 and 2016/2017 (figure 3). The country’s poverty rate was 12.7 percent in 2016/17, measured using the international poverty line of $1.90 per day per person in 2011 purchasing power parity international dollars. This level of poverty is lower than the average for Sub-Saharan Africa or for lower- middle-income countries in general (World Bank 2018). Measured using the country’s poverty lines, Ghana’s overall poverty rate in 2016/17 was 23.4 percent, whereas its extreme poverty rate was 8.2 percent. The country’s overall and extreme poverty gaps were 8.4 percent and 2.8 percent, respectively (GSS 2018). 16 Figure 3. Poverty Rate Trends, Based on International Poverty Line ($1.90-Per-Day Line) Source: Statistics obtained from the World Bank’s World Development Indicators databank. Note: Figure shows the trend for the poverty rate based on the $1.90-per-day poverty line in 2011 purchasing power parity international dollars for Ghana. It also shows the average poverty rates based on the same poverty line for Sub-Saharan Africa and lower-middle-income countries. SSA = Sub-Saharan Africa. LMICs = lower-middle-income countries. Poverty rates appear to differ markedly across areas within Ghana (figure 4). They are substantially higher in rural areas than in urban areas. In 2016/17, the overall poverty rate in rural areas was 39.5 percent, five times higher than the 7.8 percent for urban areas. The extreme poverty rate was 15.6 percent in rural areas, compared to 1 percent in urban areas. Overall and extreme poverty rates in 2016/17 are also markedly higher in the regions of Volta, Upper East, Northern, and Upper West than in other parts of the country. These high-poverty regions are situated in the northern and eastern parts of Ghana. 17 Figure 4. Extreme and Overall Poverty Rates, by Region, 2016/17 Source: Statistics obtained from GSS (2018). Note: Figure shows extreme and overall poverty rates for the country as whole, by urban versus rural areas, and by region in 2016/17. The classification of regions is per Ghana’s 2010 Population and Housing Census. Poverty levels also vary within high-poverty regions. Maps for 2010 for poverty levels across administrative districts in the country show that poverty levels appear to be higher in the western parts of the high-poverty regions (figure 5). Spatial patterns in where infrastructure, services, and markets are lacking show strong overlap with the spatial pattern in where poverty levels are high, suggesting that these factors may be constraining households from achieving sufficiently high incomes to escape poverty (World Bank 2018). High-poverty regions also face adverse ecological conditions, which are considered to play an important role in constraining growth, development, and poverty reduction (World Bank 2018). 18 Figure 5. Distribution of Poverty across Districts, 2010 a. Poverty rate (percent) b. Number of poor persons c. Poverty gap1 (percent) d. Poverty severity2 (percent) Source: Statistics obtained from GSS (2015). Note: Maps show the distribution of poverty across districts based on various indicators for 2011/12. 1Poverty gap is a measure of how far the poor are from the poverty line. 2Severity of poverty is the square of the poverty gap, which gives greater attention to the needs of the poorest. It takes account of the distribution of poverty among the poor, giving greater weight to the poorest of the poor. The classification of districts is per Ghana’s 2010 Population and Housing Census. 19 The extent of households that are classified as extreme poor, “moderate poor,” or “near poor” in 2016/17 shows some differences across selected population subgroups (figure 6). The extreme poor are those with consumption (per adult equivalent) below the extreme poverty line, the moderate poor are those with consumption between the extreme and overall poverty lines, and the near poor are those with consumption above the overall poverty line but below 1.5 times the overall poverty line. There is no difference in poverty rates between males and females, a fact that is accounted for by the fact that consumption per adult equivalent is measured at the household level and that households, on average, have roughly equal numbers of males and females. Households headed by a female are less likely to be moderate poor and much less likely to be extreme poor than those headed by men. This pattern is often seen in countries where significant numbers of men migrate (internally or internationally) for work that pays relatively well and then send remittances back to their household. Children younger than 15 years old reside in households that are somewhat more likely to be poor than do people of working age. The same is true for people 60 years and older. Figure 6. Poverty Rates, by Selected Subgroups, 2016/17 Source: Authors’ estimates based on GLSS 2016/17 data. Note: Figure shows the shares of extreme, moderate, and near poor for selected subgroups in 2016/17. hh head = household head, as defined by the household. Orphan status is restricted to children below age 18. One parent deceased = one parent deceased or unknown. Both parents decreased = both parents deceased or unknown. Lower sec. = lower secondary school. Upper sec. = upper secondary school. Post sec./tech. = post-secondary or technical school. Across all these age groups, the differences in the extreme poverty rate are greater than those for the moderate poverty rate or the near poverty rate. Extreme poverty is highly concentrated among households that have had no formal education. Poverty rates also show a strong gradient across education- 20 attainment levels. The relationship between near poverty and education also has a negative correlation, but it is less strong than that for moderate poverty and (especially) extreme poverty. Perhaps surprisingly, for children under age 18, whether they have lost one or both parents has little relationship with their poverty status, a testament to the willingness of Ghanaian families that are better off to take in orphans. Even though extreme and overall poverty rates remained roughly constant across the country as a whole between 2012/13 and 2016/17, the rates in urban areas continued to decline apace, while those in rural areas increased (figure 7). These patterns are especially notable in the regions with large urban populations (Greater Accra and Ashanti) and some of those that are predominantly rural (Volta, Northern, and Upper East). This may be explained by the fact that 2017 was an El Niño year, which tends to produce droughts in the northern areas of Ghana. The small change in the overall poverty rate between 2012/13 and 2016/17 (–0.8 percentage points) can be decomposed into a “growth contribution” (that which is attributable to the increase in average consumption during the period) and a “redistribution contribution” (that which is attributable to a change in the share of consumption going to the poor or nonpoor) (figure 8). Nationally, economic growth contributed to a 2.3 percentage point decline in the overall poverty rate, but this was offset by increasing inequality, which led to a 1.4 percentage point increase in the overall poverty rate. Thus, overall consumption grew at a healthy pace over these four years, but the growth was uneven, favoring the nonpoor. Interestingly though, we do not see this same pattern in most of the regions in the country. In urban areas, for example, both growth in consumption and a reduction in inequality contributed to a lower overall poverty rate. The same is true in many regions, the exceptions being Western, Volta, and Upper West. In all three, the growth effect is negative, but it is offset by decreasing inequality. Overall, these patterns suggest a migration pattern in which people from poorer regions have migrated to richer ones, especially cities. This has the effect of increasing consumption in the cities and decreasing it (relatively) in the poorer rural areas, but also increasing inequality in urban areas as the new migrants are poor relative to longtime urban residents. 21 Figure 7. Poverty Rate Trends, National, by Area and by Region, 2005/06– 2016/17 a. Overall poverty rates b. Extreme poverty rates Source: Statistics obtained from GSS (2018). Note: Figure shows trends in overall and extreme poverty rates, based on measurements in 2005/06, 2012/13, and 2016/17. The trends are shown for the country as a whole, by urban versus rural areas, and by region. The classification of regions is per Ghana’s 2010 Population and Housing Census. 22 Figure 8. Growth and Redistribution Decomposition of the Change in Poverty Rate between 2012/13 and 2016/17 a. National and by area b. By region Source: Statistics obtained from GSS (2018). Note: Figure shows the decomposition of the observed percentage-point change in the poverty rate between 2012/13 and 2016/17 in Ghana into the percentage-point change in the mean value of consumption, assuming that inequality in welfare remained unchanged (the “growth” effect), and the percentage-point change in inequality of welfare, assuming that mean consumption remained unchanged (the “redistribution” effect). These decompositions are performed at the national, urban, rural, and regional levels. Classification of regions is per Ghana’s 2010 Population and Housing Census. In November–December 2020, a large-scale national household sample survey was fielded for a comprehensive food security and vulnerability analysis (CFSVA), which captured data on household food insecurity. The survey report 23 indicates that 11.6 percent of households across the country were classified as food insecure, of which 6.5 percent were classified as moderately food insecure and 5.2 percent as severely food insecure (GSS et al. 2021). The prevalence of food insecurity was higher in rural than urban areas (18.2 percent versus 5.5 percent, a threefold difference) and much higher in the northern regions than in the southern regions (figure 9). Regional food insecurity prevalence rates ranged from a low of 3.5 percent in Greater Accra to 48.7 percent in Upper East (a 14-fold difference). The household correlates of the likelihood of food insecurity correspond qualitatively to the household correlates of the likelihood of poverty (GSS et al. 2021; World Bank 2018, 2020). Figure 9. Food Insecurity Rates, 2020 Source: Statistics obtained from GSS et al. (2021). Note: Figure shows food insecurity rates for the country as a whole, by urban versus rural areas, and by region. 2.5. Inequality Ghana’s level of inequality in household consumption per capita, measured by the Gini index, was 43.0 percent in 2016/17 (figure 10). It was higher in rural areas (41.8 percent) than in urban areas (37.9 percent). In terms of regions, it was lowest in Greater Accra (35.1 percent) and highest in the Upper East (48.1 percent). Indeed, inequality levels appear to be lower in the richer, southern regions than in the poorer, northern regions. These patterns run counter to what is usually observed in other countries. Between 2005/06 and 2016/17, the level of inequality increased in Ghana (by about one percentage point), driven by the increase in inequality in rural areas (figure 11). 24 Figure 10. Inequality, National, by Area, and by Region, 2016/17 Source: Statistics obtained from GSS (2018). Note: Figure shows inequality in consumption, measured by the Gini index, for the country as a whole, by urban versus rural areas, and by region in 2016/17. Classification of regions is per Ghana’s 2010 Population and Housing Census. Figure 11. Inequality Trends, National, by Area, and by Region, 2005/06–2016/17 Source: Statistics obtained from GSS (2018). Note: Figure shows trends in inequality in consumption, based on measurements in 2005/06, 2012/13, and 2016/17. The trends are shown for the country as a whole, by urban versus rural areas, and by region. Classification of regions is per Ghana’s 2010 Population and Housing Census. 25 2.6. Shocks While Ghana has conducted several nationally representative household sample surveys in recent years, these surveys lack well-constructed modules that capture data on shocks experienced by individuals and households. However, the national household sample survey fielded for the 2020 CFSVA did administer a shocks module (GSS et al. 2021). The types of shocks that the survey asked about appear to be a combination of presumably more chronic, adverse circumstances (for example, lack of money to buy food or to cover basic needs, or high food prices) and sharp, acute developments (such as loss of employment by a household member), with the latter corresponding more tightly to what we would typically consider to be a shock. The 2020 CFSVA survey report indicates that, overall, about 90 percent of households experienced at least one of the 15 different types of shocks listed. The most common shocks experienced by households were the coronavirus pandemic (63.8 percent of households reported this shock), high food prices (34.1 percent), and delayed rains or drought (21.6 percent). Urban households were more likely to report pandemic or high food price shocks than were rural households; while rural households were more likely to report delayed rain or drought shocks. The survey report also provides a more detailed analysis of the pandemic shock. The predominant way that the pandemic affected households was by keeping them from (fully) pursuing their self- and wage-employment activities. The main, specific factors were curfew and lockdown measures, reduced working time, temporary layoffs, and closed workplaces. The report’s analysis of correlates suggests that wealthier households were more likely to report a pandemic shock. 2.7. Economic growth and poverty movements spanning the coronavirus pandemic Ghana’s pattern of real GDP growth rates through the coronavirus pandemic suggests a V-shaped trajectory. Real GDP grew by 6.2 percent in 2018 and by 6.5 percent in 2019. In 2020, real GDP growth fell sharply to 0.4 percent, before rebounding to 5.4 percent in 2021 (IMF 2021, 2022). In 2022 and 2023, the economy is expected to grow by an estimated 3.6 and 2.8 percent in real terms, respectively (IMF 2022). The last actual measurement of poverty for Ghana was in 2016/17, based on the Ghana Living Standards Survey. As noted earlier, the estimated poverty rate for that year was 12.7 percent, based on the international poverty line of $1.90 per day in 2011 purchasing power parity dollars. Using the same poverty line, annual poverty rates predicted as a function of actual GDP per capita in constant cedis are available for Ghana through 2021 (World Bank 2022). The pattern of evolution in these predicted poverty rates consists of a steady decline through the late 2010s, a slight increase during the first year of the pandemic, from 8.4 percent in 2019 to 9.2 percent in 2020, and then down again (World Bank 2022). Of course, the counterfactual for the effect of the pandemic on poverty may have been a monotonic downward trajectory over the period. 26 2.8. Fiscal position Figure 12 shows total government expenditures, revenues (including grants), and debt as shares of GDP since 2010; it also shows debt service as a share of total government expenditures for the same period. Expenditures have consistently exceeded revenues, often by large amounts and especially in election years (figure 12a). 7 The fiscal deficit averaged 4.8 percent of GDP from 2010 to 2019, then rose sharply to 14.7 percent with the onset of the pandemic. In 2021, the deficit declined to 11.4 percent of GDP, still much larger than prepandemic levels. Although smaller than in 2011, the deficit in 2022 remained high compared to the average level in the 2010s. As a consequence of persistent deficits, the government’s debt position worsened steadily throughout the 2010s and dramatically with the pandemic. General government debt as a share of GDP rose steadily from 2010 to 2019, though external public debt as a share of GDP stabilized from 2015 to 2019 (figure 12b). Instead, the government financed its spending with increasing amounts of domestic debt. The large pandemic-era deficits have been funded both internally and externally, with Ghana’s overall debt level rising to concerning levels in 2022. A further consequence of Ghana’s persistent deficits is a steady rise in debt service. Interest payments as a share of total government expenditure grew by 10 percentage points between 2010 and 2014 and were then on a roughly stable trend until the advent of the pandemic (figure 12c). Interest payments rose sharply in 2021 and 2022 in the wake of the heavy pandemic-era borrowing, a development that is expected to reduce the fiscal space for public spending on the social protection programs reviewed in this study. 7 Younger (2016) discusses political business cycles in Ghana. 27 Figure 12. Fiscal Position a. Government revenue and b. Government debt expenditure c. Debt service (interest payments as a percent of government expenditure) Source: Statistics obtained from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows trends in government revenue, expenditure, and debt as a percentage of GDP; it also shows the trend in government debt service, as a percentage of government expenditure. In the x-axis labels, “F” indicates a forecasted value. 2.9. Social protection Here, we present the basic contours, composition, and legal underpinnings of the government’s social protection programs. The basis for the public provision of social protection derives from the Constitution of the Fourth Republic of Ghana, specifically the Directive Principles of State Policy. Social protection is covered under the National Social Protection Policy introduced in 2015 (MOGCSP 2015). This policy follows the Social Protection Strategy introduced in 2007 and revised in 2012. The policy defines social protection to be “a range of actions carried out by the state and other parties in response to vulnerability and poverty, which seek to guarantee relief to those sections of the population who for any reason are not able to provide for themselves.” The policy presents the notion of a “social protection floor,” which is composed of access to basic 28 healthcare for all as well as minimum income security (i) to meet the basic needs of children, (ii) for people of working age, and (iii) for the elderly. The policy also specifies three (partially overlapping) target groups for its services and programs, namely the chronic poor, the economically at risk, and the socially vulnerable. And it lays out incremental steps the country should take over a 15- year period towards achieving universal social protection. While the policy puts an emphasis on rights protection and social services, followed by social safety net (or social assistance) programs, it gives marginal attention to the social insurance agenda and associated programs. The national social protection policy assigns the Ministry of Gender, Children, and Social Protection (MOGCSP) primary responsibility for implementing the policy and coordinating activities and actors in the social protection space. MOGCSP was established in 2013, succeeding the Ministry of Women and Children’s Affairs. Key government partners for implementing the social protection policy comprise the Ministries of Education; Employment and Labour Relations; Health; Finance; Food and Agriculture; and Local Government, Decentralization, and Rural Development. The policy underscores the need for decentralized delivery of the main social protection programs, with the involvement of the Local Government Service; metropolitan, municipal, and district assemblies; regional coordinating councils (through regional planning coordinating units); district social protection committees; and community social protection committees. The social protection policy acknowledges multiple other policies pertinent for the social protection agenda, including the National Gender Policy (introduced in 2014), the National Youth Policy (2010), the National Ageing Policy (2010), the National Local Economic Development Policy (2013), and the National Employment Policy (2015). The National Labor-Intensive Public Works Policy, introduced in 2016, draws substantive links with the national social protection, employment, and local economic development policies (MELR and MLGRD 2016). The national social protection policy considers the LEAP program, the LIPW program, and GSFP as main (or “flagship”) public social protection programs. It also characterizes registration, renewal, or premium payment exemptions for NHIS participation received by certain population groups such as the “indigent” or pregnant women as a main public social protection measure (not NHIS in its entirety). Finally, the policy denotes capitation grants, per-student subsidies, provided to government basic schools as another public main social protection measure. Capitation grants are mainly used to help defray the cost of teaching and learning materials. The policy presents these five interventions as Ghana’s social protection “basket.” The SSNIT pension scheme is not classified as a key social protection program, although the national social protection policy recognizes the program’s relevance in the social insurance area. For the purposes of our performance review, along with SSNIT pensions, we examine all the interventions in the policy’s social protection basket except for capitation grants for basic schools as we consider this measure as much more of a standard public education 29 intervention than a public social protection one. None of the main public social assistance programs—the LEAP program, the LIPW program, or GSFP—is covered by legal provisions. With respect to social insurance and social security, NHIS and SSNIT pensions are covered by legal provisions. Specifically, NHIS was established through the National Health Insurance Act of 2003 (Act 650); the program is currently covered by the National Health Insurance Act of 2012 (Act 852). The current SSNIT pensions schemes were established through the National Pensions Act of 2008 (Act 766). Additionally, the government requires that employers provide for sickness and maternity benefits through the Labor Act of 2003 (Act 651) and the National Health Insurance Act of 2012, respectively. For work-related injuries, the government requires employers to provide, as applicable, disability, medical care, survivor, or funeral benefits. These benefits are mandated through the Workmen’s Compensation Act of 1987 (Act 187). Several international organizations have provided or currently provide technical and financial assistance to the government for its rights protection and social welfare services and for its social protection programs. These organizations include the Africa Development Bank (AfDB), the European Union (EU), the International Labour Organization (ILO), the United Kingdom Foreign, Commonwealth, and Development Office or UK FCDO (formerly the United Kingdom Department for International Development or UK DFID), the United Nations Children’s Fund (UNICEF), the United Nations Development Programme (UNDP), the United Nations High Commissioner for Refugees (UNHCR), the United Nations Population Fund (UNFPA), the United States Agency for International Development (USAID), the World Bank, and WFP. 30 3. PROGRAMS In this section, we provide an overview of the government of Ghana’s main social protection programs: the Livelihood Empowerment Against Poverty (LEAP) program, the Labor-Intensive Public Works (LIPW) program, the Ghana School Feeding Programme (GSFP), Social Security and National Insurance Trust (SSNIT) pensions, and the National Health Insurance Scheme (NHIS). The LEAP program, the LIPW program, and GSFP are social assistance programs, while SSNIT pensions and NHIS are social insurance programs. The overview discusses key program design and implementation parameters. It also discusses findings for the programs from available operational and beneficiary assessments and impact evaluations. The government also administers another social assistance program, called the Complementary Livelihood and Asset Support Scheme (CLASS). Targeting LEAP and LIPW program beneficiary households, CLASS provides selected household members (one member per household) training, small capital grants, and post-training mentoring and coaching to help improve the performance of beneficiaries’ farm and nonfarm income-generating activities. A pilot of the program was administered between 2014 and 2018, covering around 7,100 individuals in eight districts in the Upper East region. CLASS was launched in 2020, with the plan to reach different sets of communities in rounds, in five regions in the north of the country (Upper East, Upper West, North East, Northern and Savannah). The program is currently in its second round of implementation and remains small in scale. It covered about 9,000 individuals under the first round and is currently covering about 11,900 individuals under the second round. Given CLASS’s small scale and the absence of relevant data and rigorous research to assess its performance, we do not discuss the program in this study. With respect to the LEAP program, the LIPW program, GSFP, NHIS, and SSNIT pensions, we discuss recent trends and current patterns in selected key program indicators. We also specifically discuss levels and patterns in key program indicators in 2017, which overlaps with the reference period for our performance review of the programs based on GLSS 2016/17, discussed in section 4. 3.1. Livelihood Empowerment Against Poverty Program Following the recommendation made in the National Social Protection Strategy of 2007, the government introduced the Livelihood Empowerment Against Poverty (LEAP) program in 2008 with the main aim of reducing household poverty and vulnerability. To this end, the program provides unconditional cash transfers to extreme-poor households in impoverished geographic areas. Overall responsibility for the program currently lies with the LEAP Management Secretariat (LMS) under the Ministry of Gender, Children, and 31 Social Protection (MOGCSP). 8 Responsibility for ground-level implementation lies with district officials, community leaders and volunteers, and contracted private parties. Key organizational structures and personnel at the ground level include District LEAP Implementation Committees (DLICs), Community LEAP Implementation Committees (CLICs), District Social Welfare Officers (DSWOs), and Beneficiary Welfare Associations (BWAs). Several international organizations have provided financial and technical assistance for the conceptualization, design, and implementation of the program at different points in time and over different durations. The main organizations include EU, UK FCDO, UNICEF, USAID, and the World Bank. Launched as a pilot in February 2008, the program initially covered 1,700 extreme-poor households with orphans and vulnerable children (orphaned or made vulnerable by HIV/AIDS) ages 0-15 years in 21 districts across all regions in the country. 9 In 2009, the program was extended to extreme-poor households with elderly persons (those age 65 or older) without economic support and household with persons with severe disabilities without productive capacity. Since then, the program has been incrementally scaled out to cover extreme- poor households with “eligible” members in at least one of three sociodemographic categories: (1) orphans and vulnerable children (“OVC”)10 (below age 18), (2) individuals age 65+, and (3) individuals with severe disabilities. In 2015, MOGCSP introduced a sister cash transfer program called LEAP 1000, with the official aim of financially supporting the health and nutrition status of pregnant and lactating mothers and young children. 11 In that year, the program reached 6,100 extreme-poor households with pregnant women, mothers of children below 12 months of age, and children below 12 months of age in 10 districts in the Northern and Upper East regions. LEAP 1000 was subsumed within the larger LEAP program soon after, in 2016, adding a fourth sociodemographic category to the program. At the end of 2021, the LEAP program covered 344,000 households in all districts in Ghana, equivalent to 6.6 percent of the national population. 12 Figure 13 shows the trend in program household numbers from 2008, when the program was initiated, through 2021. Figure 14 maps the distribution of program households across districts between 2016 and 2021. Over the four-year 8 The LEAP Management Secretariat was established in 2015. Prior to that, overall responsibility for the program lay with a unit within the Department of Social Welfare under the former Ministry of Employment and Social Welfare (MOESW). 9 The program started off by targeting needy households with orphans and vulnerable children because of a preexisting program in the pilot districts by the MOESW where premiums for participation in the National Health Insurance Scheme were waived for orphans and vulnerable children. 10 A child in a needy household has to be classified as both an orphan (the death of one or both parents) and vulnerable to be considered as eligible. 11 The name of the program refers to the first 1,000 days of life, a period considered critical for health and nutrition-enhancing inputs for the child’s long-term growth and development. The design and introduction of LEAP 1000 was supported by UNICEF and USAID. 12 To convert the number of households to number of individuals, an average household size of 6.3 members in LEAP program households estimated using GLSS 2016/17 data was used. 32 period from 2018 to 2021, more than 50 percent of program households were from districts in three out of the country’s 10 regions (regional classification per the 2010 population and housing census): Northern, Upper East, and Upper West regions. In 2017, the LEAP program covered 197,000 households, equivalent to 4.1 percent of the national population. Figure 13. LEAP Program Household Numbers Source: LEAP program administrative data obtained from MOGCSP. Note: Figure shows the trend in the number of LEAP program households. LEAP = Livelihood Empowerment Against Poverty. Program targeting was based on a four-level process. These levels were region, district, community, and household. The targeting rules sought to hone in on the poorest locations in the country. MOGCSP selected the regions and districts for the program, based on region- and district-level poverty statistics obtained from the Ghana Statistical Service (GSS). Communities within districts were selected by DLICs following guidelines provided by MOGCSP. These guidelines called for a qualitative determination of the poverty status of the community using proxy indicators such as the availability of various infrastructure, amenities, and services; school participation and health outcomes; and dwelling quality of residents. The District LEAP Implementation Committees communicated to MOGCSP the communities they had selected for the LEAP program. 33 Figure 14. Distribution of LEAP Program Households, by District, 2016–21 Source: LEAP program administrative data obtained from MOGCSP. Note: The maps show the distribution of LEAP program households across districts in Ghana, in each year between 2016 and 2021. The program household numbers correspond to the numbers that were part of the last LEAP program payment round in each year. LEAP = Livelihood Empowerment Against Poverty. 34 Within communities selected for the program, the household-selection approach has evolved over time. From program inception until 2015, community-based selection was used to identify households for the program. The Community LEAP Implementation Committees identified extreme-poor households that met the categorical criteria for program eligibility. In 2010, MOGCSP introduced a proxy means test (PMT) criterion for program eligibility with the aim of better identifying extreme-poor households for the program. Using relevant data gathered from community-selected households through a questionnaire, MOGCSP estimated the household’s PMT score. 13 Community- selected households that had a PMT score that fell below predefined cutoffs were selected by MOGCSP for the program, and this information was communicated to DLICs and CLICs. The Community LEAP Implementation Committees then informed and registered the households selected for entry into the program. MOGCSP launched a national social registry (called the Ghana National Household Registry) in October 2015, and is in the process of rolling it out across regions of the country. This national social registry is expected to serve as the main source of data for the proxy means test (MOGCSP 2019). MOGCSP stipulates that the eligibility status of existing LEAP program households should be reassessed and recertified at least once every two to four years (MOGCSP 2019). Program households that no longer meet the PMT score criteria in the reassessment are to be deemed ineligible, even if the PMT was not originally applied in assessing the household for entry into the program. MOGCSP recently initiated its first reassessment and recertification exercise for program households. LEAP program benefit amounts were originally set as a function of eligible members of the household. Eligible members are those individuals who fall under the sociodemographic categories for program eligibility. Based on the latest benefit amount (last revised in September 2015), households with one eligible member received GH¢32 per month; two eligible members, GH¢38 per month; three eligible members, GH¢44 per month; and four or more eligible members, GH¢53 per month. In the last few years, MOGCSP has dropped the sociodemographic categorical targeting of households for the program. Accordingly, benefit amounts are now set as a function of household members (instead of eligible household members), while all other benefit structure parameters are maintained unchanged (MOCGSP 2019). To date, there has been no significant intake of new households into the program where this benefit structure has been applied. A significant intake is expected after the first reassessment and recertification exercise is completed. Benefit amounts are not automatically adjusted for price inflation, but are revised upward over time in irregular intervals based on a deliberative process undertaken by MOGCSP. Since the start of the program through December 2022, benefit amounts have been raised twice: in July 2013 (roughly 5-1/12 years after 13The original PMT model used by MOGCSP was constructed based on GLSS 2005/06 data. The PMT model currently in use is based on GLSS 2016/17 data. 35 program inception) and in September 2015 (after roughly another two years). The government recently announced that the benefit amounts will be doubled starting from fiscal year 2023 (Government of Ghana 2022); the monthly benefit amount will now range between GH¢64 and GH¢106. Figure 15a presents the trend in nominal benefit amounts from program inception (February 2008) through August 2022, whereas figure 15b presents the trend in real benefit amounts over the same period. From September 2015 (when benefit levels were last revised) to August 2022, the real value has fallen by 60 percent. According to program administrative data, the average annual benefit was GH¢471 in 2021 and 2022. In 2017, the corresponding average was GH¢469. Between program inception and December 2015, Ghana Post served as the payment service provider for the program, contracted by the former Ministry of Employment and Social Welfare. Ghana Post provided cash benefits to beneficiaries or their caregivers at payment points in or near LEAP program communities. To receive their benefits at payment points, beneficiaries or their caregivers presented their LEAP program (or other official) identification card and provided their thumbprint for eligibility verification and payment confirmation. After a short pilot in two phases (in November–December 2013 and January 2014), the benefit payment system was changed in December 2015 to delivery through banks with local branches in LEAP program districts contracted by the Ghana Interbank Payment and Settlement Systems (GhIPSS), a payment service provider. 14 GhIPSS was in turn contracted by MOGCSP for the program. Under this current system, program beneficiaries received biometric smart cards, called “e-zwich” cards. Partnering banks adopted the payment points previously used by Ghana Post, or established new ones where necessary. Program beneficiaries or their caregivers could collect their benefits at payment points, after verification of eligibility using their e-zwich cards. Alternatively, program beneficiaries or their caregivers could visit the partnering banks to collect their benefits following the same verification process. 14 GhIPSS is a subsidiary of the Bank of Ghana. 36 Figure 15. Nominal and Real Changes in Stipulated LEAP Program Benefit Levels a. Nominal b. Real Source: Program administration information obtained from MOGCSP. CPI statistics obtained from GSS. Note: Figure shows the trend in officially stipulated LEAP program benefit levels, in nominal and real terms. Real benefit levels are calculated using the national CPI series (base month and year is February 2008, when the LEAP program was launched). LEAP = Livelihood Empowerment Against Poverty. 37 The LEAP program stipulates a benefit payment periodicity of every two months (six payment rounds per year), following a fixed calendar. In each payment round, two months’ worth of benefits are to be delivered to program households. The transition from Ghana Post to GhIPSS and partnering banks produced a payment delivery gap spanning several months. Once the new payment system was fully operational at the needed scale, the payment periodicity improved greatly. From 2016 through 2018, the program was able to deliver six payment rounds (two months of benefits per payment round), even if the fixed calendar was not perfectly followed (figure 16). However, starting from late 2019 and through 2022, the program began to slip in its payment periodicity-it delivered fewer payment rounds and, at times, compensated by delivering four months of benefits in a payment round. The slip in payment periodicity was mainly due to delays in government budgetary releases for the program. Figure 16. Periodicity of LEAP Payment Rounds and Number of Monthly Benefits per Payment Round, 2016–22 Source: LEAP program administrative data obtained from MOGCSP. Note: Figure shows the periodicity of LEAP benefit payment rounds as well as the number of monthly benefits per payment round. Months in the x-axis are denoted numerically (i.e., 1 = January; 12 = December). LEAP = Livelihood Empowerment Against Poverty. While the LEAP program provides cash benefits to households without any conditionalities, it encourages certain coresponsibilities. These include, as relevant for the program household, enrollment and attendance of children ages 6–15 years at local schools; uptake of maternal, newborn, and child health and nutrition services at local health facilities; participation in locally provided financial and nutrition education interventions, and enrollment in the National Health Insurance Scheme and renewal of membership annually, under 38 membership categories that exempt the household from paying registration or renewal fees and premiums (MOGCSP 2019). The key organizational structures at the district and community levels for the program are expected to facilitate program households in meeting these coresponsibilities. An impact evaluation of the LEAP program finds that the intervention by and large did not have any positive effects on various socioeconomic indicators, assessed two and six years after baseline measurement (Handa et al. 2017). At baseline in 2010, the study matches a sample of prospective LEAP program households in the Brong-Ahafo, Central, and Volta regions to similar households interviewed in a separate national survey. 15 The program and counterfactual samples are then tracked over time, with follow-up data collected in 2012 and in 2016. Program impacts are estimated based on a difference-in- differences strategy. The authors of the study argue that the counterfactual group appears to be weak. Nevertheless, the evaluation finds that welfare improved dramatically over time among a sample LEAP program households. Further, consistent with the pattern in payment rounds discussed above, the evaluation finds that benefit payments became much more regular in the later evaluation period (2012–16) than the earlier period (2010–12), linking the pattern to the shift in the benefit payment system from Ghana Post to GhIPSS and banks in 2015. However, at the same time, average time to payment points and the likelihood of leakage at payment points increased while the likelihood of feeling secure at payment points decreased in the later evaluation period compared to the earlier period. The authors of the evaluation speculate that these issues may have been teething problems with the changeover to the new system. Another impact evaluation finds more favorable results for a certain variant of the program (LEAP 1000 Evaluation Team 2018). The research examines the pilot LEAP 1000 intervention which targeted pregnant women and mothers with young infants with cash transfers coupled with fee waivers for NHIS membership. As noted earlier, this pilot program and its beneficiaries were subsequently subsumed within the overall LEAP program. Exploiting the PMT cutoff for household entry into the pilot intervention, the evaluation compares outcomes between those households just below and just above the cutoff, using data collected in 2015 (baseline) and 2017 (follow-up). Those households below the cutoff received the intervention. The evaluation period spans 13 payment rounds. Program impacts are mixed. The evaluation finds significant positive impacts on multiple individual and household welfare measures, but not on several others. Further, most significant positive impacts are modest in size. In terms of consumption and poverty, average program impacts are GH¢9 for monthly overall household consumption per adult equivalent (in August 2017 prices), GH¢7 for monthly household food consumption per adult equivalent, –2.1 percentage points for the poverty rate, 15 Matching is performed based on propensity scores. 39 and –2.6 percentage points for the poverty gap. 16 The impacts are found over a period when consumption fell and poverty increased in the overall evaluation sample. The research however notes that program impacts are expected to be modest because the evaluation looks at the LEAP program households near the proxy means test cutoff (that is, relatively more well-off LEAP program participants) and not the more typical LEAP program household. Based on the same identification strategy and data noted above, studies have examined the impacts of the pilot LEAP 1000 intervention (cash transfers combined with fee waivers for NHIS membership) on intimate partner violence (Peterman, Valli, and Palermo 2022) and on NHIS membership (Palermo et al. 2019). The latter study finds significant, large positive impacts. Several other quantitative and qualitative studies have examined samples of LEAP program households (for example, FAO 2013 and CDD-Ghana 2016). Their findings on levels and changes in outcomes are consistent with the corresponding findings for LEAP program households from more rigorous evaluations (that is, those that incorporate a reasonable counterfactual group). Beneficiary feedback assessments indicate high satisfaction rates among LEAP program households (for example, P&OD Consult 2018). The introduction of the smart card payment system for beneficiaries appears to be an important factor behind this result. Interviews with program beneficiaries and administrators suggest that program implementation rules and procedures are generally well defined but that their application can fall short because of insufficient knowledge, competency, and capacity of local program administrators. A careful operational assessment conducted in 2012 indicates no widespread or severe issue with program implementation across the delivery chain (Handa et al. 2012). Beneficiary feedback and operational assessments also identify the need for greatly improved communication and awareness- raising efforts targeted toward program households, communities, and local program administrators on various aspects of program design and implementation. 16 The average baseline values for the comparison group were GH¢120 for monthly overall household consumption per adult equivalent, GH¢90 for monthly household food consumption per adult equivalent, 91 percent for the poverty rate, and 49 percent for the poverty gap. 40 3.2. Labor-Intensive Public Works Program The main aim of the Labor-Intensive Public Works (LIPW) program is to reduce poverty and vulnerability by offering rural, poor households temporary labor earnings opportunities during the agricultural off-season. These opportunities are engendered through public works employment activities that result in the creation, rehabilitation, or maintenance of public or community infrastructure (referred to by the authorities as community “assets”). The agricultural lean season runs from November to March or April. Typical public works activities (referred to by the authorities as “subprojects”) involve feeder roads, basic water resources infrastructure (small earth dams and dugouts), and soil and land conservation (such as work on seedling and plant nurseries and community tree planting). The LIPW program is covered under the National Labor-Intensive Public Works Policy (MELR and MLGRD 2016). The government approved the policy in August 2016 and it took effect in January 2017. Overall current responsibility for administering the program lies with the Rural Development Coordination Unit (RDCU) under the Ministry of Local Government, Decentralization, and Rural Development (MLGDRD). Current responsibility for administering the program on the ground lies with District Assemblies (DAs) with support from Zonal Coordinating Officers (ZCOs) of RDCU and community facilitators (who come from the community). Subprojects are either fully managed by communities, or DAs contract private contractors to manage them. Private contractors are usually contracted to manage road or water-resources infrastructure subprojects. The program is fully financed by the World Bank. The LIPW program was launched in 40 districts across all regions in 2011. Figure 17 plots the number of participants and subprojects per year from inception through 2021, whereas figure 18 plots the distribution of participants across districts for each of the years between 2016 and 2021. Individuals and subprojects are not necessarily unique across consecutive pairs of years as the works activity can span two years. In 2018, the program was implemented (that is, it engaged participants) for only part of the year; in 2019, it did not engage any participants; and, in 2020, program implementation with the engagement of participants restarted. The hiatus in the engagement of participants was connected to interrupted donor financing. 17 In 2021, the program supported 30,000 participants engaged in 352 subprojects in 80 districts across all regions. In 2017, it supported 32,900 participants engaged in 321 subprojects in 80 districts across all regions. 17 Donor financing was interrupted by a protracted design period for the World Bank-supported Ghana Productive Safety Net Project, which became effective in June 2019, following the closing of the World Bank-supported Ghana Social Opportunities Project in May 2018. 41 Figure 17. LIPW Program Beneficiary and Subproject Numbers a. Beneficiary individuals b. Subprojects Source: LIPW program administrative data obtained from MLGDRD. Note: Figure shows the trend in the number of LIPW program beneficiary individuals and of LIPW subprojects. The LIPW program was only partly active in 2018 and it was fully inactive in 2019, due to an interruption in donor financing. Program beneficiary individuals and subprojects are not necessarily unique across consecutive year pairs. LIPW = Labor Intensive Public Works. 42 Figure 18. Distribution of LIPW Program Individuals, by District, 2016–21 Source: LIPW program administrative data obtained from MLGDRD. Note: The maps show the distribution of LIPW program beneficiary individuals across districts in Ghana in various years. The map for 2019 is blank because there were no LIPW program beneficiary individuals that year due to the lack of program financing. LIPW = Labor Intensive Public Works. 43 Program targeting for each agricultural off-season follows a three-level process. First, MLGDRD selects districts with the worst levels of poverty based on data provided by the Ghana Statistical Service (GSS). Second, the poorest communities within the selected districts are identified jointly by DAs and RDCU using information from GSS, District Medium-Term Development Plans, and other sources. In selecting communities, DAs and RDCU prioritize existing LEAP program communities, assess the willingness and interest of the community to participate in the LIPW program, and assess whether a subproject proposed by the community qualifies. For a subproject to ultimately qualify, among other criteria, the percentage of the subproject’s total budget for labor costs should exceed the stipulated minimum, which varies by the nature of the subproject. The selected subproject should also preferably last for two agricultural off-seasons. Third, within a selected community, a combination of individual self-selection and community selection is used to identify participants for the LIPW program. District Assemblies and community facilitators for the program make households aware of the program along with expectations and rules for participation. Individuals between ages 18 and 65 (adults) can seek to participate in a subproject by registering with the community facilitator. If the number of registered individuals exceeds the labor needs for the subproject, guidelines indicate that the poorest individuals among those registered are to be selected by the community, led by the community facilitator and two community leaders – this meant starting off with selecting registered participants from LEAP program households. The list of selected participants for the subproject is to be made publicly available at common gathering places in the community, and complaints can be made on the selection process or outcome through several channels. Program guidelines call for at least 60 percent of program participants selected by DAs and community facilitators to be women, to be achieved through community sensitization, outreach, and conducive work activities, working conditions and hours, and amenities. Administrative data on a sample of individual LIPW program participants in 2016 and 2017 indicate that 57 percent were women. The average age of participants was roughly 39 years. The average age of female participants was slightly lower than for male participants, by less than a year. While the stipulated minimum age of 18 years for program participation appears to have been consistently applied, a small percentage (less than 5 percent) of participants were older than 65 years, the stipulated maximum age. Qualifying participants can work a maximum of 200 days per year during the agricultural off-season, for up to two consecutive seasons, but this amount of work is not guaranteed to every participant and in every subproject. Each qualifying household can register two capable adult members to work, a primary and a secondary participant, in case the primary participant is unavailable to work on a given day. Only the primary participant is eligible to collect program payments. The daily program wage benefit in a given year is pegged to the official daily minimum wage (rounded up to the nearest integer) for an eight-hour working 44 day in that year. The official daily minimum wage is adjusted yearly through a national tripartite (government, organized labor, employers) deliberative process, which considers increases in the cost of living. Figure 19a shows the trend in the stipulated daily minimum wage between 2011 and 2022. LIPW program beneficiaries are, however, only required to work a maximum of six hours per day (but are paid for an eight-hour work day). In 2022, the daily minimum wage was GH¢13.5. In 2017, the daily minimum wage was GH¢8.8 (the same wage for both sexes). This minimum wage level was 21 percent of average daily wage earnings for rural male wage workers ages 18–65 and 56 percent of average daily wages for female counterparts. 18 Figure 19b shows the trends in nominal and real average benefit levels between 2011 and 2021, based on program administrative data. In 2021, the average total wage payment per program participant was GH¢1,176. In 2017, the average was GH¢498. LIPW program participation and benefit data were not gathered in the GLSS 2016/17. Consequently, we cannot estimate average LIPW program benefits relative to household consumption or to the country’s poverty lines among LIPW program households, as we do for the other programs in section 4.3. However, based on a survey conducted in November–December 2015 which included a sample of LIPW program beneficiaries, Osei-Akoto et al. (2016) report an average total wage payment per program participant of GH¢455 in the north of the country and GH¢828 in the south. They also report that the average LIPW program benefit relative to household consumption was 25.9 percent for extreme-poor program households, with 87 percent of program households determined to be extreme poor. Since 2016, program participants have been paid monthly using the same payment system as the LEAP program. 19 The Rural Development Coordination Unit has contracted GhIPSS, which, in turn, has contracted financial institutions to administer payment points in or near LIPW program communities. Biometric smart cards were provided to primary program participants to use at payment points. The authorized payment amount is based on information on the number of days worked by the participant captured in the program’s DASH (Daily Attendance Sheet) system and transmitted by DAs to GhPISS via RDCU. 18 Average daily wage earnings are estimated based on GLSS 2016/17 data. 19 See the discussion of GhIPSS and e-zwich cards in section 3.1. 45 Figure 19. Stipulated Minimum Wage and Average LIPW Program Wage Payment Levels a. Stipulated daily minimum wage b. Nominal and real average total wage payments Source: LIPW program administrative information obtained from MLGDRD. Annual national CPI data obtained from the World Bank’s World Development Indicators databank. Note: Figure shows the trend in the officially stipulated daily minimum wage and in the nominal and real average total wage payments received by LIPW program beneficiary individuals. The LIPW program was only partly active in 2018 and it was fully inactive in 2019 due to an interruption in donor financing. Real average wage payments were calculated by using annual national CPI data (base year = 2010). LIPW = Labor Intensive Public Works. 46 Figure 20 shows the results of LIPW works subprojects in each year. Between 2011 and 2020, the LIPW program yielded 1,603 kilometers of rural roads, 37.3 million cubic meters of dam and dugout volume, and 4,616 hectares of afforestation and reforestation. The effects of the LIPW program on individual and household outcomes have been evaluated (Osei-Akoto et al. 2016). The evaluation is based on a randomized phase-in design, in which works subprojects are randomly assigned into two implementation phases, the first phase in 2015 and the second phase in 2016. LIPW program beneficiary households in a sample of phase-1 program communities (stratified by the type of subproject) are compared to statistically matched households in a sample of phase-2 communities after the completion of the first phase of implementation. Based on a cross-sectional analysis using follow-up household survey data in November–December 2015, the study finds that the program increased individual employment and earnings outcomes, measured variously. These results are consistent across the different types of subprojects. Depending on the type of subproject and outcome indicator, the study also finds mixed evidence that the program influenced household crop output and sales; nonfarm enterprise activity; borrowing, lending, and savings; migration; income and consumption; and the use of health and education services. The evaluation data are also used to conduct an operational review and an assessment of beneficiary satisfaction (Osei-Akoto et al. 2016). For this, the follow-up survey in November–December 2015 is complemented by information from focus group discussions conducted earlier in July–August 2015, closer to the end of phase-1 implementation of the LIPW works subprojects. While program beneficiaries generally report program activities and aspects such as awareness raising, registration, household selection, the wait time from registration to actual participation, and payment mode to be satisfactory, large shares of households report delays in benefit payments (the stipulated payment period was every two weeks during the period of the study data) and conflict between days and hours (within a day) for LIPW works activities and for their agricultural activities (LIPW works activities are supposed to occur during the agricultural off-months). 47 Figure 20. Civil and Environmental Works Results from LIPW Program Subprojects a. Small dam and dugouts, capacity, b. Roads, length volume c. Afforestation and reforestation, area Source: LIPW program administrative data obtained from MLGDRD. Note: Figure shows the trends in the civil, conservation, and environmental works results of LIPW subprojects, by type of subproject. The LIPW program was only partly active in 2018 and it was fully inactive in 2019, due to an interruption in donor financing. LIPW = Labor Intensive Public Works. 3.3. Ghana School Feeding Programme According to the country’s National School Feeding Policy (approved in and effective since 2016), the Ghana School Feeding Programme (GSFP) primarily seeks to increase school enrollment, attendance, and retention in poor, rural communities; improve the nutrition status of young, school-age children; encourage local food production and consumption; and increase rural income (MOGCSP 2016). Specifically, the program aims to provide one cooked, nutritious meal every school day to each student in public preprimary, primary, and special education schools in poor, rural communities covered by the program. Meals are to be 48 provided by local private caterers contracted by the government for the program. Caterers are expected to purchase most of their foodstuff for the program from local farmers. The Ghana School Feeding Programme was launched in late 2005. From 2005 to 2014, primary overall responsibility of the program lay with the former Ministry of Local Government and Rural Development 20 through its National School Feeding Secretariat. In 2015, this responsibility was assumed by MOGCSP, through the establishment of its own National School Feeding Secretariat. The main actors engaged in implementation are Regional Coordination Offices (RCOs), District Assemblies (DAs), District Implementation Committees (DICs), School Implementation Committees (SICs), caterers, and cooks. RCOs are expected to perform a liaison function for the program between the National School Feeding Secretariat and DAs, along with reporting and monitoring functions. District Assemblies, through their DICs, have primary responsibility for ground-level implementation of the program. Program implementation in communities and schools is supported by SICs composed of local community members and school staff. Importantly, adopting a “caterer model” for the program, District Assemblies contract private caterers to procure food, prepare meals, and serve students at program schools. These private caterers typically come from the program communities; they also hire cooks from the program communities. Various international organizations have provided financial and technical assistance to the program at different points during conceptualization and implementation. These organizations, importantly, include UN FAO, the Partnership for Child Development (PCD), the Netherlands Embassy, SNV Netherlands Development Organization, UNICEF, USAID, WFP, and the World Bank. Several civil society organizations have also supported DICs, SICs, schools, and caterers to perform their roles. Since 2011, the program has been fully financed by the government, through general revenues. The program was launched as a pilot in 2005, covering approximately 1,900 students in 10 schools, one in each region (region classification per Ghana’s 2010 population and housing census). Since then, the program has been rapidly scaled up. In academic year 2016/17 (which overlapped with the fielding of GLSS 2016/17), the program covered 1.67 million students in 5,682 schools served by 4,975 caterers, across all regions. In academic year 2018/19 (the most recent year before program implementation was disrupted by the pandemic), the program covered 2.94 million students in 9,162 schools served by 9,561 caterers. Figure 21 maps the distribution of program students across regions for academic years 2016/17 through 2018/19. The Ashanti region is a positive outlier in terms of the share of program schools and students. 20 Now, the Ministry of Local Government, Decentralization, and Rural Development. 49 Figure 21. Distribution of GSFP Beneficiaries, by Region, 2016/17–2018/19 Source: Program administrative data obtained from MOGCSP. Note: Maps show the distribution of GSFP individuals across regions in various academic years. For these statistics, the academic year runs from September through July. The classification of regions is per Ghana’s 2010 Population and Housing Census. GSFP = Ghana School Feeding Programme. Program targeting is at the community level. District Assemblies have primary responsibility for the assessment and selection of public schools and communities covered under the program. Community selection for the program is expected to be based on criteria stipulated by MOGCSP. These criteria include low levels of schooling, high gender disparities in schooling, low adult literacy levels, high poverty levels, and low levels of access to potable water. The assessment also includes criteria such as strong intracommunity relations and management capacity in the prospective community, willingness of the prospective community to provide cash or in-kind contributions for basic amenities for the program (e.g., use of kitchen or storeroom), and the absence of any other feeding program in prospective schools. Student enrollment numbers for the program are transmitted by the schools and the DAs to the National School Feeding Secretariat (Dunaev and Corona 2019). Caterers contracted by the DAs are required to follow menus designed by the National School Feeding Secretariat in consultation with the DAs. The meals under these menus are formulated to provide at least 30 percent of the recommended dietary allowance of essential nutrients on a daily basis (Dunaev and Corona 2019). To cover the costs of food procurement, preparation, and service, caterers receive a payment amount per student per meal. This payment amount is not automatically adjusted over time for price inflation. In academic years 2015/16 and 2016/17, caterers were paid GH¢0.8 per student per meal. Starting in the 50 second term of the academic year 2017/18, the payment amount was raised to GH¢1. The government recently announced that the payment amount will be increased starting from the 2023 academic year (Government of Ghana 2022). 21 The actual per-student expense would have to be lower than the payment amount received by the caterer if the caterer is to earn any profit. However, independent costing of the menus using local prices indicates that school meals would cost more than the payment amounts received by caterers if the menus were strictly adhered to. Indeed, large shares of caterers report that their expenses exceed their payments. The contract that caterers sign stipulates that they should spend at least 60 percent of their payment amount on expenses. Available evidence suggests that the purchase of food accounts for on average 80 percent of total expenses by caterers, while transportation of food, payment of the salaries of cooks, and the purchase of water and cooking fuel account for the remainder (Dunaev and Corona 2019). Available evidence also suggests caterers often face chronic and long delays in receiving their payments. Caterers are contractually obligated to prefinance meals for one academic term, but the delays tend to extend for several months into the subsequent term and even multiple subsequent terms. The low payment amount (compared to caterers’ expenses) and the long delay in receiving payments appear to have led caterers to reduce the frequency, size, and nutritional value of meals (Dunaev and Corona 2019). GSFP has been rigorously evaluated (Gelli et al. 2019; Aurino et al. 2023). Under the research, from a sample of 58 districts, two communities in each district were randomly assigned into either program or nonprogram status. Assessed in 2016, more than two academic years after baseline measurement in 2013, the research finds that GSFP improved academic achievement among girls, children from poor households, and children from the northern regions (these regions had better program implementation than elsewhere). It also finds that the program improved nutritional status (specifically, standardized height- for-age) among girls, younger children, and younger children from poor households. These positive program impacts are found despite significant implementation deficiencies due to delayed and insufficient payments from the government to caterers, as pointed out in the research. 21The academic year was changed from September through July to January through December starting from 2021. 51 3.4. Social Security and National Insurance Trust Pension Scheme Since 2010, Ghana has had a three-tiered pension scheme. The first tier is the National Social Security Scheme, a defined benefit pension scheme. This scheme is the focus of this discussion. The second tier is a defined contribution pension scheme managed by various private trustees. The third tier includes all other pension and provident funds. The first two tiers are mandatory for almost all workers in the formal sector and voluntary for those in the informal sector. The third tier is voluntary for all. The Social Security National Income Trust (SSNIT) is a public trust whose main purpose is to manage the first tier of the pension scheme. SSNIT is governed by a board of directors that includes representatives from the President’s Office, organized labor, the Ghana Employers’ Association, the National Pensioners’ Association, the Ministry of Finance, and the security services (excluding the military). Operations are managed by an Executive Committee comprised of full-time SSNIT employees. International organizations that have provided technical assistance for SSNIT pensions include ILO and the World Bank. All workers in Ghana are eligible to join SSNIT. In the formal sector, employers are required to register their employees and make the stipulated contributions on their behalf. In the informal sector, registration is voluntary and rare, as we discuss below. Some workers are exempt from joining SSNIT, if they belong to other pension schemes in its stead. These include employees in the armed forces, police, the Prisons Service, and the National Fire Service, and a few very senior civil servants stipulated in the Constitution. Figure 22 shows the total number of active contributors between 2005 and 2020; it also shows total active contributors as a share of the total labor force. Over this period, the number of active contributors increased from 0.9 million to 1.6 million, with an average annual growth of 4 percent and a cumulative growth of 82 percent for the period. The program participation share of the total labor force has also increased from 8–9 percent during the earlier years of the period to 10–12 percent during the latter years. Notwithstanding, the shares are low because the formal sector is small in Ghana. 22 22 Based on the GLSS 2016/17 data, 25 percent of workers were wage employed, and, of them, 44 percent had a written contract and 35 percent had employer-provided social security (Nxumalo and Raju 2020). Using these conditions (written contract, social security coverage) as indicators of formal employment status, roughly 10 percent of workers were formally employed in Ghana. 52 Figure 22. Active Contributor Levels, SSNIT a. Active contributors b. Active contributors as a percentage of the labor force Source: Program administrative statistics obtained from SSNIT. Labor force estimates obtained from the World Bank’s World Development Indicators databank. Note: Figure shows the trend in the number of active contributors to the SSNIT pension scheme and in the share of active contributors in the overall labor force. SSNIT = Social Security and National Insurance Trust. 53 The service sector has the overwhelming majority of active contributors. In 2020, 86 percent of active contributors were involved in services. This includes many public employees (civil servants, teachers, medical staff), but also financial institutions, most of which are in the formal sector. The percentage of active contributors from the services sector has increased steadily between 2009 and 2020 from the low 70s in the earlier years of the period. The majority of active members are also from the formal private sector. In 2020, 64 percent were from the formal private sector, 35 percent from the public sector, and less than 1 percent from the informal private sector. The distribution has largely remained stable between 2009 and 2020. Workers contribute 5.5 percent of their basic salary (withheld by the employer), while employers contribute the equivalent of 13 percent of each worker’s salary. The self-employed must contribute the entire 18.5 percent. Of this 18.5 percent, 11 percent goes to SSNIT, 5 percent goes to the tier two defined contribution pension scheme, and 2.5 percent goes to the National Health Insurance Fund (NHIS) as premia for SSNIT members (contributors and pensioners) who are automatically enrolled in NHIS. The Social Security and National Insurance Trust offers four types of benefits: retirement pensions, invalidity pensions, survivor’s pension (paid as a lump sum), and an emigration benefit for retirees leaving Ghana. To qualify for a retirement pension, the member must be 60 years old or older and have made contributions for 15 years if she or he joined SSNIT after 2008 and 20 years if she or he joined before. The amount of the monthly pension, payable until death, is the average of the employee’s best 36 months of salary times a “pension right.” For those who joined the system before 2008, the pension right is 37.5 percent for the first (required) 180 months of contributions and an additional 1.125 percent for each additional year of contributions. For those who joined after, the pension right is 2.5 percent per year for the first 15 years of contributions and 1.125 percent for each additional year, up to a maximum of 60 percent. There is an early retirement option for those 55 to 60 years old. The requirements are the same, including the minimum number of contributions, but the amount of the monthly pension is reduced by 60 percent for those age 55 up to 90 percent for those age 59. For a member who does not qualify for a retirement pension, her or his contributions are refunded with interest upon retirement. To qualify for an invalidity pension, a member must have made contributions in at least 12 of the 36 months preceding her or his invalidity, and must be permanently incapable of any normal gainful employment. The disability status must be confirmed by a medical officer and certified by a Regional Medical Board that includes a SSNIT Medical Officer. The amount of the monthly pension, payable until death, is calculated in the same way as the retirement pension, but the minimum age limit is not enforced. For those who have not made the minimum 15 years of contributions, the pension right is 37.5 percent. The survivors’ pension is paid to a member’s dependents on her or his death before age 75, whether or not the member is working at the time of death. If the 54 member has made 12 contributions in the 36 months preceding death, the survivor’s pension is a lump sum equal to the present value of 15 years’ pension as calculated above. When the member is already retired, the lump sum is the present value of the unexpired pension payments up to age 75. The emigration benefit is for non-Ghanaians only, and only for those who will leave Ghana permanent. It is a lump sum payment of whatever benefit is due to the member. For those who qualify for a pension, this is the present value of the member’s pension payments. For those who do not qualify for a pension, their contributions are returned with interest. Pension benefits are not indexed to inflation but adjusted at irregular intervals through a deliberative process undertaken by SSNIT linked in part to considerations around current and predicted future funds availability. Figure 23 shows the number of pensioners over the past decade. The number has increased steadily from 99,000 in 2009 to 227,000 in 2020, a growth of 8 percent per year and a cumulative growth of 131 percent over the period. In 2017, the number of pensioners was 190,000. Figure 23 also shows the average pension payment (monthly and lump sum payments combined) in nominal and real terms over the past decade. Both the nominal and real values have increased steadily between 2009 and 2020, for a cumulative increase of 542 percent and 89 percent, respectively. In 2020, the average nominal pension payment was GH¢14,524. The corresponding average nominal pension payment was GH¢11,551 in 2017. Because of the relative youth of the Ghanaian labor force, the country has many more working age contributors (more than 1.6 million in 2020) than pensioners (227,000 in 2020), so an important aspect of SSNIT’s activities is to invest the accumulated difference between contributions collected and pensions paid to guarantee SSNIT’s ability to pay pensions in the future. These investments include government debt, equity in listed and unlisted firms, real estate, and hospitals and health centers. Figure 24 shows the real rate of return on SSNIT’s portfolio over the past decade. While very strong until 2014, real returns have become more volatile and generally lower since, threatening SSNIT’s financial sustainability. An external actuarial review conducted in 2017 indicates that the “SSNIT scheme is not financially sustainable over the period covered by the projections from the report” (SSNIT 2019, p. 38). Nevertheless, the same review reports that at a hypothetical real rate of return of 4.25 percent (less than the past decade’s average real return), SSNIT has funds sufficient to last until 2038. 55 Figure 23. SSNIT Beneficiary (Pensioner) Numbers and Average Benefit Levels a. Pensioners b. Nominal and real average pension benefits Source: Program administrative statistics obtained from SSNIT. CPI statistics obtained from the World Bank’s World Development Indicators databank. Note: Figure shows the trend in the number of SSNIT pensioners and in the average pension benefits received by pensioners, in nominal and real terms. Real benefit levels are calculated by using the national CPI series (base year = 2010). SSNIT = Social Security and National Insurance Trust. 56 Figure 24. SSNIT Portfolio Performance, Actual Real Return Source: Program administrative statistics obtained from SSNIT. Note: Figure shows the trend in actual real returns from SSNIT’s investments of pension funds. SSNIT = Social Security and National Insurance Trust. Dorfman (2015) provides a general review of pension systems in Sub-Saharan Africa, which allows for a comparison of the SSNIT pension scheme to other pension schemes. Schwarz (2017) provides a critical review of Ghana’s pension system, including SSNIT pensions. Both studies show that the contribution rates of the SSNIT pension scheme are among the highest in Sub-Saharan Africa. Schwarz concludes that the SSNIT pension scheme has little room to increase contribution rates further. This is important because an external actuarial analysis performed in 2017 shows that the SSNIT pension scheme is not sustainable in the long run (SSNIT 2019), a conclusion reached by several earlier actuarial analyses of the scheme as well. SNNIT pensions are based on only the best 36 months of salary. Schwarz argues that this gives a strong incentive for employers to underreport salaries up to an employee’s last three years of work, showing how sharply reported salaries rise during employees’ last years of work. Increasing the number of years of salary that serve as the basis for pensions would reduce this incentive for underreporting and thus allow the SSNIT pension scheme to raise contribution revenue without increasing contribution rates. 57 3.5. National Health Insurance Scheme The National Health Insurance Scheme (NHIS) was passed into law in 2003 and rolled out across the country in 2005/06. The program allows for members to receive free healthcare services at participating facilities. NHIS is managed by the National Health Insurance Authority (NHIA), a public agency charged with overseeing both NHIS and private insurance schemes, but the latter are rare in Ghana. The Ministry of Health (MOH) sets broad policy guidelines for the National Health Insurance Scheme, but NHIA is responsible for their implementation. NHIA maintains regional, district, and metropolitan offices where members can register, healthcare providers can seek accreditation and file claims, and both can file complaints. Several international organizations have provided or currently provide technical and financial assistance to NHIS: Agence Française de Développement (AFD), the Global Fund, ILO, Japan International Cooperation Agency (JICA), Marie Stopes International (MSI), UK FCDO, USAID, and the World Bank. National Health Insurance Scheme membership is, in practice, voluntary. 23 Figure 25 shows the active membership size (measured at the end of the year) between 2010 and 2020. Active membership has increased (unevenly) from 7.4 million individuals in 2010 to 11.8 million individuals in 2020, a cumulative growth of 58 percent. Figure 25 also shows active membership as a share of population. Over the period, active membership has ranged between 30 percent and 40 percent of the national population. In 2017, active membership was 10.7 million individuals, equivalent to 35.3 percent of the national population. To obtain membership, individuals must pay a registration fee, which is currently GH¢8, or an annual renewal fee of GH¢5, and an annual premium of GH¢7 in rural areas and GH¢48 in urban areas. The premium is waived for several selected groups, including (i) those under 18 years, (ii) those aged 70 years or older, (iii) “indigents,” (iv) SSNIT contributors and pensioners, and (v) pregnant women and women with infants under three months (and the infants). The indigent, pregnant women, and women with young infants also do not have to pay the registration or renewal fee. NHIA defines an indigent as an individual who (i) “does not have any visible source of income,” (ii) “does not have a fixed place of residence,” (iii) “does not live with a person who is employed and has [a] fixed place of residence,” and (iv) “does not have a consistent source of support from another person.” NHIA currently relies on MOGCSP to identify indigent persons. 23 Since 2012, membership has technically been mandatory. But this is not enforced except for formal sector employees, whose employers are required to ensure that they are enrolled. 58 Figure 25. NHIS Beneficiary Levels a. Beneficiaries b. Beneficiaries as a percentage of population Source: Program administrative statistics obtained from NHIA. Population estimates obtained from the World Bank’s World Development Indicators databank. Note: Figure shows the trend in the number of NHIS beneficiary individuals and in the share of beneficiary individuals in the overall population. Beneficiary individual numbers correspond to end-of-year active membership numbers. NHIS = National Health Insurance Scheme. 59 In 2019, looking at end-of-year active membership, 34 percent were nonexempt members. About 46 percent of members were in the below-18 group. The other exempted groups each constituted 5 percent of total membership. There have been large swings between 2006 and 2019 in the member numbers (and hence the membership distributional shares) in the exempt groups of indigents and pregnant women and women with young infants. Figure 26 shows average nominal and real benefit levels over the last decade. For each year, average benefits are obtained by dividing reported annual expenditures on claims by the number of active members at the end of the year. Average nominal benefits rose (unevenly) from GH¢58 in 2010 to GH¢80 in 2019, a cumulative increase of 38 percent over the period. On the other hand, real benefits decreased over the period (a cumulative decline of 51 percent). In 2017, the average nominal benefit was GH¢124. Figure 26. Nominal and Real Average NHIS Benefits Source: Program administrative statistics obtained from NHIA. Annual national CPI statistics obtained from the World Bank’s World Development Indicators databank. Note: Figure shows the trend in average NHIS benefits received by beneficiary individuals, in nominal and real terms. For each year, the benefit level is calculated by dividing expenditures on claims by active membership at the end of the year. Real benefit levels are calculated by using the annual national CPI series (base year = 2010). NHIS = National Health Insurance Scheme. The National Health Insurance Scheme only covers healthcare services provided by participating hospitals, clinics, healthcare centers, and pharmacies. To participate, providers must register with the National Health Insurance Authority, giving proof that they are in good standing with the appropriate regulatory authority (the Health Facilities Regulatory Authority or the Pharmacy Council) and providing documentation to show that the professional staff are in good standing with the appropriate regulatory authorities. Providers 60 may be public, private, or mission facilities. For providers whose staff are not paid by government through other mechanisms, NHIS reimbursements are somewhat higher, though these facilities complain that the additional amounts they receive are insufficient to cover their staff costs (Laar, Asare, and Dalinjong, 2021). 24 In 2020, Ghana had 4,793 NHIS-accredited facilities, of which 69 percent were government run, 26 percent were private run, and 5 percent were run by missions. In 2017, there were 4,010 facilities, with roughly similar proportions of government-, private-, and mission-run facilities. National Health Insurance Scheme members are not required to make any copayment when they receive services, including medicines. Instead, healthcare providers must file a claim with the National Health Insurance Authority to be paid for the service provided. Originally, such claims were made on a fee-for- service basis, but this led to concerns that providers were overbilling NHIA. In 2012, for all payments except medicines, NHIA switched to a Diagnosis Related Groupings (DRG) approach, providing payment based on a standard set of diagnoses stipulated by NHIA in cooperation with the providers. Medicines continue to be reimbursed on a fee-for-service basis. The same act of Parliament that allowed for diagnosis related groupings also allowed for capitation payments as a means to control costs. 25 The National Health Insurance Authority, piloted a capitation scheme in the Ashanti region in 2012, but the pilot was politically contentious, was never expanded, and was eventually abandoned (Abiiro, Alatinga, and Yamey 2021). NHIS fees and payment rates for healthcare services provided by healthcare facilities are not indexed to inflation but adjusted based on a deliberative process undertaken by NHIA. Claims reimbursement is the aspect of NHIS that receives the overwhelming majority of complaints from providers. An extensive review of the literature on Ghana’s national insurance scheme found eight studies reporting slow reimbursement as a complaint of healthcare facilities (Christmals and Aidem 2020). Agyepong et al. (2016) also record slow or nonexistent reimbursement for medicines in interviews with key informants. Anecdotal evidence suggests that this problem is so severe that some participating facilities turn away NHIS members, or charge them for services despite that the members should have no copays. Raju and Younger (2022) find quantitative support for this observation, based on GLSS 2016/17 data. The National Health Insurance Scheme is an unusual insurance scheme in that 24 It is important to note that staff costs at public facilities and some quasi-public facilities such as the Christian Health Association of Ghana (CHAG) are paid by the central government through a separate mechanism, not through NHIA. This means that for anyone seeking care at a public facility, the staff costs are “insured” by the central government whether they are an NHIS member or not. This implies that, at public facilities, NHIS only covers nonstaff (and noncapital) costs, mostly medicines and diagnostic services. 25 Capitation payments pay providers a fixed annual amount for each member who enrolls with that provider as his or her “preferred primary provider.” 61 the fees and premia charged to members, even with increases in recent years, are far short of what would be actuarially fair. Instead, NHIS receives the bulk of its funding—roughly three-fourths—from a special 2.5 percent addition to the VAT called the National Health Insurance Levy (NHIL). A further one-fifth of its funding comes from the Social Security National Income Trust (SSNIT), which pays NHIS fees for its members (Wang, Otoo, and Dsane-Selby 2017). The NHIL, SSNIT contributions, registration and renewal fees, and premia are all pooled in the National Health Insurance Fund (NHIF), which is managed by the National Health Insurance Authority to pay provider claims and the operating expenses of the NHIA. Despite the large share of VAT revenue earmarked for NHIS, there have long been concerns that it is not financially sustainable as it is currently operated (Wang et al. 2017; Aikens et al. 2021). The National Health Insurance Scheme is meant to cover all Ghanaians though, as we have seen, coverage only reaches 30 percent to 40 percent of the national population. Given that the insurance scheme covers 95 percent of diagnosed conditions found in Ghana and that there are no copays, it may seem surprising that so few Ghanaians enroll in NHIS. Figure 27 based on GLSS 2016/17 data shows the distribution of responses to the question “Why have you never been registered with any health insurance scheme?”, for the overall population and for the poor. Almost half of the overall population report financial inability to pay for membership (or, inversely, the high cost of membership), and 70 percent of the poor report the same. Sizeable percentages of the overall population also report that they do not need health insurance or that they do not have confidence in NHIS. GLSS 2016/17 also asked respondents who once had NHIS membership but no longer do “If you are no longer covered, why?” Figure 27 also shows the distribution of the responses to this question, for the overall population and the poor. The results are similar to those for the “never” question, though financial inability to pay is more likely here as a response. In the operational literature on issues with NHIS, the greatest concern is with claims reimbursement. As we have noted, anecdotal and qualitative evidence suggests that this problem is so severe that participating providers sometimes turn away NHIS members or insist that they pay for services that should be free. GLSS 2016/17 respondents, though, do not often select a reason for not registering or renewing that is consistent with that argument. 26 Instead, the upfront registration, renewal, and premium fees seem to be the main deterrents to membership enrollment and continuation. 26 “Don’t have confidence in operators of scheme” and “Health insurance does not cover the services I need” would seem to be the two responses most consistent with the proposition that providers are not providing the services they should free of charge. 62 Figure 27. Reasons for Never or No Longer Being an NHIS Member a. Overall population b. Poor population Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows the distribution of responses as to why the individual no longer is, or never was, an NHIS member, for the overall population and for the poor population. In the figures, bars reflecting values less than 1 percent are not labeled. NHIS = National Health Insurance Scheme. The academic literature on the impact of NHIS on healthcare usage, health status, and out-of-pocket expenditures (OOP) is quite large, perhaps because NHIS was one of the first attempts to provide universal health insurance in Africa. Blanchet and Acheampong (2013), Okoroh et al. (2018), and Degroote et al. (2020) provide useful literature reviews. There is a clear consensus across 63 studies using a variety of methods that: • NHIS membership increases utilization of healthcare services; • NHIS members have lower out-of-pocket healthcare expenditures than nonmembers; • NHIS members have lower “catastrophic” healthcare expenditures, variously defined; and • NHIS members usually do not have better health outcomes than nonmembers. While we do not pretend to review all these studies, our interest in describing the distributional consequences of social assistance programs leads us to summarize three papers that examine the impact of NHIS on out-of-pocket healthcare expenditures since this affects beneficiaries’ disposable income (net of health care costs) directly. Powell-Jackson et al. (2013) investigate a randomized allocation of free healthcare in one district of Ghana, Dangme West, in 2004, just before the advent of NHIS. They find that this comprehensive insurance reduced out-of- pocket healthcare expenditures by 27 percent on average and that the reduction was higher at the upper end of the healthcare expenditure distribution, suggesting that those faced with the largest healthcare expenditures benefited the most from the insurance. 27 García-Mandicó et al. (2021) use data from GLSS 2005/06, whose fielding straddled the rollout of the NHIS at that time. Thus, they are able to use a difference-in-differences estimator to compare survey respondents in the same districts, some of whom gained access to NHIS during the survey fielding period and some of whom did not. They find a mean decrease in out-of-pocket healthcare expenditures of 18 percent and a 29-percent decrease for households that suffered “more illness.” 28 They also report small but statistically significant reductions in the probability of catastrophic healthcare expenditure, defined as out-of-pocket expenditure that is more than 10 percent of total household expenditure. Aryeetey et al. (2016) use instrumental variable methods to estimate the impact of NHIS membership on out-of-pocket and catastrophic expenditures. To instrument NHIS membership, they use the data and results of a cluster- randomized control trial meant to study the effect of a “multi-stakeholder problem-solving program” on NHIS enrollment. Comparing the difference between insured and uninsured households before the intervention and after, they find that NHIS cover reduces out-of-pocket expenditures for outpatient care by GH¢20 on average, compared to an average expenditure in both samples of GH¢32. This is a much larger percentage reduction in out-of-pocket expenditure than found in Powell-Jackson et al. (2013), but a small cedi amount, perhaps because it considers only out-of-pocket expenditure for outpatient 27 They do not, however, estimate the impact on catastrophic healthcare expenditures as usually defined. 28 This is defined as being above the sample median for the total number of days ill for adults in the household divided by the total number of adults in the household times 14 days. 64 services. 29 The study also estimates the impact of NHIS membership on catastrophic healthcare expenditures, defined as 40 percent or more of nonfood consumption. They find a 4-percentage-point reduction in the number of households suffering catastrophic expenditures, on average, compared to an average across the two samples of 23 percent. 29Their results for inpatient services actually suggest that NHIS increases out-of-pocket expenditures. 65 4. PERFORMANCE ANALYSIS We discuss findings from our analysis of the performance of the selected social protection programs in six main parts. In section 4.1, we discuss the levels of, and patterns and trends in, spending on programs. In sections 4.2–4.4, we discuss findings for program coverage, incidence, and effectiveness. In section 4.5, we discuss the hypothetical results on program incidence and outlay from simulating reforms in coverage and benefit levels for the LEAP program. In section 4.6, we discuss the findings from investigating the association between shocks and social assistance program participation. The findings in sections 4.1 are based on government administrative data on programs, the findings in sections 4.2–4.5 are based on GLSS 2016/17 data, and the findings in section 4.6 are based on high-resolution climate map data and government administrative data on programs. As data on LIPW program participation was not gathered in the GLSS 2016/17, we do not discuss coverage, incidence, and effectiveness for this program. 4.1. Spending How much expenditure went towards benefits under the government’s main social protection programs, that is, the LEAP program, GSFP, the LIPW program, SSNIT pensions, and NHIS? For each program, we specifically discuss the level of benefit spending in 2017, which mostly overlaps with the period for our ensuing performance analysis based on the GLSS 2016/17, and we discuss the level of benefit spending in the most recent year for which we have data. We examine the level of benefit spending relative to GDP and to overall government spending across years, bounded by 2021, the latest year for which we have actual (instead of forecasted) GDP and overall government spending statistics for Ghana. The periods for our benefit spending data differ from program to program. All program benefit spending data were obtained directly from the relevant program implementing agencies. Finally, pooling across relevant programs, we also discuss patterns and trends in benefit spending under social assistance (namely, the LEAP program, GSFP, and the LIPW program), social insurance (namely, SSNIT pensions and NHIS), and social protection (social assistance and social insurance combined). This study attempted to obtain data on spending on program administration, but they were either not shared by the program-implementing agencies (SSNIT, NHIA), or the data shared were potentially problematic because the program- implementing agencies shared administrative arrangements and personnel across multiple programs and initiatives (MLGDRD, MOGCSP). For all programs apart from SSNIT pensions, a fair amount of the yearly fluctuation in benefit spending is explained by delays in the internal flow of required funds to program-implementing agencies. In the case of the LIPW program, it is also attributable to the availability of donor financing. Across programs, periods of an uptrend in benefit spending in the data are largely explained by increases in the numbers of beneficiaries. In the case of SSNIT pensions, they are also attributable to nominal increases in the value of pension 66 payments. We obtained data on benefit spending under the LEAP program between 2016 and 2022 and under GSFP between 2016 and 2019. Over the period of our data, the LEAP program is partly financed through donor financing channeled through the government, whereas GSFP is financed by the government through general revenues. In 2017, LEAP program benefit spending was GH¢93 million, equivalent to 0.035 percent of GDP and to 0.19 percent of overall government spending (figure 28). In 2022, total program benefit spending was GH¢162 million. In 2017, GSFP benefit spending was GH¢259 million, equivalent to 0.10 percent of GDP and to 0.5 percent of overall government spending (figure 29). In 2019, the latest year for which we have data and also the year before implementation of the program was interrupted by public school closures resulting from the coronavirus pandemic, total program benefit spending was GH¢445 million, equivalent to 0.12 percent of GDP and to 0.6 percent of overall government spending. We obtained LIPW program benefit spending data between 2011 and 2021. The LIPW program is fully financed by donor funds channeled through the government. There was little spending in 2011 when the program was launched (so we exclude this year from our examination), and there was little to no spending in 2018 and 2019 when donor financing was unavailable (we retain these years in our examination). Except for the years when donor financing was unavailable, program benefit spending has generally increased over time (figure 30). Program benefit spending is divided between labor payments and payments for nonlabor components of the small civil works. Across years with significant program scale and spending, labor payments ranged from a low of 41.6 percent of total benefit spending (in 2017) to a high of 70.1 percent (in 2021). In 2017, total benefit spending under the LIPW program was GH¢39 million, equivalent to 0.015 percent of GDP and to 0.079 percent of overall government spending. In 2021, total benefit spending under the program was GH¢50 million, equivalent to 0.011 percent of GDP and to 0.029 percent of overall government spending. 67 Figure 28. Spending on the LEAP Program a. Absolute program benefit spending b. Total program benefit spending relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from MOGCSP and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trend in absolute benefit spending as well as in benefit spending relative to GDP and to overall government spending for the LEAP program. Panel (b) does not report statistics for 2022 because actual GDP and overall government spending information is unavailable. LEAP = Livelihood Empowerment Against Poverty. 68 Figure 29. Spending on GSFP a. Absolute program benefit spending b. Total program benefit spending relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from MOGCSP and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trend in absolute benefit spending and in benefit spending relative to GDP and to overall government spending for GSFP. GSFP = Ghana School Feeding Programme. 69 Figure 30. Spending on the LIPW Program a. Absolute program benefit spending b. Total program benefit spending relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from MLGDRD and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trend in absolute benefit spending and in benefit spending relative to GDP and to overall government spending for the LIPW program. LIPW = Labor Intensive Public Works. Benefit spending = labor and nonlabor payments. 70 The SSNIT pension benefit spending data cover the period between 2010 and 2020. In 2017, total spending on pension benefits was GH¢2,189 million, of which monthly pension benefits accounted for 84.6 percent and lump-sum payments for 13.4 percent (figure 31a). This level of spending on pension benefits was equivalent to 0.83 percent of GDP and to 4.4 percent of overall government spending (figure 31b). The benchmarking of program benefit spending to overall government spending should not be interpreted as reflecting how much the government spends on the program benefits. SSNIT pensions are financed by contributions from employers and workers provided to SSNIT and returns from investments of contribution funds by SSNIT. Government spending on program benefits applies to the extent that the government pays contributions on behalf of its employees to SSNIT. In 2020, the latest year of our data, total spending on pension benefits was GH¢3,303 million, equivalent to 0.84 percent of GDP and to 2.9 percent of overall government spending. NHIS benefit spending data (that is, NHIS spending on claims) are for 2010 to 2019. Program spending is almost fully financed by earmarked taxes and NHIS premium payments by SSNIT on behalf of its contributors and pensioners. Across the years, earmarked taxes have tended to account for roughly three- quarters of funds received by NHIA for the program. In 2017, total NHIS benefit spending was GH¢1,325 million, equivalent to 0.50 percent of GDP and to 2.6 percent of overall government spending. In 2019, the latest year of our data, total program benefit spending was GH¢979 million, equivalent to 0.27 percent of GDP and to 1.6 percent of overall government spending (figure 32). The caveat for our benchmarking against overall government spending that we mentioned in relation to SSNIT pensions applies to this program as well, albeit to a much lesser degree given that the bulk of funds for NHIS comes from earmarked taxes (that is, government financing). 71 Figure 31. Spending on SSNIT Pensions a. Absolute program benefit spending b. Total program benefit spending relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from SSNIT and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trend in absolute benefit spending and in benefit spending relative to GDP and to overall government spending for SSNIT pensions. SSNIT = Social Security and National Insurance Trust. 72 Figure 32. Spending on NHIS a. Absolute program benefit spending b. Total program benefit spending relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from NHIA and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trend in absolute benefit spending and in benefit spending relative to GDP and to overall government spending for NHIS. NHIS = National Health Insurance Scheme. Benefit spending = spending on claims. 73 Combining benefit spending under the LEAP program, the LIPW program, and GSFP, we examine total benefit spending under social assistance programs between 2016 and 2019, a period which subsumes our program performance analysis subperiod of 2016/17. GSFP benefit spending accounts for at least two- thirds of total benefit spending under social assistance programs (figure 33). Social assistance program spending is equivalent to 0.15–0.2 percent of GDP and to 0.8–1.0 percent of overall government spending. This level of spending relative to GDP is far below the corresponding average levels of social assistance programs in Sub-Saharan African countries and in low- and middle- income countries more generally for which we have estimates of social assistance program spending based on national household sample survey data (figure 34). The average level of social assistance program spending in these two groups of countries is 1.55 percent of GDP. 30 We also examine total social protection program benefit spending combining benefit spending under social assistance programs with benefit spending under social insurance programs (the latter composed of NHIS and SSNIT pensions) between 2016 and 2019. Benefit spending under social insurance programs accounts for at least four-fifths of total social protection program benefit spending (figure 35). Social protection program benefit spending is equivalent to 1.3–1.5 percent of GDP and to 6–8 percent of overall government spending. As discussed before in relation to benefit spending levels of NHIS and SSNIT pensions, the noted percentage of overall government spending should be interpreted with caution given the important nongovernment sources of financing of NHIS and SSNIT pensions. 30These international averages are likely a large underestimate given that total spending is estimated based solely on reported relevant data from household surveys (that is, without any corrections for coverage or benefit level discrepancies vis-à-vis administrative data, or for missed programs). Household surveys tend to capture relevant information on programs that offer regular cash transfers over extended periods; relevant information is frequently absent in these surveys on programs that offer monetary benefits but allow for short-duration participation and on programs that offer in-kind benefits such as food transfers. 74 Figure 33. Spending Across Social Assistance Programs a. Absolute benefit spending b. Relative benefit spending c. Overall social assistance benefit spending, relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from MOGCSP and MLGDRD and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trends in benefit spending for the LEAP program, the LIPW program, and GSFP, in absolute and relative terms; the figure also shows the trend in overall social assistance benefit spending relative to GDP and to overall government spending. LEAP = Livelihood Empowerment Against Poverty. LIPW = Labor Intensive Public Works. GSFP = Ghana School Feeding Programme. Social assistance = LEAP program, LIPW program, and GSFP. 75 Figure 34. Social Assistance Program Spending: Ghana in International Comparison Cumulative distribution of annual social assistance program spending as a percentage of GDP 1.0 0.8 Cumulative share 0.6 0.4 0.2 0.0 0 2 4 6 8 10 Ghana, 2016-19 SSA, global Percent All developing countries Sub-Saharan African countries Source: Global data from the World Bank’s Atlas of Social Protection Indicators of Resilience and Equity (ASPIRE) database (accessed May 19, 2020), http://datatopics.worldbank.org/aspire/home. Statistic for Ghana is own estimate based on government administrative data for the LEAP program, the LIPW program, and GSFP for 2016 through 2019, and on GDP statistics from the World Bank’s WDI databank. Note: Figure shows the cumulative distribution functions for annual social assistance spending as a percent of GDP for all developing countries and Sub-Saharan African countries, respectively, with the levels for Ghana between 2016 and 2019 and the mean levels for two country samples specified. For countries in the global data, the mean year for the data is 2014. Sample size for all developing countries is 124; sample size for Sub-Saharan African countries is 45. The blue vertical line labeled “SSA, global” indicates the mean value for all developing countries and for Sub- Saharan African countries. (The mean values for the two country samples are roughly the same, so they are treated as identical for the sake of graphing.) The red vertical lines labeled ”Ghana, 2016–19” reflect the band for values for Ghana between 2016 and 2019. For the purposes of this figure, social assistance programs in Ghana comprise the LEAP program, the LIPW program, and GSFP. SSA = Sub-Saharan Africa. 76 Figure 35. Social Protection Spending, Social Assistance versus Social Insurance a. Absolute benefit spending b. Relative benefit spending c. Overall social protection benefit spending, relative to GDP and to overall government spending Source: Authors’ estimates based on program administrative data from the various agencies and on GDP and overall government expenditure statistics from the World Bank’s Ghana Macro-Poverty Outlook Datasheet for April 2023. Note: Figure shows the trend in benefit spending on social assistance and social insurance programs, in absolute and relative terms; the figure also shows the trend in overall social protection benefit spending relative to GDP and to overall government spending. Social protection = LEAP program, LIPW program, GSFP, NHIS, and SSNIT pensions. Social assistance = LEAP program, LIPW program, and GSFP. Social insurance = NHIS and SSNIT pensions. 4.2. Coverage What is the extent of coverage of individuals and households by Ghana’s main government-provided social protection programs? GLSS 2016/17 captured information on program participation at the individual level for GSFP, NHIS, and SSNIT pensions. The survey also gathered information on participation at the household level for the LEAP program. For details on the construction of the program-participation variables, see appendix A. Across the programs, table A1 reports survey estimates of beneficiary numbers and compares these estimates to beneficiary numbers from government administrative data. Under the assumption that the government administrative 77 data are accurate, survey-based beneficiary numbers generally fall short. The shortfall is largest for the LEAP program, but these shortfalls are a concern for all programs we examine here. As an important caveat, this suggests that coverage rates for the programs are likely to be underestimated by a significant amount in the survey data. While we have no a priori reason to believe that the likelihood of underreporting in GLSS 2016/17 is correlated with any of the individual and household characteristics we examine in our subgroup analysis of coverage rates, if it is in fact correlated, it will bias our results. We examine program coverage rates for individuals (including “presumed eligible” or “target” individuals, based on applying some of the eligibility criteria for program benefit receipt to the GLSS 2016/17 data) and for households. For the definition of the coverage rate, see appendix B. We examine coverage rates separately by poverty group (extreme poor, moderate poor, near poor, and “other nonpoor”). The extreme poor are those with consumption (per adult equivalent) below the extreme poverty line, the moderate poor are those with consumption between the extreme and overall poverty lines, the near poor are those with consumption above the overall poverty line but below 1.5 times the overall poverty line, and the “other nonpoor” are those with consumption above 1.5 times the overall poverty line. At the national level, 8.2 percent of the population are extreme poor, 15.2 percent are moderate poor, 17.1 percent are near poor, and 59.5 percent are “other poor.” To allow for international comparisons, we estimate program coverage rates separately by household consumption quintiles, reported in tables C1 and C2. We also estimate program coverage rates for rural and urban areas, reported in the same tables. The discussion that follows focuses on program coverage rates at the national level for individuals and households, disaggregated by poverty group. For NHIS and SSNIT pensions, which are available to the national population more generally, we also discuss individual coverage rates across regions. NHIS reaches 34 percent of Ghana’s population (figure 36a). The other programs reach far fewer Ghanaians. GSFP reaches 4 percent of the national population, the LEAP program 1.5 percent, and SSNIT pensions 0.4 percent. Coverage is pro-poor for the LEAP program and for GSFP (figure 36b). NHIS coverage rates are roughly similar across poverty groups. Coverage by SSNIT pensions is essentially zero except for the “other nonpoor,” at 0.6 percent. We also examine the extent of GSFP and SSNIT pension coverage of their respective target groups. GSFP covers 25 percent of public preprimary and primary school students across Ghana, and 61 percent of public preprimary and primary school students in GSFP areas (that is, those primary sampling units in GLSS 2016/17 where there is at least one GSFP beneficiary) (figure 37). For GSFP’s target group, the progressiveness of program coverage falls when we shift from looking at the country as a whole to looking at GSFP areas within the country, indicating that, like for the LEAP program, GSFP’s progressiveness is driven by its geographic targeting to poor areas. SSNIT pensions cover 5 percent of those aged 60 or older. For this target group, the program’s coverage rate is 8 percent for the “other nonpoor,” and between 0.5 percent and 2 percent for the other poverty groups. 78 Figure 36. Coverage Rates of Individuals, by Program a. Overall b. By poverty status Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows individual coverage rates by program. These coverage rates are also estimated by poverty status (extreme poor, moderate poor, near poor, and other nonpoor); for definitions of the poverty statuses, see appendix A. The LEAP program benefit is treated as a household benefit; given this, all members in LEAP program households are considered beneficiaries. Bars with values lower than 0.5 percent are not labeled. LEAP = Livelihood Empowerment Against Poverty program. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. 79 Figure 37. Coverage Rates of Individuals in Target Categories for GSFP and SSNIT a. Overall b. By poverty status Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows individual coverage rates for GSFP and SSNIT pensions in target categories (where target categories are defined in the data by the authors). The coverage rate for GSFP is also estimated in GLSS 2016/17 primary sampling units where at least one GSFP household was present. Coverage rates are also estimated by poverty status (extreme poor, moderate poor, near poor, and other nonpoor); for definitions of the poverty statuses, see appendix A. Bars with values lower than 0.5 percent are not labeled. GSFP = Ghana School Feeding Programme. SSNIT = Social Security and National Insurance Trust. 80 As noted earlier, NHIS covers 34 percent of the national population. However, coverage rates vary widely across regions, from 22 percent in the Central region to 49 percent in the Upper East region (a spread of 27 percentage points) (figure 38). SSNIT pension coverage rates range from a low of virtually zero percent in the Northern region to about 1 percent in the Greater Accra region (figure 39a). Limiting the examination to those aged 60 or older, SSNIT pensions’ coverage rates across regions of course increase: They range from 0.2 percent in the Northern region to 14 percent in the Greater Accra region (figure 39b). Figure 38. NHIS Coverage Rates of Individuals, by Region Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows NHIS coverage rates of individuals by region, for the overall population. Classification of regions is per Ghana’s 2010 Population and Housing Census. NHIS = National Health Insurance Scheme. 81 Figure 39. SSNIT Pensions Coverage Rates, by Region a. Overall b. Individuals aged 60+ Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows SSNIT pension coverage rates of individuals by region, for the overall population and for those aged 60 or above. Classification of regions is per Ghana’s 2010 Population and Housing Census. SSNIT = Social Security and National Insurance Trust. We also estimate program coverage rates for households. In the case of individual-level benefits such as for GSFP, NHIS, and SSNIT pensions, a household is classified as a beneficiary if at least one member receives the program benefit. NHIS covers 48.5 percent of households nationally, followed by GSFP at 7.9 percent, SSNIT pensions at 1.4 percent, and the LEAP program at 0.9 percent (figure 40). 82 Figure 40. Coverage Rates of Households, by Program a. Overall b. By poverty status Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows coverage rates of households by program. These coverage rates are also estimated by poverty status (extreme poor, moderate poor, near poor, and other nonpoor); for definitions of the poverty statuses, see appendix A. A household is assigned as a beneficiary of a program if any member receives the benefit from the program. Bars with values lower than 0.5 percent are not labeled. LEAP = Livelihood Empowerment Against Poverty program. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. For GSFP, NHIS, and SSNIT pensions, coverage rates for households are somewhat higher than those for individuals. This is because in beneficiary households, some members directly receive the individual-level benefits, while 83 others do not. For the LEAP program, the coverage rate for households is a little lower than for individuals. This is because LEAP program households tend to be larger in size and so count more heavily in the individual-level coverage estimates. 31 Program coverage rates for households have similar patterns to program coverage rates for individuals across poverty groups. Finally, we examine coverage overlaps across programs at the individual and household beneficiary levels. The extent of overlaps is positively correlated with the extent of coverage for each individual program, with SSNIT pensions having the smallest coverage and NHIS the largest (figure 41). As expected, program coverage overlaps at the household level are markedly higher than those at the individual level. LEAP program, GSFP, and SSNIT pension households tend to have at least one member who is enrolled in NHIS, at 75 percent for LEAP program households, 65 percent for GSFP households, and 68 percent for SSNIT pension households. A sizeable share of LEAP program households—46 percent— also have at least one member who receives GSFP benefits. Figure 41. Individual- and Household-Level Coverage Overlaps across Programs Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows the rates of individual and household coverage overlap across programs. For the examination of coverage overlap among households, the household is considered a beneficiary of a program if any member of the household receives the benefit from the program. Even in the examination of coverage overlap among individuals, the LEAP program benefit is treated as a household benefit; as such, all members in LEAP program households are considered beneficiaries. Bars with values lower than 1 percent are not labeled. LEAP = Livelihood Empowerment Against Poverty program. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. 31 In LEAP program households, each member is considered to be covered. 84 4.3. Incidence Program incidence for the poor is measured by accounting for who receives a social assistance program benefit and the value of the benefit. For details on the construction of the program benefit variables, see appendix A. Before discussing our incidence results, we touch on program benefit levels. Based on our construction, the average LEAP program benefits received by LEAP program households is GH¢469. Analogously for each of the other programs, it is GH¢313 for GSFP households, GH¢291 for NHIS households, and GH¢7,734 for SSNIT-pension households. How large are these program benefits in relative terms? We examine this question based on two measures. The first measure is total benefits from a given program as a percentage of total household consumption averaged over households in that program (average percentage of consumption for short). This average percentage of consumption is lowest for NHIS, at 3.6 percent, and highest for SSNIT pensions at 60.3 percent (figure 42a). For the LEAP program and GSFP, the average percentages are 12.8 percent and 5.2 percent, respectively. 32 Breaking down further for the LEAP program and GSFP, if the sample is restricted to poor LEAP program households, the average percentage of consumption for LEAP program benefits is 15.1 percent, whereas it is 20.9 percent if restricted to extreme-poor LEAP program households. Likewise, if the sample is restricted to poor GSFP households, the average percentage of consumption for GSFP program benefits is 9.5 percent, whereas it is 14.5 percent if restricted to extreme-poor GSFP households. The second measure is total benefits from a given program as a percentage of the poverty line calculated at the household level averaged over households in that program (average percentage of the poverty line for short). We estimate this measure separately with respect to the overall and extreme poverty lines. 33 The LEAP program, GSFP, and NHIS have average percentages of the poverty lines that are low, specifically between 4 percent and 19 percent depending on the program and the poverty line (overall or extreme) (figure 42b). SSNIT pensions’ average percentage is 226 percent of the overall poverty line and 404 percent of the extreme poverty line. 32 As a comparison, in their sample of program households for an evaluation of the LEAP program, Handa et al. (2017) find that program benefits received by a program household in 2016 as a percentage of the household’s total consumption in 2010 averaged across program households was 18.3 percent. 33 We obtain household poverty line for each household by multiplying the stipulated individual poverty line by the number of adult equivalents in the household (see appendix A for more information on the poverty lines). 85 Figure 42. Relative Benefit Levels, Program Households, by Program a. Average percentage of consumption b. Average percentage of national poverty lines Source: Authors’ estimates based on data from GLSS 2016/17. Note: Figure shows total program benefits as a percentage of total household consumption averaged across program households, for each program; it also shows total program benefits as a percentage of the overall and extreme poverty lines at the household level averaged across program households, for each program. LEAP = Livelihood Empowerment Against Poverty program. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. 86 As a first way to assess program incidence, we examine the distributions of program beneficiaries and program benefits by poverty status. Starting with the analysis of the distribution of program beneficiaries by poverty status, LEAP program beneficiaries are disproportionately concentrated among the extreme and moderate poor (figure 43a). Seventy-four percent of LEAP program beneficiaries are extreme or moderate poor. In comparison, 23.4 percent of the population in general are either extreme or moderate poor. Notwithstanding, 17.9 percent of beneficiaries are near poor, and another 8.1 percent of beneficiaries are “other nonpoor.” GSFP beneficiaries are also disproportionately concentrated among the poor (46.1 percent of beneficiaries are either extreme or moderate poor). SSNIT pension beneficiaries are almost entirely nonpoor (96.6 percent of beneficiaries). What’s more, 91.1 percent of SSNIT pension beneficiaries are “other nonpoor.” NHIS beneficiaries are also mostly nonpoor—17.0 percent of beneficiaries are near poor and 63.0 percent are “other nonpoor.” Looking at the distribution across poverty statuses of NHIS beneficiaries by membership category, while beneficiaries categorized as “indigent” are more likely to be poor (45.7 percent of indigent beneficiaries are either extreme or moderate poor) compared to the population in general (23.4 percent), 42.2 percent of indigent beneficiaries are “other nonpoor” (figure 43b). NHIS beneficiaries in membership categories “below age 18,” “age 70+,” and “free maternal care” are all distributed (with respect to poverty status) similarly to the population in general. NHIS beneficiaries in the categories “employer paid” and “SSNIT paid” are much less likely to be poor: 89.3 percent of employer-paid beneficiaries and 95.5 percent of SSNIT-paid beneficiaries are “other nonpoor.” The results from the analysis of the distribution of program benefits by poverty status are qualitatively similar to those from the analysis of the distribution of program beneficiaries by poverty status (figure 44). Additionally, one consistent pattern is that, across programs and across NHIS membership categories, the share of benefits that go to the “other poor” exceeds the share of beneficiaries in this poverty category, indicating that average program benefits are lower for beneficiaries that are extreme, moderate, or near poor (as a collective group). To allow for a finer division than the poverty statuses we discuss here, figures C1 and C2 present the distribution of program beneficiaries and program benefits by consumption deciles. 87 Figure 43. Distribution of Program Beneficiaries, by Poverty Status a. By program b. NHIS, by membership category Source: Authors’ estimates based on data from GLSS 2016/17. Note: Panel (a) shows the distribution of beneficiaries across poverty status (extreme poor, moderate poor, near poor, and other nonpoor), by program; for definitions of the poverty statuses, see appendix A. Panel (b) shows the distribution of NHIS beneficiaries across poverty status, by category of beneficiary. Bars with values lower than 5 percent are not labeled. LEAP = Livelihood Empowerment Against Poverty program. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. FMC = Free Maternal Health Care. 88 Figure 44. Distribution of Program Benefits, by Poverty Status a. By program b. NHIS, by membership category Source: Authors’ estimates based on data from GLSS 2016/17. Note: Panel (a) shows the distribution of total benefits across poverty status (extreme poor, moderate poor, near poor, and other nonpoor), by program; for definitions of the poverty statuses, see appendix A. Panel (b) shows the distribution of total NHIS benefits across poverty status, by category of beneficiary. Bars with values lower than 5 percent are not labeled. LEAP = Livelihood Empowerment Against Poverty program. HH.con.pae = household consumption per adult equivalent. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. FMC = Free Maternal Health Care. 89 Concentration coefficients offer another way to assess program incidence. 34 A coefficient value closer to minus one indicates that the benefits from a program are more concentrated among poorer households, while a value closer to one indicates that they are more concentrated among richer households. LEAP program benefits are strongly concentrated among the poor (figure 45). The program’s concentration coefficient of –0.529 compares favorably with similar cash transfer programs in a large set of low- and middle-income countries. 35 GSFP benefits are also concentrated among the poor, but less so than the LEAP program. SSNIT pension benefits, on the other hand, are strongly concentrated among the richest. The program’s concentration coefficient of 0.736 far exceeds the concentration coefficient for household consumption per adult equivalent, at 0.417. For GSFP and especially for the LEAP program, the concentration coefficient in program areas is less negative than for the country as a whole, indicating that benefits from these programs are less well targeted in program areas than they are nationally, though even in program areas the poor still benefit disproportionately. Nevertheless, much of the success of these programs, especially the LEAP program, is accounted for by good geographic targeting (that is, the identification of program areas). Overall, benefits of NHIS membership are moderately concentrated among the nonpoor (0.194), but are less concentrated than household consumption per adult equivalent. Disaggregated by membership status, as to be expected, benefits to the indigent are concentrated among the poor (–0.138), but less so than either the LEAP program or GSFP. 36 Benefits for all other NHIS membership categories are not concentrated among the poor. 37 Further, benefits to members who participate in NHIS through the payment of premiums by employers or by SSNIT are concentrated among the richest, with their concentration coefficients (0.669 and 0.722, respectively) far exceeding the concentration coefficient for household consumption per adult equivalent. 34 For a definition of the measure, see appendix B. 35 The average concentration coefficient value for direct transfers of any type among 55 countries is –0.27, and the lowest coefficient value is –0.63 (http://commitmentoequity.org/datacenter) (accessed May 15, 2022). 36 This could be because the indigent that lack a fixed address are not captured in the GLSS sampling frame. 37 For reference, the concentration coefficients for being under 18, 70 or older, or pregnant are –0.11, –0.01, and 0.02, respectively. 90 Figure 45. Concentration Coefficients a. By program b. NHIS, by membership category Source: Authors’ estimates based on data from GLSS 2016/17. Note: Panel (a) shows concentration coefficients, by program. Panel (b) shows concentration coefficients for NHIS, by category of beneficiary. LEAP = Livelihood Empowerment Against Poverty program. GSFP = Ghana School Feeding Programme. NHIS = National Health Insurance Scheme. SSNIT = Social Security and National Insurance Trust. HH con.pae = household consumption per adult equivalent. FMC = Free Maternal Health Care. 91 Still another way to assess a program’s incidence is through its marginal effects on inequality and poverty. 38 These effects are defined as the change that each program produces in the Gini index, the poverty rate, or the poverty gap, with the sign reversed so that a positive value indicates a reduction in inequality or poverty. The poverty rate and the poverty gap are measured using Ghana’s overall poverty line. Marginal effects depend on who receives the program benefit and the value of the benefit received, reflected by the concentration coefficient and the size of the program, defined as the program benefits divided by household consumption averaged across households. At the national level, the base levels for evaluating the marginal effects are 23.4 percent for the overall poverty rate, 8.4 percent for the overall poverty gap, and 41.7 percent for the Gini index. The marginal effects of the LEAP program, GSFP, NHIS, and SSNIT pensions nationally on inequality and poverty nationally are uniformly small: Virtually all the marginal effects are less than fifth of a percentage point (table 1a). The estimates are small because the sizes of the programs are small. The marginal effects of the LEAP program in LEAP program areas on inequality and poverty are at least an order of magnitude larger than the LEAP program’s marginal effects nationally. This is entirely due to the fact that the size of the LEAP program (2.1 percent of household consumption on average) is much larger in program areas. Restricting GSFP to GSFP areas yields changes in program size and in marginal effects similar to when we restricted the LEAP program to LEAP program areas, but these changes are less pronounced. Compared to GSFP nationally, the program size and the marginal effects of GSFP in GSFP areas roughly triple in magnitude. As noted earlier, we estimate far fewer LEAP program beneficiaries based on GLSS 2016/17 than we find based on government administrative records. This could be because the survey was not designed to produce a representative sample of LEAP program beneficiaries and just happened to miss the areas where this program is available. Or it could be because survey respondents do not report themselves as program beneficiaries. To adjust for this, we predicted an additional set of LEAP program households who are similar to those receiving LEAP program benefits in the survey but report that they do not receive it. For details on the prediction exercise, see appendix E. We add predicted LEAP program households to the set of actual LEAP program households, until we reach the administrative number of LEAP program households in January 2017 (roughly 198,000). We refer to this augmented set of program households as “LEAP+”. 38 For a more detailed definition of the measure, see appendix B. 92 Table 1. Marginal Effects Marginal effects Program size (percentage points) (percent) Inequality Poverty rate Poverty gap (Gini index) (1) (2) (3) (4) a. By program LEAP 0.03 0.06 0.05 0.14 LEAP, in LEAP areas 0.53 0.82 0.83 2.08 LEAP+ 0.11 0.13 0.17 0.51 GSFP 0.03 0.08 0.04 0.11 GSFP, in GSFP areas 0.08 0.27 0.11 0.33 NHIS 0.06 0.21 0.10 0.45 SSNIT pensions 0.00 0.03 0.18 0.22 b. NHIS, by membership category Indigent 0.00 0.00 0.00 0.00 FMC 0.00 0.01 0.00 0.02