101230   overty Reduction in Ghana P Progress and Challenges 2015 Vasco Molini and Pierella Paci Poverty Reduction in Ghana Progress and Challenges 2015 Vasco Molini Pierella Paci © 2015 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 18 17 16 15 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. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free  to ­ollowing copy, distribute, transmit, and adapt this work, including for commercial purposes, under the f conditions: Attribution—Please cite the work as follows: Molini, Vasco, and Pierella Paci. 2015. “Poverty Reduction in Ghana 2015: Progress and Challenges.” World Bank, Washington, DC. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official content or error in this translation. World Bank translation. The World Bank shall not be liable for any ­ Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Cover photo: Jonathan Ernst/World Bank. Permission required for reuse. Cover design: Bill Pragluski/Critical Stages. Contents Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Executive Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Framing Ghana’s Success. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Definitions and Technical Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Two Decades of Expanding Prosperity for All. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Shared Prosperity Led to a Reduction in Poverty and Vulnerability. . . . . . . . . . . . . . . . . . . . . . . . . . 7 The Poor over Time: Better Attributes and Living Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Better and More Equal Opportunities for Sustainable Progress. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2 Drivers of Poverty Reduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Key Economic Developments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 The Main Drivers of Poverty Reduction and Consumption Growth. . . . . . . . . . . . . . . . . . . . . . . . . 28 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3 Remaining Challenges. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Growing Inequality and Polarization in Household Consumption. . . . . . . . . . . . . . . . . . . . . . . . . . 35 Persistent Spatial Disparities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 The Deteriorating Macroeconomic Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4 A Roadmap for Policy Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 A Tale of Success. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 A Less Positive Outlook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 A Roadmap for Policy Action. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Appendix A  Computing Poverty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Appendix B  Regression Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Boxes 1.1 How Has the Typical Poor Household Changed between 1991 and 2012?. . . . . . . . . . . . . . . . . 13 3.1 Poverty Maps in Ghana . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Contents iii Figures ES.1 Real GDP Growth, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix ES.2 Trends in Poverty and Extreme Poverty, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x ES.3 Distribution of Employment, by Sector and Consumption Quintile. . . . . . . . . . . . . . . . . . . . . . xi ES.4 Workforce Educational Attainment, by Sector of Employment, 1991–2012 . . . . . . . . . . . . . . . . xi ES.5 Decomposition of Poverty Changes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii ES.6 Poor Individuals in Rural and Urban Areas, by Region, 1991–2012. . . . . . . . . . . . . . . . . . . . . xiii 1.1 Real GDP Growth, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.2 Growth Incidence Curves, 1991–2012 and 2005–12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3 Trends in Poverty and Extreme Poverty, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.4 Growth, Inequality, and Poverty Decomposition, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.5 Growth Elasticity of Poverty, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.6 Vulnerability as Distance from the Poverty Line, Nationwide, 1991–2012 . . . . . . . . . . . . . . . . . 9 1.7 Vulnerability and Poverty Depth as Distance from the Poverty Line, Rural and Urban Areas, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.8 Trends in Regional Poverty Headcounts, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 1.9 Main Sectors of Employment, Household Heads, by Quintile, 2012. . . . . . . . . . . . . . . . . . . . . . 13 1.10 Access to Basic Services, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1.11 Asset Index, Rural and Urban Areas and Nationwide, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . . 16 2.1 Sectoral Composition of GDP, 1991–2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.2 Poverty and Access to Employment, Adults Aged 15–64, 2012. . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.3 Production Indexes, Major Crops, 2000–10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.4 Output per Worker, by Sector, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.5 Distribution of Employment, by Sector, Region, and Poverty Status . . . . . . . . . . . . . . . . . . . . . 24 2.6 Labor Force with at Least Basic Education, by Region, 1991 and 2012. . . . . . . . . . . . . . . . . . . . 25 2.7 Education and Employment, 1991 and 2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.8 Variation in Population, by Rural and Urban Area and Region, 1991–2012. . . . . . . . . . . . . . . 27 2.9 Household Heads Who Were Not Born in Their Current Places of Residence. . . . . . . . . . . . . 27 2.10 Decomposition of Poverty Changes, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2.11 Education Coefficients, by Percentile and Year. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.12 Employment Category Coefficients, by Percentile and Year. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.13 Decomposition of Household Consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.14 Decomposition of Changes in Characteristics, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.1 Decomposition of the Gini Index, by Household Characteristic, 1991–2012 . . . . . . . . . . . . . . 36 3.2 Gini Indexes in Sub-Saharan Africa. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3.3 Relative Consumption Distribution, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.4 Median-Adjusted Relative Consumption Distribution, 1991–2012 . . . . . . . . . . . . . . . . . . . . . . 38 3.5 Median-Adjusted Relative Consumption Distribution Series, Ghana, 1991–2012. . . . . . . . . . 39 3.6 Regional Price Indexes: Total and Nonfood, 2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 A.1 Distribution of Highest Education Level of Household Head, by Percentile and Year. . . . . . . 56 A.2 Distribution of Type of Employment, by Percentile and Year . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 A.3 Distribution of Household Characteristics, by Percentile and Year. . . . . . . . . . . . . . . . . . . . . . . 57 iv Contents Maps 3.1 Poverty Maps, 2000 and 2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.2 Additional Indicators of Welfare, 2011 and 2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 A.1 Administrative Map of Ghana. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Tables ES.1 Infant and Under-5 Mortality, Vaccination, and Fertility Rates, 1988–2014. . . . . . . . . . . . . . . . x ES.2 Interquartile Consumption Ratios, by GLSS Wave, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . xii I.1 Ghana Living Standards Surveys 1–6, 1987–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1 Mean and Median Real Consumption (Adult Equivalent), 1991–2012 . . . . . . . . . . . . . . . . . . . . 6 1.2 Poverty Rates, the Poverty Gap, and the Severity of Poverty, 1991–2012. . . . . . . . . . . . . . . . . . . 8 1.3 The Profiles of Poor and Non-Poor Households, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.4 Educational Attainment, Bottom 20 and Top 20, by Gender, 1991–2012. . . . . . . . . . . . . . . . . . 14 1.5 Infant and Under-5 Mortality, Vaccination, and Fertility Rates, 1988–2014. . . . . . . . . . . . . . . 15 2.1 GDP Growth Decomposition, by Per Capita Value Added, 1991–2012. . . . . . . . . . . . . . . . . . . 20 2.2 Employment, by Major Sector, 1991–2012 (% of Total Employment) . . . . . . . . . . . . . . . . . . . . 21 3.1 Measures of Per Capita Household Consumption Expenditure, 1991–2012. . . . . . . . . . . . . . . 36 3.2 Interquartile Consumption Ratios, by GLSS Wave, 1991–2012. . . . . . . . . . . . . . . . . . . . . . . . . . 37 A.1 Price Deflators (2005/06–2012/13). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 A.2 Poverty Rates Using the Revised and the Old Poverty Lines with Different Price Deflators. . . . 55 B.1 Probit Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 B.2 Quantile Regression, 1991. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 B.3 Quantile Regression, 1998. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 B.4 Quantile Regression, 2005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 B.5 Quantile Regression, 2012. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 B.6 Oaxaca-Blinder Poverty Decomposition by 40th and 60th Percentiles, Variation between 1991 and 2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Contents v Acknowledgments This report was prepared by a core team com- Sara  Johansson (Consultant), and Michele prising Vasco Molini (Task Team Leader, Tuccio (Consultant) provided important inputs. GPVDR); Pierella Paci (Lead Economist, We also thank Ghana Statistical Service for their GPVDR); Fabio  Clementi, Dan  Pavlesku, and collaboration and willingness to share their data. Francesco Schettino (Consultants), under the ­ enefited from discussions with gov- The report b guidance of Pablo Fajnzylber (Practice overall ­ ernment ­ officials, development partners, the Manager, GPVDR); Yusupha Crookes (Country World Bank Africa Poverty team, and partici- Director, AFCW1); Kathleen Beegle (Program pants in a workshop held in Accra in October Leader, AFRW1); and Andrew Dabalen (Lead 2015. Martin Buchara (Team Assistant, GPVDR) Economist, GPVDR). Rose Mungai, Maddalena provided excellent logistical assistance in the Honorati, Aly Sanoh, Ayago Wambile, Marco preparation of this report. Final editing was done Ranzani, and  Dhiraj Sharma (GPVDR), by Robert Zimmerman. Acknowledgments vii Executive Summary Ghana has posted a strong growth ­performance Between 1991 and 2006 the impact of during the past two decades. After more than a growth on poverty in Ghana was higher than decade of stable annual growth in gross domestic in the rest of Sub-Saharan Africa (SSA). Until product (GDP) at between 4 and 5 percent, 2005 Ghana enjoyed an elasticity of poverty to growth began to pick up in the early 2000s and growth well above 2, closer to that of other devel- reached a steady rate of nearly 8  percent after oping countries and well above the SSA average. 2006 (figure 1). Over the last 20 years, the However, since then the elasticity has declined to Ghanaian economy has almost always grown a more modest 0.7, close to the SSA average but more quickly than have the economies of other considerably below that of other developing Sub-Saharan African countries (Africa hereafter) countries. and at rates similar to those of lower-middle-­ Progress has gone beyond the reduction of income countries. consumption poverty. Ghana has also substan- Rapid growth translated into accelerated tially improved various nonmonetary indicators poverty reduction. The poverty rate fell by more of poverty. For example, infant mortality than half between 1991 and 2012, from declined from 57 deaths per 1,000 live births in 52.7 percent to 21.4 percent (figure 2). The coun- 1998 to 41 in 2014, and under-5 mortality try seems easily on track to reduce the poverty declined by more than half (table 1). Fertility is rate in line with Millennium Development also decreasing and this has led to reduction in Goal 1. Its performance compares well with that the dependency ratio. of other countries in Africa. In 2012, the poverty These improvements in poverty reduction rate in Ghana was less than half the African aver- occurred during a period of rapid change in age of 43 percent, while in 1991, it had been only the economic and sociodemographic structure 10 percent lower than the African rate.1 The of the country. Three are the factors associated extreme poverty rate declined even more quickly, with the reduction in poverty: (1) structural dropping from 37.6 percent in 1991 to 9.6 percent transformation, (2) the growing skills of the labor in 2012.2 force, and (3) geographical mobility. Figure ES.1  Real GDP Growth, 1991–2012 16 14 12 10 Percent 8 6 4 2 0 –2 94 04 10 09 99 00 98 03 93 08 92 12 96 02 06 11 91 01 95 05 97 07 20 20 20 19 20 19 20 19 19 19 20 19 20 20 19 20 20 19 20 19 20 20 Ghana Lower-middle income (worldwide) Sub-Saharan Africa (developing only) Source: World Development Indicators (WDI) database. Executive Summary ix Figure ES.2  Trends in Poverty and Extreme Poverty, 1991–2012 52.7 50 37.6 39.8 40 Percent 27.1 28.5 30 21.4 18.1 20 9.6 10 0 1991 1992 2005 2012 Poverty Extreme poverty Source: Calculations based on Ghana Living Standards Survey (GLSS) 3–6. Table ES.1  Infant and Under-5 Mortality, Vaccination, and Fertility Rates, 1988–2014 Rate 1988 1993 1998 2003 2008 2014 Infant mortality 77 66 57 64 50 41 Under-5 mortality 87 57 54 50 31 19 Vaccination — 54 62 69 79 84 Fertility 6.4 5.2 4.4 4.4 4.0 4.2 Source: Demographic and Health Surveys 1988–2015: STATcompiler (DHS Program STATcompiler) (database), ICF International, Rockville, MD, http://www.statcompiler.com/. The recent economic growth is associated the relative growth in the service ­ sector—​ with a shift of the economy out of agriculture. 23.9 percent of total GDP growth—was a hetero- The share of agriculture in GDP declined by geneous group of activities including financial nearly 40 percent, from over a third of GDP in and business services, public administration, 1991 to 23 percent in 2012. As a result, by 2011, education, health care, social protection, and agriculture was the smallest sector in the econ- other services. This was followed by “transport, omy in terms of value added, although it is still storage, and communication,” 18.5 percent of the main sector of employment, representing overall GDP growth, and “wholesale, retail trade, 43.2 percent of total employment. Agriculture restaurants, and hotels,” 13.8 percent. also saw a stable increase in productivity and a The type of job done is strongly correlated substantial reduction in the workforce. The pro- with poverty status. In 2012, agriculture was by duction of cocoa and other cash crops was the far the dominant sector of employment among main driver of these changes. Beginning in the the poorest 20 percent of the consumption distri- late 1990s, cocoa production expanded quickly, bution, but it accounted for only 14 percent of and Ghana became the world’s third-largest the jobs among the richest 20 percent, 42 percent cocoa producer. of whom are wage employees (figure 3). Workers leaving agriculture moved mostly The labor force has become better educated into services and, to a lesser extent, industry. and higher level of education translated into Employment in the service sector expanded from better job opportunities over the past two 28.8 percent of total employment in 1991 to decades. Between 1991 and 2012, the share of the 42.0 percent in 2012, and the share of construc- labor force without schooling almost halved— tion more than doubled over the two decades. from 41 to 24 percent—and in 2012 the majority The share of employment in industry also of workers had completed at least junior second- increased, although at a lower rate, from ary education, compared to 39 percent in 1991. 10.7 percent to 14.9 percent. The main driver of Figure 4 also shows that education is an important x Executive Summary Figure ES.3  Distribution of Employment, by Sector and Consumption Quintile Richest Fourth Third Second Poorest 0 20 40 60 80 100 Percent Wage, private Wage, public Self, nonagriculture Self, agriculture Source: Calculations based on GLSS 3–6. Figure ES.4  Workforce Educational Attainment, by Sector of Employment, 1991–2012 100 90 80 70 60 Percent 50 40 30 20 10 0 1991 2013 1991 2013 1991 2013 1991 2013 1991 2013 None Primary Junior secondary Senior secondary Tertiary Wage, public Wage, private Self, nonagriculture Self, agriculture Source: Calculations based on GLSS 3 and 6. driver of higher earnings within each type of work opportunities. It is therefore not surprising that and it also facilitated access to more productive, moving to the faster growing areas of the South and better paid, activities. It is particularly strik- and Ashanti has been seen by many as an effective ing that more than half of Ghana’s tertiary edu- way of escaping poverty. In absolute numbers, cated workers are employed in the public wage since 1991, Accra and Ashanti gained over sector; with another thirty percent in the private 2.4 million inhabitants each, around half of them wage sector, and only 14 percent in nonfarm self-­ in the last decade. employment. The high concentration of highly Which of these developments were the main educated workers into the public sector may act as drivers of poverty reduction and shared pros- barrier to the development of a modern private perity? Using econometric analysis and standard sector. decomposition techniques, the report concludes Location was another major correlate of that, on average over the period changes in the poverty. Spatial inequalities in the incidence of characteristics of the household, in what they do poverty are striking and patterns of poverty are and where they live are just as important in closely linked to a divergence in employment explaining the positive performance as changes in Executive Summary xi Figure ES.5  Decomposition of Poverty Changes 1991–2012 1991–2005 2005–12 –0.20 –0.15 –0.10 –0.05 0 0.05 0.10 Variation in poverty rate Endowments Coefficients Interaction Source: Calculations based on GLSS 3–6. Table ES.2  Interquartile Consumption Ratios, by GLSS Wave, 1991–2012 Year p90/p10 p90/p50 p10/p50 p75/p25 p75/p50 p25/p50 Gini 1991 5.23 2.42 0.46 2.37 1.56 0.66 0.38 1998 6.00 2.48 0.41 2.60 1.64 0.63 0.38 2005 6.36 2.46 0.39 2.63 1.62 0.61 0.41 2012 6.73 2.65 0.39 2.68 1.66 0.62 0.41 Source: Calculations based on GLSS 3–6. the rewards to these characteristics. However, In 1991, consumption per capita was about five before 2005 the improvements were mainly times greater among the top decile of the distri- driven by high returns to certain household bution than among the bottom percentile characteristics while further improvements after (table 2). By 2012, the gap had widened to nearly 2005 were driven by accumulation of productive seven times, and the Gini index rose 8 percent, endowments—i.e., to changes in some of those from 37.5 to 40.8. However, Ghana still compares characteristics (figure 5). Among these charac- favorably with other African countries; its Gini teristics, it is worth noting the importance of is  still below the median and one of the low- access to basic infrastructures, increased educa- est  compared with rapidly growing African tional attainment, and modification in the house- countries. hold structure. Much of the increase in inequality is the Despite Ghana’s success in reducing poverty reflection of increased regional disparities, and promoting shared prosperity, three signifi- although within the region inequalities are also cant challenges remain: growing inequality pronounced. Poverty rates have fallen below and polarization in household consumption, 20  percent in the large area that includes the large spatial disparities in welfare, and the Ashanti, Eastern, Greater Accra, and Western deteriorating macroeconomic environment. regions, southern Brong Ahafo Region, and Consolidating the progress made in poverty coastal Volta Region. Poverty has also declined reduction and shared prosperity in recent well below the 40 percent in the central belt. decades will require addressing these challenges Recent improvements notwithstanding, the pov- effectively. erty rate is far above 40 percent in most districts Inequality in household consumption in the north. As a result poverty has increasingly increased, particularly between 1998 and 2005. become concentrated in rural areas and in the xii Executive Summary Figure ES.6  Poor Individuals in Rural and Urban Areas, by Region, 1991–2012 Number of individuals in thousands 1,000 500 0 Ce n Ea ta A rn No afo pe n rn ra Ea ta A rn No afo pe n ra pe ast t t t rA l i te al i es as es on nt on nt te tra er Up her Up her cc l l e te e cc ea ntr Vo Vo Ah Ah t st st rW rW Up r E rE Br sha Br sha n es es rA rt rt e W W C g g pe Up ea Gr Gr Rural Urban 1991 2012 Source: Calculations based on GLSS 3–6. Northern part of the country: one out of three immediate policy priority. The country’s long- poor people lives in the northern rural areas while term growth prospects remain positive, but, to in 1991 it was only one out of five (figure 6). realize its full potential, Ghana needs to succeed Ghana is facing deteriorating macroeco- in the implementation of the stabilization and nomic prospects. Since 2014 GDP growth has reform program undertaken in 2014. The success halved and is projected to slow further to of the program hinges on a sustained commit- 3.4  percent in 2015 as energy rationing, high ment to fiscal discipline, rapid progress in struc- inflation, and ongoing fiscal consolidation con- tural reforms, and the reduction of inflation. tinue to weigh on economic activity. Simulations Tackling inequalities in outcomes and indicate that the decline in household purchas- opportunities is a longer-term development ing power alone could cause an increase of the challenge, but it is key for consolidating the current poverty rate of about 5 percent, with the country’s middle-income status. Ghana entered urban lower deciles being the most affected. a new stage of development with its designation Looking forward, the development chal- as a middle-income country in 2011. It will now lenge faced by Ghana is to consolidate its pov- be difficult to achieve sustained progress in pov- erty reduction successes in the context of erty reduction and shared prosperity without difficult internal and external economic condi- broadening the reach of the development process tions and a rapidly changing economic and to those people who have so far been left behind. social environment. The deteriorating macro- The main challenge is to improve access to economic outlook and the persistent inequalities opportunities across the entire population with- threaten future progress given the strong link out stifling the energy of the economy. This will between economic growth and poverty reduc- require a multifaceted, well-targeted, and fiscally tion and the tendency of higher inequality and sustainable package of policies that balance the polarization to translate into a lower growth elas- needs of the poor with the needs of the most ticity of poverty reduction. dynamic economic sectors. Preventing further deterioration in the A small set of win-win policy areas emerge macroeconomic environment is the most as priorities to consolidate Ghana’s success in Executive Summary xiii poverty reduction and shared prosperity. rights and more efficient land markets will facili- Improving the business climate is a must to enable tate structural transformation by allowing Ghana a modern private sector to flourish and create to benefit from geographical agglomeration. high-productivity, well-paying jobs. A continued Finally, experimenting with innovative ways of commitment to investment in infrastructure and expanding the coverage of social protection and skills development will also be key to increase improving its targeting can help reduce the high productivity in agriculture, create modern jobs, vulnerability to shocks of Ghanaian households and ensure that workers have the skills needed to and increase the productive potential of individ- take advantage of the new employment opportu- uals by breaking the vicious circle that links nities. Increased connectivity between rural and inequality of income to unequal opportunities urban areas, combined with clearly defined land over generations. Notes 1. The average poverty rate in Africa has been cal- 2. For comparing the 4 survey rounds analyzed in culated using the international poverty line of the  report, they used the 1999 poverty line and US$1.25. food poverty line. See also GSS (2014). xiv Executive Summary Introduction Ghana has long been at the vanguard of devel- milestone performance by profiling the opment in Africa. As the first country on the changes and identifying their main drivers. continent to gain independence from European The last 25 years have borne witness to perhaps rule, in 1957, Ghana continues to provide a the greatest advances in the country’s history. For model of stability, democracy, and prosperity to this reason, this study takes a medium-term per- low- and middle-income countries in Africa and spective to identify and understand more closely further afield. In addition, it has played a leading the circumstances in which poverty has fallen so role in its neighborhood by promoting peace, quickly and steadily since the 1990s. The report economic development, and regional coopera- profiles the progress made during this period in tion in West Africa. reducing poverty and increasing the consump- tion of households in the bottom 40 and identi- fies the main drivers of this success. The country’s rapid economic growth has Framing Ghana’s Success been the main engine of poverty reduction, but Progress made since the 1990s has been partic- challenges remain. Ghana has been more effective ularly remarkable. The country has achieved than other African countries in sharing the dramatic gains in living standards, public health, increased prosperity and transforming it into pov- and educational attainment, and it has enjoyed a erty reduction. Recent successes notwithstanding, stout increase in consumption among the bot- several challenges remain and, using a combina- tom 40 percent of the consumption distribution tion of standard and more innovative techniques— (the bottom 40). These achievements have been such as poverty mapping and polarization accompanied by strong economic performance, measures—the report highlights three areas of leading to the country’s landmark achievement concerns that are seen as particularly important: of middle-income status in 2010, a decade earlier the growing inequalities of consumption and than anticipated. opportunities, the persistent spatial disparities, and Ghana’s success story has been underpinned the deteriorating macroeconomic environment. by the ability of the economy to generate an Despite the challenges, however, what fol- essential dividend from growth: poverty reduc- lows herein is ultimately an extraordinary suc- tion. Both absolute and extreme poverty rates cess story with lessons that may be learned and have dropped dramatically in the last 25 years. applied elsewhere in the developing world. In Since 1991, the national poverty rate has fallen by itself, the country’s graduation to middle-income more than half, from 52.7 percent that year to status is not without challenges. More Ghanaians 21.4 percent in 2012,1 a feat few other countries than ever are well educated and seeking employ- can claim, and one that sets Ghana on course ment, while civil society is thriving. As more to  meet the Millennium Development Goal people are lifted from poverty, their expectations Objective 1. Moreover, extreme poverty rates will rise, along with their living standards. have dropped dramatically in the last 25 years, Maintaining progress will therefore require and the share of the extreme poor in the popula- laying the groundwork for the country’s next tion declined from 37.6 percent in 1991 to stage of development. The policy agenda out- 9.6 percent in 2012. lined at the end of the report is intended to guide Within such a background, this poverty the Ghanaian government in identifying a compre- assessment seeks to shed light on Ghana’s hensive and fiscally sustainable policy package to Introduction 1 successfully address the challenges and consoli- GLSS to estimate the total consumption of each date Ghana’s successes in the years to come. household. This covers the consumption of both This report consists of four chapters. food and nonfood items (including housing). Chapter 1 profiles the trends in household con- Food and nonfood consumption commodities sumption and poverty rates, and in the character- may be explicitly purchased by households or istics of the poor observed between 1991 and acquired through other means (own production 2012. Descriptive statistics of consumption and activities or through receipts). The household selected poverty indexes are presented and a consumption measure takes into account all ­ profile of the characteristics of the poor is these sources in the various modules of the sur- given. The chapter concludes with an analysis of vey questionnaires. vulnerability. Chapter 2 uses descriptive and ­ The availability of comparable survey data econometric techniques to identify the drivers beginning in 1991 allows us to carry out an of  Ghana’s success over the last two decades. updated, detailed analysis of poverty in Ghana. Chapter 3 examines the main challenges Ghana The surveys have improved in quality over the continues to face: widening inequalities, a persis- years and involved the collection of data on both tent spatial divide, and the deteriorating macro- the monetary and nonmonetary dimensions of economic environment. Chapter 4 provides a welfare, thereby permitting an accurate analysis roadmap for policy action to effectively address of poverty and inequality over time. The GLSS these challenges and consolidate Ghana’s success has emerged as one of the most important as a middle-income economy. tools for the welfare monitoring system in Ghana. It provides the basis not only for official welfare measures and analysis, but also for detailed ­ information on several socioeconomic and Data demographic characteristics, the household con- The analysis in this report is based on data of sumption of purchased and home-produced the Ghana Living Standards Survey (GLSS) goods, asset ownership, and remittances. produced by the Ghana Statistical Service The GLSS is based on a two-stage (non-​ (GSS). The GLSS is a multipurpose survey that stratified) sample design; in data analysis, collects detailed information on individual and ­sampling weights are therefore used to account household characteristics and on basic indicators for the survey design. To enhance the compara- of living standards (table I.1). Six rounds of GLSS bility of consumption data over the four latest, data have been collected since 1987, thereby pro- comparable waves, all expenditures are deflated viding over 20 years of comparable data. However, across both space and time, expressed in 2005 only the last four rounds, from GLSS (round) 3 constant prices, and converted, if necessary, from to GLSS 6, have been based on the same ques- the Ghanaian second cedi (1967–2007) to the tionnaire and are therefore fully comparable. The Ghanaian third cedi (2007–), that is, to accom- GSS collects sufficient information through the modate GLSS 3 to GLSS 5. Each of the four waves Table I.1  Ghana Living Standards Surveys 1–6, 1987–2012 Dataset Collection period Sample size, number of households Representativeness Comparability GLSS 1 September 1987–August 1988 3,172 National, urban, and rural GLSS 1 and 2 are comparable GLSS 2 October 1988–August 1989 3,194 National, urban, and rural GLSS 1 and 2 are comparable GLSS 3 September 1991–August 1992 4,523 National, urban, and rural GLSS 3–6 are comparable GLSS 4 April 1998–March 1999 5,998 National, urban, and rural GLSS 3–6 are comparable GLSS 5 September 2005–August 2006 8,687 National, urban, and rural GLSS 3–6 are comparable GLSS 6 October 2012–October 2013 16,772 National, urban, and rural GLSS 3–6 are comparable Source: GLSS 1–6. 2 Introduction is organized into four modules, which are stored unless otherwise specified. The headcount index in the individual, labor force, household, and measures the proportion of the population with household expenditure files. per adult equivalent consumption below the value The availability of comparable and extensive of a minimum basket of food and nonfood items, information covering over two decades rep- that is, the poverty line. This population share resents a success on its own. Ghana is one of the with per adult equivalent consumption below the few countries in Africa that has produced compa- value of a minimum basket of food (food poverty rable, high-quality household data covering over line) only is called the extreme poor.3 two decades.2 This is an important achievement The report uses the 1999 poverty line. In because the availability of such rich and compara- 2013, the GSS revised the poverty line and modi- ble information beginning in 1991 allows an fied the basket to take into account the significant updated and detailed analysis of the country’s changes that had occurred in the consumption recent successes in poverty reduction, including basket (see appendix A).4 However, given the the drivers behind the reduction. The quality focus of this report on comparisons over time, we improvements of the surveys over the years and maintain the 1999 line as the poverty threshold the fact that they collect data on both the mone- throughout the entire period under analysis. tary and the nonmonetary dimensions of welfare This does not appear to have unduly affected the are particularly welcome because this permits the poverty rate for 2012, which we estimate at 21.4 establishment of an accurate picture of poverty percent, compared with the official 24.3 percent and inequality over time. based on the revised poverty line. The poverty gap index measures the extent to which individuals fall below the poverty line (the poverty gap) as a proportion of the poverty line. Definitions and Technical Notes The sum of the poverty gaps so calculated yields The selected measure of welfare is consumption the minimum cost of eliminating poverty if per adult equivalent. Consumption has proven transfers were perfectly targeted. The poverty gap ­ preferable to income as a measure of poverty index (or the index of the severity of poverty) because it is less volatile (for example, see Deaton squared averages the squares of the poverty gaps and Zaidi 2002; Haughton and Khandker 2009). relative to the poverty line. In agricultural economies in particular income is The growth elasticity of poverty (GEP) is the more volatile and more highly affected by the percentage reduction in the poverty rate that is growing and harvest seasons, so that relying on associated with a percentage change in mean income as an indicator of welfare might under- or (per capita) income. A numerical example overestimate living standards significantly. ­ clarifies the concept. In Ghana between 2005 and Consumption is a better measure of long-term 2012 the GEP assumed value −0.7 (see chapter 2). welfare also because households can borrow, draw This value implies that a 1 percent increase in per down savings, or receive public and private trans- capita income was associated with a 0.7  percent fers to smooth short-run fluctuations. The GLSS decrease in the poverty rate. collects sufficiently detailed information to facili- The growth incidence curve plots the growth tate estimates of the total consumption of each rate at each quintile of per capita consumption household. It relies on consumption per adult (or income). Graphs of growth incidence curves equivalent to capture differences in need by age allow us to compare differences in the incidence and economies of scale in consumption. Scales of of growth between poorer and richer segments of consumption by age and sex are computed by the population or with the rate of growth of mean the GSS. consumption (or income) (see chapter 2). The terms ‘poverty’ and ‘incidence of The cumulative consumption curve shows the ­ poverty’ refer to the poverty headcount index, level of welfare enjoyed by various percentiles Introduction 3 of  the population in any particular year. differences in the effects of these determinants Distributions to the right on the curve reflect an (coefficients effect), on the other. improvement in the overall welfare of the popu- Polarization is the combination of the diver- lation, that is, they are statistically dominant and gence from the global mean income and the con- can be considered. The horizontal axis represents vergence toward local mean incomes. Polarization consumption measured as a percentage of the differs from inequality because the latter is the poverty line. The vertical axis represents the overall dispersion of the distribution, that is, the percent share of the population, and each point distance of every individual from the median or on the distribution function shows the share of mean income. In income-polarized societies, peo- the population below a certain percentage level ple are clustered around the group means and tend of the poverty line. The point on the vertical axis to be remote from the mean or median of the over- that corresponds to the vertical line that indi- all distribution. Within each group, there is income cates 100 percent of the poverty line yields the homogeneity and often narrowing income inequal- poverty rate. There are several measures of the ity. Thus, we may talk of increasing identification. vulnerability to poverty. In this report, we use a Between the two groups, we talk, rather, about simple, unsophisticated ­ measure: 140 percent increasing alienation (Duclos, Esteban, and Ray and 180 percent of the poverty line. 2004). The overall impact of the forces of identifi- The Oaxaca decomposition (Oaxaca 1973), cation and of alienation between two groups of sig- explains the gap in the means of an outcome nificant size leads to effective opposition, a situation variable (consumption in our case) between that may give rise to social tensions and conflict two groups (e.g., between two survey rounds, (Esteban and Ray 1999, 2008, 2011). Also, the 1991 and 2012). The gap is decomposed into the group at the top of the distribution possesses voice, part that is due to group differences in the mag- while the other group, which is made up of those at nitudes of the determinants (endowments effect) the bottom, is voiceless in matters that affect the of consumption, on the one hand, and group welfare and society at large. Notes 1. For comparing the 4 survey rounds analyzed in internal and external consistency and harmoniza- ­ overty the report, the 1999 poverty line and food p tion across a common set of variables. line were used. See also GSS (2014). 3. Alternatively, the extreme poor are those people 2. The GLSS is also part of the Survey-Based whose standard of living is insufficient to meet Harmonized Indicators Program, which is run by basic nutritional requirements even if they were to the World Bank to combine and harmonize house- devote their entire consumption budgets to food. hold surveys across various African countries, 4. The old poverty line was C | 370.89 per adult equiv- including Burkina Faso, Cameroon, Ethiopia, alent per year in 2005 prices. The food poverty line The Gambia, Ghana, Kenya, Madagascar, Malawi, | 288.47 per adult equivalent per year in 2005 is C and Zambia. The program involves verification of prices. See also GSS (2014). 4 Introduction Chapter 1 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity The economic development of Ghana has been Two Decades of Expanding a tale of success over the last two decades. The Prosperity for All country has achieved sound economic growth and a stout increase in consumption among the Ghana has posted a solid growth performance bottom 40 percent of the consumption distribu- during the last two decades. After over a decade tion (the bottom 40). It has cut poverty by half of stable annual growth at between 4 and and considerably reduced vulnerability. Strong 5 percent, gross domestic product (GDP) began growth in income and consumption has been to pick up in the early 2000s and reached a steady accompanied by substantial improvement in growth rate of nearly 8 percent after 2006. There nonmonetary indicators of living standards. was then an impressive peak in 2011 mainly Newborns in Ghana today are expected to live because of the discovery of oil and the rebasing two years longer than newborns in 2005, and of GDP.1 Since 2008, Ghana has grown more children are more than twice as likely to be quickly than other African economies and, since enrolled in secondary school. They live in houses 2010, more quickly than the average among that are more than twice as likely to have electric- lower-middle-income countries (figure 1.1). ­ ity and improved sanitation facilities. The striking It  reached the middle income status in 2010. progress made in providing better opportunities These achievements have been important espe- for all is more than an achievement in its own cially in light of concurrent external shocks, such right. It has also strengthened the prospects for as the global financial crisis of 2008–09, reduced strong and inclusive growth in the future. trade revenues, and fluctuating oil prices. Figure 1.1  Real GDP Growth, 1991–2012 16 14 12 10 Percent 8 6 4 2 0 –2 94 04 10 09 99 00 98 03 93 08 92 12 96 02 06 11 01 91 95 05 97 07 20 20 20 19 20 19 20 19 19 19 20 19 20 20 19 20 20 19 20 19 20 20 Ghana Lower-middle income (worldwide) Sub-Saharan Africa (developing only) Source: World Development Indicators (WDI) database. A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 5 Table 1.1  Mean and Median Real Consumption (Adult Equivalent), 1991–2012 Cent Indicator 1991 1998 2005 2012 Mean Nationwide 588 719 952 1,132 Urban 822 953 1,281 1,397 Rural 464 569 702 805 Median Nationwide 435 542 706 834 Urban 640 778 1,005 1,070 Rural 355 435 553 594 Source: Calculations based on GLSS 5–6. Figure 1.2  Growth Incidence Curves, 1991–2012 and 2005–12 a. 1991–2012 b. 2005–12 National National 5 5 4 4 3 3 2 2 1 1 0 0 10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100 Percentiles Percentiles 95% confidence bounds Growth rate Growth rate in mean Growth rate at median Source: Calculations based on GLSS 5–6. The impressive GDP growth of the last growth rate was greater than the national average decade was associated with a substantial increase between 1998 and 2005, but slowed significantly in average household consumption. Mean and between 2005 and 2012. In rural areas, growth median real consumption almost doubled, with slowed during the latter period, but was, overall, the most rapid growth occurring between 1998 more rapid than that in urban areas. and 2005, when consumption expanded by about The benefits of growth were broadly shared 30 percent (table 1.1). The rates of growth in rural among the population, although the top per- and urban areas were almost the same in centile received relatively more. The growth 1991–2012, though there were substantial differ- incidence curves in figure 1.2 show the varia- ­ ences across subperiods. In urban areas, the tion  in  consumption between 1991 and 2012 6 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity (top quadrant) and between 2005 and 2012 (bottom 31.3 ­percentage points, from 52.7 percent in 1991 to quadrant) across percentiles of the distribution. 21.4 percent in 2012.2 The country seems on track Overall, the growth of consumption in 2005–12 was to reduce the poverty rate by half in line with less than the growth over the whole period: an aver- Millennium Development Goal 1. Ghana’s perfor- age rate of 2.5 percent a year against 3.2 percent. On mance in reducing poverty compares well with that the basis of their performance, three subgroups of other African countries. In 2012, the rate of pov- stand out: the top 20 percent of the distribution (the erty in Ghana was less than half the African average top 20), the bottom 25 percent (the bottom 25), and of 43 percent, while in 1991, it had been only 10 the percentiles in between. The top percentiles grew percent lower.3 consistently more rapidly than the national average, However, the highest poverty reduction, and this contributed to a widening in inequality (see 13.0 percent, was recorded between 1991 and below). The bottom 25 performed less well than the 1998. Since then, the speed of poverty reduction rest over the whole period, but its performance has declined, to 11.0 percent in 1998–2005 and to improved, in relative terms, in the last decade: it 7.1 percent in 2005–13. This is despite increasing continued to grow at about 2.5 percent a year, GDP growth (see figure 1.1). which, over the decade, coincided with average con- Extreme poverty declined more rapidly. The sumption growth. The middle percentiles grew at share of the population with consumption below rates slightly below the mean (2.8 percent versus the food poverty line declined from 37.6 percent 2.9 percent), but still at a sufficiently high rate to jus- in 1991 to 9.6 percent in 2012 (figure 1.3). tify considering their performance particularly Rapid growth in average consumption was the good. This group roughly corresponds to those peo- driving force behind the impressive poverty ple who, in 1991, were poor or vulnerable to poverty reduction. Figure 1.4 decomposes the change in and who had managed to move out of poverty (or, the incidence of poverty experienced between 1991 among some of them, out of vulnerability) by 2012 and 2012 into a component caused by the expan- (see subsequent chapters). sion in average household consumption and a com- ponent caused by changes in inequality (Kolenikov and Shorrocks 2003). The sharp decline in poverty Shared Prosperity Led to a was clearly driven almost exclusively by the increase Reduction in Poverty and in average household consumption. Inequality changed little over the period, and the impact of Vulnerability growth was large enough to offset the potential rise Poverty has been cut in half over the last two in poverty associated with the small increase in decades. The estimated poverty rate fell by inequality ­ experienced after 1998. Figure 1.3  Trends in Poverty and Extreme Poverty, 1991–2012 52.7 50 37.6 39.8 40 Percent 27.1 28.5 30 21.4 18.1 20 9.6 10 0 1991 1998 2005 2012 Poverty Extreme poverty Source: Calculations based on GLSS 3–6. A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 7 The important impact of economic growth and the rest of the developing world (ROW) also on poverty reduction is reflected in the compar- shows that the elasticity was higher in Ghana than atively high poverty elasticity of economic in the rest of Africa and closer to the elasticity of growth. Overall, the poverty elasticity of growth other developing countries in 1998–2005, but has over the last decade has been low in Africa com- since fallen dramatically also in relative terms. pared with the rest of the developing world, but The depth of poverty has also declined. Both Ghana has been an exception. It experienced an the poverty gap and the severity of poverty fell average ­elasticity of 1.36 over 1991–2012 (figure quickly, in particular until 2005 (table 1.2). This, 1.5). However, the value declined from well above in combination with the outcomes among the 2.0 to a more modest 0.7 after 2005. The compari- extreme poor, suggests that important changes son with the averages of Africa (SSA in the figure) occurred to the consumption of people living below the poverty line (figure 1.6). The share of Figure 1.4  Growth, Inequality, and Poverty people living below the poverty line by a certain Decomposition, 1991–2012 fixed amount (80 percent or 60 percent) can be used to measure the depth of poverty; 80 percent 5 of the poverty line is close to the value of the food 0 poverty line. The reduction in the incidence of Variation in poverty incidence –5 poverty was accompanied by a sharp drop in the –10 depth of poverty because the share of people liv- –15 ing below 60 percent or 80 percent of the poverty –20 Table 1.2  Poverty Rates, the Poverty Gap, –25 and the Severity of Poverty, 1991–2012 –30 Severity of Year Poverty rates Poverty gap poverty –35 1991 52.7 19.1 9.1 –40 1998 39.9 13.9 6.6 Growth Inequality Poverty contribution contribution variation 2005 28.5 9.6 4.6 1991–98 1998–2005 2005–12 2012 21.4 6.6 3.0 Source: Calculations based on GLSS 3–6. Source: Calculations based on GLSS 3–6. Figure 1.5  Growth Elasticity of Poverty, 1991–2012 0 –0.5 –0.7 –0.8 –1.0 –0.9 –1.5 –1.6 –2.0 –2.0 –2.2 –2.5 –2.4 –2.5 –2.5 –3.0 GHA SSA ROW 1991–1998 1998–2005 2005–2012 Source: Calculations based on GLSS 3–6 and WDI. Note: GHA = Ghana; ROW = rest of the world; SSA = Sub-Saharan Africa. 8 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity Figure 1.6  Vulnerability as Distance from the Poverty Line, Nationwide, 1991–2012 1 0.8 Share of population 0.6 0.4 0.2 0 20 60 80 100 140 180 220 260 300 % of poverty line 1991 1998 2005 2012 Source: Calculations based on GLSS 3–6. line narrowed considerably. That the distance of around 30 percent. The decrease in the num- from the 1991 incidence curve is greater around ber of the vulnerable was particularly significant the poverty line suggests that the expansion in in 1991–98 and 1998–2005. consumption was most rapid among those Vulnerability is the norm in rural areas, households with consumption levels closer to the and  the progress in urban areas has slowed poverty threshold. This is consistent with the appreciably since 2005. A rural-urban disaggre- conclusion that the extreme poverty rate fell gation shows that the increase in consumption in more quickly than the poverty rate. urban areas occurred mostly in 1991–98 and, Vulnerability remains widespread despite especially, in 1998–2005, but slowed consider- the recent increase in the number of Ghanaians ably in 2005–12, that is, the curves for the last with consumption levels close to the poverty two periods almost overlap (figure 1.7). By con- line. Non-poor households with consumption trast, in rural areas, progress was more modest, levels slightly above the poverty line are often but more evenly distributed. However, more than labeled vulnerable because even comparatively half of rural households are still extremely small shocks may push them into poverty. In the ­ vulnerable to shocks, that is, their levels of profile of the vulnerability of the Ghanaian pop- consumption are less than 140 percent of the ­ ulation, vulnerability thresholds of 140 percent poverty line. and 180 percent of the poverty lines are particu- larly relevant because a loss of less than US$0.50 a day in consumption can push households below The Poor over Time: Better the national poverty line of around US$1.30 a day. The cumulative curves of consumption in Attributes and Living Standards figure 1.6 show that the share of the population Poverty has always been predominantly rural. living under 140 (180) percent of the poverty line In recent years, however, poverty has become declined from around 73 (83) percent in 1991 to even more concentrated in rural areas. By 2012, 39 (51) percent in 2012. In both cases, the drop poverty rates were 38.2 percent in rural areas was slightly greater than the reduction in poverty and 10.4 percent in urban areas, and the median A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 9 Figure 1.7  Vulnerability and Poverty Depth as Distance from the Poverty Line, Rural and Urban Areas, 1991–2012 Rural Urban Share of population for both Share of population for both 1 1 0.8 0.8 rural and urban rural and urban 0.6 0.6 0.4 0.4 0.2 0.2 0 0 20 60 100 140 180 220 260 300 20 60 100 140 180 220 260 300 % of poverty line % of poverty line 1991 1998 2005 2012 Source: Calculations based on GLSS 3–6. consumption per capita was almost two times north in 2012, but only 17 percent of the greater in urban areas than in rural areas. For ­ population. The joint profile of residence in the every poor person in urban areas, there were north and residence in rural areas accounted for nearly four poor people in rural areas. The share 34  percent of the poor in 2012, but only 20 of the poor among rural residents did not percent in 1991. change much between 1991 and 2012, but the Poor households tend to be larger than non-​ share of rural residents in the total population poor households. In 2012, poor households had fell. Whereas in 1991, 67 percent of the total an average of 7.5 members, nearly 50 percent population was living in rural areas, the share more than non-poor households, and, on aver- had dropped to 50  percent by 2012. The inci- age, they also had more children (table 1.3 and dence of poverty was especially high among box 1.1). In  poor households half of the mem- people living in the rural savanna. These people bers  are below 14 years of age, and 20 percent accounted for more than 40 percent of the over- are  under 5  years of age, while among non-­ all poverty rate. poor  households, the shares are, respectively, The poor live mostly in the north. The trends 40 percent and 10 percent. in the number of the poor by region confirm that The heads of poor households are likely to their concentration has been relatively greater in be older men who are self-employed. The aver- the north than in the rest of the country. This is age head of a poor household is approach- caused by the combination of less favorable cli- ing  50  years of age—3.5 years older than the mate, distance from the sea, and lack of infra- average head of a non-poor households—and is structure. However, these disadvantages appear ­ considerably less likely to be a woman (by to have grown over the last two decades to the 33 percent). Some 89 percent of heads of poor extent that, whereas both the poverty rate and households are self-employed, 78 percent of the absolute numbers of the poor have declined them in agriculture. This compares with, respec- in the more populous southern and central tively, 66 percent and 35 percent of heads of non-­ regions, the number of the poor has risen in the poor households. Agricultural self-employment Northern Region and Upper West since 1991 remains the prevalent economic activity among (figure 1.8). As a result of these divergent trends, household heads in the bottom three quintiles nearly 40 percent of the poor were living in the (figure 1.9). 10 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity Figure 1.8  Trends in Regional Poverty Headcounts, 1991–2012 Number of poor individuals (thousands) 1,500 1,000 500 0 Western Central Greater Volta Eastern Ashanti Brong Northern Upper Upper Accra Ahafo East West 1991 1998 2005 2012 Source: Calculations based on GLSS 3–6. Table 1.3  The Profiles of Poor and Non-Poor Households, 1991–2012 Poor Non-poor Household characteristic, % unless indicated 1991 1998 2005 2012 1991 1998 2005 2012 Age of household head, mean 47.5 48.3 48.4 49.6 44.3 45.7 45.8 46.1 Sex of household head Female 30.5 34.0 15.5 18.2 42.5 42.8 26.4 26.8 Male 69.5 66.0 84.6 81.8 57.5 57.2 73.6 73.2 Makeup of household Mean number of members 7.1 6.6 7.7 7.5 5.3 5.2 5.1 5.2 Dependency ratio, age 0–14 and 65– relative to 15–64 0.5 0.5 0.5 0.5 0.5 0.5 0.4 0.4 Share of 0–14 age-group 0.5 0.5 0.5 0.5 0.4 0.4 0.4 0.4 Share of 0–4 age-group 0.2 0.2 0.2 0.2 0.1 0.1 0.1 0.1 Highest educational attainment, household head No education, other, not known 55.3 49.6 58.7 52.2 35.2 26.2 26.7 21.2 Primary 11.4 18.8 15.3 22.5 11.4 17.8 15.1 20.3 Secondary 32.1 29.6 25.6 24.1 48.9 48.9 50.5 50.0 Higher 1.2 1.9 0.5 1.2 4.6 7.1 7.7 8.5 Household head, occupation Wage, private sector 3.9 3.5 5.1 6.8 8.7 10.3 14.9 17.5 Wage, public sector 8.6 4.9 2.1 2.1 18.2 12.1 9.9 8.3 Nonagricultural self-employment 13.1 15.1 9.0 10.9 29.7 32.9 26.3 31.1 Agricultural self-employment 72.4 72.6 80.7 77.6 35.9 35.1 41.6 35.1 Other, including unemployed 2.0 3.9 3.0 2.6 7.4 9.8 7.4 8.0 Region of residence Western 11.3 6.9 6.6 7.8 8.4 13.7 11.5 9.6 Central 8.7 14.6 6.1 6.7 12.2 9.6 9.8 9.5 table continues next page A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 11 Table 1.3  continued Poor Non-poor Household characteristic, % unless indicated 1991 1998 2005 2012 1991 1998 2005 2012 Greater Accra 5.8 2.8 5.8 3.9 18.4 22.9 17.2 19.7 Volta 7.5 17.4 8.3 11.9 9.0 12.2 7.1 7.8 Eastern 15.5 10.2 6.9 8.6 12.9 10.7 16.0 11.0 Ashanti 12.4 15.9 12.1 12.1 19.6 19.1 18.7 21.6 Brong Ahafo 14.4 7.0 9.5 10.6 8.6 7.4 9.0 9.6 Northern 11.3 11.2 22.0 21.8 7.4 3.2 8.1 6.8 Upper East 9.1 4.6 11.8 7.7 1.3 0.4 2.0 3.1 Upper West 4.0 9.4 11.0 9.1 2.2 0.8 0.6 1.2 Area of residence Rural 82.4 81.3 85.8 80.1 50.5 53.7 52.9 41.9 Urban 17.6 18.7 14.2 19.9 49.5 46.3 47.1 58.1 Water supply Protected 38.0 49.0 69.4 71.3 55.8 62.0 76.8 60.4 Unprotected 60.4 51.0 30.5 28.6 43.0 38.0 23.2 39.5 Vendor, truck 0.9 0.8 0.4 3.6 2.0 5.3 4.0 28.5 Connection of electricity in dwelling None 86.7 82.3 79.8 61.9 56.3 46.3 44.5 25.7 Connected 13.3 17.8 20.2 38.1 43.7 53.7 55.5 74.3 Main cooking fuel Firewood 86.1 86.5 83.5 85.3 55.4 50.9 52.6 40.6 Other 13.9 13.5 16.5 14.7 44.6 49.1 47.5 59.4 Main toilet facility Toilet, improved pit latrine 6.1 20.8 19.1 33.2 20.4 41.5 50.1 63.4 Other 93.9 79.2 80.9 66.8 79.7 58.6 49.9 36.6 Ownership of durables Radio and television 36.2 43.0 73.2 65.1 55.0 66.0 80.4 72.7 Television 3.5 8.0 8.7 25.6 20.8 36.3 39.9 64.0 Bicycle 28.0 27.4 48.3 39.6 17.8 17.2 23.5 23.6 Motorcycle 0.0 0.8 4.3 11.3 0.0 2.0 3.7 10.0 Source: Calculations based on GLSS 3–6. The poor are largely unskilled. Over have some connectivity. A majority of poor house- 50 percent of the poor in 2012 had no education, holds did not have access to electricity (62 percent) compared with only 21 percent of the non-poor. or adequate sanitation (69 percent), and 85 percent An additional 23 percent of the poor had only relied on firewood for cooking. This is in sharp primary education, compared with 20 percent of contrast to the availability of services of the non-​ the non-poor. In addition, the proportion of poor, which mostly enjoyed centrally provided the  poor without education has declined little electricity (74  percent) and adequate sanitation since 1991 while it has sharply declined among (63 percent) and used fuels other than firewood for the non-poor. cooking (59 percent). Nonetheless, even the poor In 2012, the majority of the poor still had lit- enjoy some degree of connectivity with the outside tle access to basic infrastructure, though they did world at least via radio. 12 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity Figure 1.9  Main Sectors of Employment, Household Heads, by Quintile, 2012 Richest Fourth Third Second Poorest 0 20 40 60 80 100 Percent Public wage Private wage Self-employment, nonagricultural Self-employment, agricultural Source: Calculations based on GLSS 6. BOX 1.1  How Has the Typical Poor Household Changed between 1991 and 2012? ­ ypical Despite the considerable socioeconomic changes Ghana experienced in the 1991–2012 period, the t poor Ghanaian household in 1991 and the typical poor household in 2012 were not dissimilar: 1. Each had seven to eight members; 2. They were both headed by a man approaching 50 years of age who had no education and worked in agriculture; 3. They had no access to electricity or adequate sanitation and used firewood for cooking; and 4. They were connected to the outside world through the radio (see table 1.3). However, a more careful analysis reveals that, compared to its 1991 counterpart, the typical poor house- hold in 2012 was considerably better off in nonmonetary indicators. It had: 1. Considerably better access to basic services and infrastructure; 2. Higher ownership of durables; and 3. Nearly twice the probability of enjoying some level of educational attainment. Better and More Equal measured through monetary and consumption indicators of household welfare. However, income Opportunities for Sustainable or consumption alone do not provide a complete Progress picture. The nonmonetary dimensions of depriva- Welfare includes more than reduced poverty and tion are also important not only in their own greater consumption among the less well off. right,  but also because they are associated with Our analysis has so far focused mainly on progress inequalities of opportunity that can exacerbate A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 13 disadvantages in income or consumption. Among However, there are many persistent chal- these dimensions, asset accumulation, in particu- lenges in the education sector. Enrollment rates lar, is an important indicator of how economic in senior-secondary school and tertiary education growth is shared across a population and of the remain low, particularly among the bottom success or failure of a sustainable poverty reduc- 20  percent of the consumption distribution (the tion strategy. Likewise, improvement in human bottom 20) and among rural residents. In 2012, capital accumulation and in access to basic services only 1 percent of youth living in poor households are also key components of welfare. Ghana has had attained tertiary education, compared with achieved good progress on these indicators, too. 39 percent among youth in the 5th (richest) quin- Average educational attainment has tile. Moreover, the average child in the richest improved considerably in Ghana. By 2012, the quintile was more than twice as likely as the aver- youth literacy rate had reached 79 percent, up age child in the poorest quintile to complete 24 percentage points since 1991, and the gender primary school, 69 percent compared with ­ gap in this indicator had narrowed (table 1.4).4 30  percent. Small but persistent gender gaps in Overall, adult literacy had also improved, and, by educational attainment are also characteristic 2012, Ghana ranked 15th in Africa, with about of  the bottom income groups. In addition the 75  percent of the adult population considered quality of learning in both basic and post-basic ­literate. The net enrollment rate in primary school education remains low and the education system rose from 55  percent in 1991 to 75 percent in has ­ limited  capacity to create relevant skills for 2012. increasing  competitiveness and productivity. ­ Table 1.4  Educational Attainment, Bottom 20 and Top 20, by Gender, 1991–2012 1991 2005 2012 Level Bottom 20 Top 20 Bottom 20 Top 20 Bottom 20 Top 20 Gross primary enrollment ratio Female 65.7 96.2 97.4 117.3 97.6 113.7 Male 78.5 105.1 98.7 121.9 113.5 120.1 Gross lower-secondary enrollment ratio Female 28.9 57.1 43.8 65.8 56.1 95.5 Male 41.2 76.9 45.4 71.1 54.1 98.3 Gross senior-secondary enrollment ratio Female 5.0 24.1 7.9 31.4 14.8 71.5 Male 13.0 39.6 15.0 37.8 18.7 92.2 Net primary enrollment rate Female 42.7 59.7 57.4 77.6 59.4 84.9 Male 43.6 65.8 55.3 75.8 62.4 88.4 Net lower-secondary enrollment rate Female 13.3 28.0 16.6 40.2 19.8 65.5 Male 18.3 30.0 18.3 36.6 16.7 60.2 Net senior-secondary enrollment rate Female 3.0 7.5 5.5 19.0 6.1 39.4 Male 5.7 7.9 6.7 22.4 6.5 30.1 Youth literacy rate (ages 14–25) a Female 42.7 60.7 59.0 77.6 59.8 90.2 Male 47.6 68.3 66.3 87.1 68.5 93.6 Source: Calculations based on GLSS 3–6. a. The total number of literate females or males (based on self-reporting) aged 14–25 expressed as a percent of total females or males in the 14–25 age-group. The 1998 enrollment rates are not comparable with the rates in the 1991, 2005, and 2012 surveys because the years of education were not captured in the 1998 survey. 14 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity Table 1.5  Infant and Under-5 Mortality, Vaccination, and Fertility Rates, 1988–2014 Rate 1988 1993 1998 2003 2008 2014 Infant mortality 77 66 57 64 50 41 Under-5 mortality 87 57 54 50 31 19 Vaccinationa — 54 62 69 79 84 Fertility 6.4 5.2 4.4 4.4 4.0 4.2 Source: Demographic and Health Surveys 1988–2015: STATcompiler (DHS Program STATcompiler) (database), ICF International, Rockville, MD, http://www.statcompiler.com/. a. Children aged 12–23 months are fully vaccinated, that is, they have received BCG, measles, and three doses of DPT and polio vaccine (excluding polio 0) according to their vaccination cards or the reports of their mothers. Figure 1.10  Access to Basic Services, 1991–2012 80 70 60 Proportion of popul tion (%) 50 40 30 20 10 0 El ctricit Improv d nd flush toil t G rb coll ction 1991 1998 2005 2012 Source: Calculations based on GLSS 3–6. About two-thirds of the students who complete by more than half, from 54 to 19 deaths per 1,000 primary school do not attain proficiency in core live births (table 1.5). This is partly the result of the subject areas and there are large disparities between efforts of the government to raise vaccination rates the 100 most highly subscribed senior high schools among all children. About 84 percent of children (roughly 20 percent of students) and the rest of the across the country have been immunized.5 sector. The top 10 percent of schools produce 90 The fertility rate is declining. This has led to a percent of the students entering university while reduction in the dependency ratio. However, in students with low WASSCE examination marks 2008, fertility rates varied considerably across enroll in some form of training or apprenticeship Ghana: they were higher in rural areas—4.9 births program (World Bank, forthcoming). per woman, compared with 3.1 in urban areas— Newborns in Ghana today are expected to live and in the Northern and Upper West regions. two years longer than newborns in 2005, and There was substantial improvement in girls have higher life expectancy than boys. Over access to basic household services. The last the last decade, infant mortality declined by about decades have seen improvements in access to 30 percent, from 57 deaths per 1,000 live births in sanitation, electricity, and clean drinking water 1998 to 41 in 2014, and under-5 mortality decreased (figure 1.10). The striking expansion in the A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 15 coverage of garbage collection after 2005—from Figure 1.11  Asset Index, Rural and Urban less than 10 percent to about 60 percent—reflects Areas and Nationwide, 1991–2012 the efforts of the government to provide such a. Rural services to the urban poor through subsidies. The success of the policy has made Ghana a 1 Share of population leader in sanitation in Africa. 0.8 The positive trends in household consump- 0.6 tion and in other dimensions of household 0.4 welfare over the last decade can be summa- 0.2 rized through an asset index.6 The cumulative 0 distributions of the asset index highlight the –1.5 –1 0 1 2 rapid pace at which households along the entire Asset index values population distribution have accumulated assets b. Urban over the two decades. Thus, the curve for 2012 1 lies well to the right of the curves for previous Share of population years (figure 1.11). The progress was particularly 0.8 striking after 2005, reflecting the lagged effects 0.6 of the substantial consumption growth between 0.4 1998 and 2005 and the sustained public 0.2 investment in infrastructure, including roads, ­ 0 electricity grids, and sanitation infrastructure –1.5 –1 0 1 2 and services. Asset index values A comparison confirms the large divide c. Nationwide between urban and rural areas. The urban dis- 1 tribution is always to the right of the rural Share of population 0.8 distribution, and the probability density is ­ always  much higher above 0 than below 0 (see 0.6 ­ figure 1.11). By contrast, despite recent improve- 0.4 ments, the rural probability density continues to 0.2 lie mostly below 0. However, the considerable 0 shift to the right of the rural distribution shows –1.5 –1 0 1 2 that the economic growth of the last two decades Asset index values has been inclusive and has gone some way to 1991 1998 2005 2012 reducing inequalities in opportunity. Source: Calculations based on GLSS 3–6. Progress in the non-income dimensions of household welfare strengthens the prospects for  future growth and poverty reduction by exacerbate income disadvantages by limiting the improving the opportunities for all. Empirical ­ potential of individuals from birth, and may have evidence from across all the regions of the world long-term negative impacts on the potential for shows that disadvantages in the accumulation growth and poverty ­ reduction within countries. of  physical and human capital and in access to Ghana has taken important steps toward breaking ­services—inequality of opportunity—are p ­ ervasive, this vicious circle. Notes 1. In 2010 Ghana changed its base year from 1993 $13 billion (£8 billion) of economic activity had to 2006, and this led to a jump in GDP and been missed. As a result, Ghana was upgraded from conclusion that, in previous estimates, about the  ­ a low-income to a lower-middle-income country. 16 A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 2. See Introduction on the poverty line selected. in the Ashanti, Brong Ahafo, and Greater Accra 3. The average rate for Africa is calculated using the regions. Progress in immunization has been wide- international poverty line of US$1.25. spread, although coverage in the Northern Region 4. The youth literacy rate as used here is the percent was below 60 percent in 2008. of the population aged 18–24 years that can, with 6. The asset index, suggested by Filmer and Pritchett understanding, read and write a short, simple (2001), uses principal-components analysis to cal- statement about their everyday lives. The adult lit- culate the weights of the index. The first principal eracy rate is based on the same criterion except the component, the linear combination capturing population of reference is aged 15 years or above. the greatest variation among the set of variables, 5. Rural areas show infant (under-5) mortality rates can be converted into factor scores, which serve of 56 (90) per 1,000 live births, compared with as weights. The rationale for using this index is 49 (75) in urban areas. Infant (under-5) mortality that it captures the household’s permanent wel- rates are 70 (137) and 97 (141) per 1,000 live births fare dimension more effectively relative to simple in the Northern and Upper West regions, respec- consumption data and can provide more reliable tively, but well below 40 (65) per 1,000 live births rankings across households. A Tale of Success: Shared Prosperity, Poverty Reduction, and a Boost in Opportunity 17 Chapter 2 Drivers of Poverty Reduction The improvements in household consumption composition were also important. Moreover, and poverty reduction described in chapter 2 Falco et al. (2014), in an econometric study that occurred during a period of rapid change in uses panel data to trace the earnings of individu- the economic and sociodemographic structure als over time, find that the main determinants of of the economy. In this chapter, we look more both earnings and the growth of earnings over closely at factors associated with poverty reduc- the life span are type of job—defined as in this tion, focusing on three key developments of the chapter—and level of education. They also find past decade: (1) structural transformation, that beginning working life in a low-paying activ- (2)  increased skills among the labor force, and ity such as agriculture has a scarring effect by (3) urbanization. reducing earning prospects for the rest of the Underlying the choice of these factors is a worker’s life. simple analytical framework according to The structure of the chapter is as follows. In which per capita household consumption is the first section, we describe three key develop- driven by household composition, income ments in Ghana over the past decade. The sec- from labor and other sources, and prevailing ond section relies on GLSS data to examine how consumption patterns. Nonlabor income is usu- these developments have been associated with ally derived from public and private transfers poverty reduction and consumption growth and returns to capital. With the possible excep- and, thereby, identifies the main drivers behind tion of remittances, transfers and returns to cap- these phenomena. ital are virtually nonexistent in Ghana, as they are in most other African countries. Thus, work is the main source of income, especially among Key Economic Developments the poor. Many families escape (or fall) into pov- erty because family members obtain (or lose) Structural Transformation jobs or because the returns from work are aug- Ghana’s recent economic growth has been mented (or reduced) by factors within their con- associated with a shift of the economy out of trol (investments in education, migration) or by agriculture into services. The share of agricul- exogenous shocks (poor or abundant rainfall). ture in GDP declined by nearly 50 percent, from However, in Ghana, having a job is often not over a third of GDP in 1991 to 23 percent in 2012 sufficient to bring workers out of poverty (figure 2.1). As a result, by 2011, agriculture was because the returns to work tend to be extremely the smallest sector in the economy in terms of low. It is growing earnings from work that make value added, and its share has continued to a difference in the effort to escape poverty. Family decline since. Meanwhile, the service sector composition and demographics are also impor- expanded to nearly half of GDP, from an initial tant because they affect the dependency ratio, 34 percent in 1991. that is, the number of consumers relative to the Services account for more than half of per number of earners in the household. Azevedo et capita GDP growth. The main driver of the rela- al. (2013), in a quantitative analysis, suggest that tive growth of the service sector, accounting for changes in labor income accounted for nearly 23.9 percent of total GDP growth, was a hetero- half the reduction in poverty in Ghana between geneous group of “other activities” (table  2.1). 1998 and 2005, but that changes in household This included financial and business services, Drivers of Poverty Reduction 19 Figure 2.1  Sectoral Composition of GDP, 1991–2013 55 $4,500 50 $4,000 45 $3,500 40 $3,000 Percentage of GDP 35 $2,500 30 $2,000 25 $1,500 20 $1,000 15 $500 10 $0 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 GDP per capita, PPP (constant 2011 US$) Services value added Agriculture value added Industry value added Sources: National Accounts Main Aggregates Database; World Bank data. Table 2.1  GDP Growth Decomposition, by Per Capita Value Added, 1991–2012 Percent Contribution of Contribution of Contribution of Sector within-sector changes in changes in intersectoral Total output per worker employment shifts Agriculture 17.66 –16.97 11.64 12.33 Industry 4.83 8.94 7.40 21.17   Mining, manufacturing, utilities –0.38 4.97 2.90 7.50  Construction 3.57 3.96 6.14 13.68 Services 24.93 27.25 4.04 56.23   Wholesale, retail trade, restaurants, and hotels 4.45 16.03 −6.65 13.82   Transport, storage, and communications 1.67 3.84 12.98 18.48   Other activitiesa 13.41 7.39 3.13 23.93 Subtotals 47.42 19.22 23.08 89.72 Demographic component n.a. n.a. n.a. 10.28 Total 100.00 Total % change in value added per capita 92.06 Sources: National Accounts Main Aggregates Database; World Bank data. Note: n.a. = not applicable. a. Other activities include financial and business services, public administration, education, health care, social protection, and other services. public administration, education, health care, such as information and communications tech- social protection, and other services. This was fol- nology, finance and insurance, and real estate. lowed by “transport, storage, and communica- This growth took place mainly in Accra, which tion,” accounting for 18.5 percent of overall GDP saw a massive inflow of capital in the last decade, growth, and “wholesale, retail trade, restaurants, but also a surge in real estate prices. The other and hotels,” accounting for 13.8 percent. ­ service component is represented by those The boom in the service sector derived from ­ activities—borderline between formal and a number of factors. An important engine was ­ informal—that characterize West African towns: the rapid growth in high–value added services retail activities, construction, transport, and so on. 20 Drivers of Poverty Reduction Increased  public sector employment in educa- at which workers moved out of agriculture nearly tion, health care, and public administration also doubled after 2005, from 1.6 percent to 3.0 percent contributed to the service boom. a year (see table 2.2). However, unlike the shift in By contrast, the share of the industrial sec- GDP, workers gravitated mostly toward services tors in GDP increased, and, in 2011, for the (4.0  percent a year), while industry picked up first time since independence, it was larger only a small share of employment. than the share of agriculture. The main con- However, the major driver of growth has tributor, besides gold, was crude oil production. been within-sector increases in value added. Construction was also an important driver of The total contribution of the within-sector change GDP growth, contributing about 14  percent in value added per worker explains the 47.2 percent thanks to an increase in productivity and a posi- of per capita GDP growth (see table  2.1). Value tive intersectoral shift (see table 2.1). added per worker in agriculture rose in 1991–2012 The speed and nature of structural transfor- because of the outflow of workers and the devel- mation changed after 2005. The share of agricul- opment of commercial farming. It also increased ture in value added declined more quickly after in services. Agriculture and services explain, 2005, while the share of services remained virtu- respectively, the 37.2 percent and 52.6 percent of ally unchanged. This resulted in an expansion in the observed change in output per worker. the share of industry to 27  percent of GDP in Structural transformation alone, that is, the 2012 (see figure 2.1). intersectoral shifts, accounted for less than half The sectoral distribution of employment of GDP growth over the two decades. This is adjusted somewhat in line with the changes in likely linked to the relatively small expansion of the structure of the economy. Although agricul- employment in industry and the high concentra- ture remains the main sector of employment, at tion of workers in lower-productivity service 43.2  percent, workers have increasingly moved activities, where output per worker appears to away from this sector into services and, to a lesser have been less than in agriculture. extent, industry over the last decade. Employment The limited role of increased employment in the service sector expanded from 28.8 percent in in explaining growth is consistent with the 1991 to 42.0 percent in 2012, and the share of con- fact that most Ghanaians work. In 2012, nearly struction more than doubled over the two decades. 4 adults in 5 aged 15–64 and 9 adults in 10 aged The share of employment in industry also rose, over 25 were working (figure 2.2, panel a). As in although at a lower rate, from 10.7  percent to many African economies, the open unemploy- 14.9 percent (table 2.2). ment rate was low, around 2  percent of the The shift in the sectoral composition of labor force, and highly concentrated among the employment accelerated after 2005. The speed better-off. Table 2.2  Employment, by Major Sector, 1991–2012 (% of Total Employment) Sector 1991 1998 2005 2012 Agriculture 60.5 52.0 53.4 43.2 Industry 10.7 15.0 14.7 14.9   Mining, manufacturing, utilities 9.5 13.4 12.8 11.4  Construction 1.2 1.6 1.9 3.5 Services 28.7 33.0 31.9 42.0   Wholesale, retail trade, restaurants, and hotels 16.9 19.5 18.5 24.7   Transport, storage, and communications 1.9 2.6 3.1 4.0   Other activities 9.9 10.9 10.3 13.3 Source: Calculations based on GLSS 3–6. Drivers of Poverty Reduction 21 Figure 2.2  Poverty and Access to Employment, Adults Aged 15–64, 2012 a. Labor market status b. Employed, by consumption quintile 90 Total 15–64 80 70 60 40–64 Percent 50 40 25–39 30 20 15–24 10 0 0 20 40 60 80 100 Poorest Second Third Fourth Richest Percent Employment-to-population ratio Employed Unemployed Inactive Source: Calculations based on GLSS 6. It follows that having a job is not a guarantee productivity remains limited in many parts of the of escaping poverty. Indeed, the share of employed country because of traditional farming methods adults in the poorest consumption quintile and volatility in rainfall (Molini et al. 2010). (80 percent) is slightly higher than the corresponding Nonetheless, the poverty rate among cocoa share in the richest quintile (­ figure  2.2, panel b).1 farmers declined from 60  percent in 1991 to Instead, it is the types of jobs that are important in about 24 percent in 2005 (Breisinger et al. 2008). poverty reduction. Four broad types of employ- The positive performance of staple crops also ment are of importance: employment in agricul- helped boost consumption and reduce poverty. ture, nonfarm self-employment, wage employment Figure 2.3 shows production growth in a number in the private sector, and wage employment in the of key food crops between 2000 and 2010 using public sector. 2000 as a base year. The growth in the production Agriculture is dominated by l ­ ow-­productivity of rice, maize and, to a lesser extent, millet after smallholder farming and is mainly rainfed. 2007 is striking: relative to 2000, the production The  most competitive cash crop is cocoa beans of rice was 2.5 times greater; of millet, 1.5 times (Teal, Zeitlin, and Maamah 2006; Vigneri 2005); greater; and, of maize, almost 2.0 times greater. Ghana controls an average of 14.5 percent of the The rise in production was also rapid in other world market and is the third-largest producer. food crops, such as yams, groundnuts, soybeans, Nonetheless, the average yield (431.0 kilograms and cassava (Mohan and Matsuda 2013). Average per hectare in 2005–12) is low compared with the staple food-crop output grew much more quickly top 2 cocoa producers, Côte d’Ivoire (595.7 kilo- than the population, and per capita production grams per hectare) and Indonesia (576.0 kilo- was more than 80  percent greater in 2005–07 grams per hectare). Other cash crops include than in 1981–83. Growth in higher-value cotton, rubber, and tobacco, of which Ghana pro- ­vegetables and fruit for domestic and export mar- duces only a small share of the world output. Since kets was also encouraging (Breisinger et al. 2008). independence, raising agricultural productivity Nonfarm self-employment is dominated by and transforming agriculture from a subsistence household enterprises. It is the most important base to a market base have been priorities of the economic activity in urban areas and typically government. Various policies and interventions provides supplementary income to rural house- have been tried to boost cash crop production for holds. Despite its generally low ­income-generating internal and, especially, international trade. But capacity, it represents an initial opportunity to the results have been mixed, and agricultural quit agriculture among low-skilled youth and 22 Drivers of Poverty Reduction among adults who may not have adequate skills or restaurants, and hotels” in figure  2.4—are not experience to fill wage jobs, especially in areas ­necessarily a ticket out of poverty because the out- where medium or large private enterprises are put per worker is less than in agriculture. limited. However, some service sector activities Private and public sector wage jobs tend to with a high concentration of relatively low-skilled be concentrated in better-off urban areas and workers—grouped under “wholesale, retail trade, to attract a more highly skilled labor force by Figure 2.3  Production Indexes, Major Crops, 2000–10 300 250 200 Ind x 150 100 50 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Maize Millet Rice Sorghum Source: Ministry of Agriculture 2012. Figure 2.4  Output per Worker, by Sector, 1991–2012 12,000 10,000 Constant 2005 US$ 8,000 6,000 4,000 2,000 0 Agriculture, Mining, Construction Wholesale, Transport, Other hunting, manufacturing, retail trade, storage, and activities forestry, utilities restaurants, and communication fishing hotels 1991 2012 Sources: National Accounts Main Aggregates Database; GLSS 3 and 6. Drivers of Poverty Reduction 23 offering greater earnings and better working among the heads of households in the top 20, conditions. The public sector offers particularly 42 percent of whom are wage employees. advantageous conditions for the better educated. The recent growth in household consump- Even after we control for observed characteristics tion has been associated with a considerable and selectivity, wages among men appear to be growth in nonagricultural private employ- notably higher in public administration than in ment. Agriculture has given way to nonagricul- the private sector (Ranzani and Tuccio, forth- tural self-employment and, to a lesser extent, coming). This is a concern because the high private sector wage employment. Between 1991 returns to public employment may increase the and 2012, the share of workers in nonwage and reservation wage for employment in the private wage work outside agriculture expanded in sector, thereby inducing workers to wait in line about equal measure: from 26  percent to for government jobs and crowding out employ- 34  percent in self-employment and from ment in potentially more productive activities. 15 percent to 24 percent in wage work. There was The type of job done by the household head also a significant shift from public sector to pri- is a good predictor of poverty status. In 2012, vate sector wage employment. The share of pri- over 42 percent of the employed workforce relied vate sector wage employment tripled, from on low-productivity farming for income. Another 6 percent to 18 percent, while the share of public 34 percent were working for themselves or their sector wage employment fell, from 9 percent to families in household enterprises in the non- 6 percent. As a result of these trends, the share of farm sector. The wage sector remained small, agriculture in employment is now lower, and the accounting for about 24 percent of the workforce, share of nonfarm employment is now higher mostly in the private sector. In the north and than the average among lower-­ middle-income east, agriculture is the main source of income, countries in Africa (where it reaches 55, 31, and accounting for between 54 percent of income in 14 percent, respectively) (Filmer and Fox 2014). the Volta Region and 73  percent in the Upper However, the magnitude of the sectoral West Region (figure  2.5). The type of employ- shift in employment differed significantly ment greatly influences consumption levels. across regions. In Greater Accra, the only Agriculture was by far the dominant economic region in which agriculture had already contrib- activity among the bottom 20. By contrast, agri- uted little to employment in 1991, private sector culture accounted for only 14  percent of jobs wage employment expanded significantly at the Figure 2.5  Distribution of Employment, by Sector, Region, and Poverty Status a. By sector and region b. By sector and consumption quintile Nationwide Richest Upper West Nothern Upper East Fourth Brong Ahafo Volta Third Central Eastern Second Western Ashanti Poorest Greater Accra 0 20 40 60 80 100 0 20 40 60 80 100 Percent Percent Wage, private Wage, public Self, nonagricultural Self, agricultural Source: Calculations based on GLSS 6. 24 Drivers of Poverty Reduction expense of both public sector employment and The labor force has become better educated self-­employment. It now accounts for 42 percent over the past two decades. Between 1991 and of all jobs. Other southern regions (the Ashanti, 2012, the share of the labor force without school- Central, Eastern, and Western regions) wit- ing fell by almost half, from 41  percent to nessed a considerable shift away from agricul- 24  percent. In 2012, the majority of workers, ture toward wage employment, but the shift was 52  percent, had completed at least lower-­ smaller there than in the remaining regions of secondary education, compared with 39 percent the north (the Brong Ahafo, Northern, Upper in 1991. However, regional disparities persist East, Upper West, and Volta regions). figure  2.6). Despite some improvements, the (­ Northern, Upper East, and Upper West regions A More Skilled Workforce still have a basic education completion rate under Building a skilled workforce is an accumula- 25  percent, while, in Accra, 40  percent have tive process that depends on many years of senior-secondary or tertiary degrees, compared good-quality schooling and the creation of with 7  percent in 1991. The share of the work- job-relevant skills. Good-quality basic educa- force with tertiary education was practically tion is a fundamental requirement of skills nonexistent in 1991, but has now reached about development. Nonetheless, skills that are rele- ­ 8 percent. vant to labor markets are not synonymous with Education is an important correlate of job education. Skill is the ability to perform some ­ opportunities. The educational attainment of function (or ­ specific job) because of one’s knowl- the workforce has increased across all job types, edge (which may be acquired through ­education), though differences in educational attainment but also practice and aptitude. Yet, quality educa- across categories of work have also become more tion is an important stepping-stone to the acqui- accentuated. Over 50  percent of public sector sition of most job skills because it develops wage workers have tertiary education, compared foundational skills such as literacy and ­numeracy, with less than 20  percent among the nonfarm builds noncognitive skills, and familiarizes stu- self-employed (figure  2.7, panel a). More than dents with learning s­ ituations (Adams, Johansson half of the workers with tertiary education are de Silva, and Razmara 2013). employed in the public (wage) sector; another Figure 2.6  Labor Force with at Least Basic Education, by Region, 1991 and 2012 100 90 80 70 60 Percent 50 40 30 20 10 0 Greater Ashanti Western Central Eastern Brong Volta Upper Upper Nothern Nation- Accra Ahafo East West wide 1991 2012 Source: Calculations based on GLSS 3 and 6. Drivers of Poverty Reduction 25 30 percent in the private (wage) sector; and only areas in 1991–98; 1.9 million in 1998–2005; and 14  percent in nonfarm self-employment 4.7 million in 2005–12 (figure  2.8). By 2012, figure 2.7, panel b). Lack of education has acted (­ Ghanaians were equally split between urban and as a barrier to occupational mobility, but workers rural areas. This represented a considerable shift with primary education or above have been relative to 1991, when 70 percent of the popula- increasingly entering nonfarm employment, first tion was living in rural areas. Since 1991, Accra as self-employed and then as wage workers. and Ashanti have gained over 2.4 million inhabi- tants each, around half of them in the last decade. Urbanization and Agglomeration Migration is not the only possible explanation The spatial patterns of poverty reduction and for the growth of the urban population. There consumption growth in recent years suggest may also be other demographical phenomena at that location of residence is a major correlate play, for example, a fast increase of household of the risk of poverty and that geographical size in urban areas. However, data on the birth- mobility may be an effective strategy among place of household heads in rounds 5 and 6 of the individuals to exit poverty. Our analysis above GLSS and in related analysis support the rural-­ highlights the long-established spatial patterns of urban migration hypothesis (Molini, Pavelesku, poverty that the structural transformation and and Ranzani, forthcoming) (figure 2.9). economic growth of recent decades seem to have Urbanization is likely to have facilitated the reinforced. The patterns appear closely linked to process of structural transformation. Sustainable a widening gap in employment opportunities. economic transformation entails important socio- Moving to more rapidly developing areas in the economic changes correlated with urbanization. south has been a strategy for escaping poverty. For example, urbanization has enabled higher-­ The urban population has increased rap- quality education to reach a larger share of the idly, mostly because of internal migration. population. Larger urban areas have an advantage Around 1.5 million people flowed into urban over small towns and rural areas in their ability to Figure 2.7  Education and Employment, 1991 and 2012 a. Highest educational attainment, by sector b. Sector share of total, by educational level 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 1991 2012 1991 2012 1991 2012 1991 2012 1991 2012 1991 2012 1991 2012 1991 2012 1991 2012 Wage, Wage, Self, non- Self, None Primary Junior Senior Tertiary public private agricultural agricultural secondary secondary None Primary Junior secondary Wage, public Wage, private Senior secondary Tertiary Self, nonagricultural Self, agricultural Source: Calculations based on GLSS 3 and 6. 26 Drivers of Poverty Reduction Figure 2.8  Variation in Population, by Rural and Urban Area and Region, 1991–2012 W st rn C ntr l Gr t r Accr Volt Rur l E st rn Ash nti Bron Ah fo North rn Upp r E st Upp r W st W st rn C ntr l Gr t r Accr Volt Urb n E st rn Ash nti Bron Ah fo North rn Upp r E st Upp r W st –1,000 –500 0 500 1,000 1,500 2,000 2,500 Numb r of individu ls in thous nds 1991–98 1998–2005 2005–12 Source: Calculations based on GLSS 3–6. Figure 2.9  Household Heads Who Were Not Born in Their Current Places of Residence 1,000,000 800,000 Number of individuals 600,000 400,000 200,000 0 ra an an an an an an an an an cc b b b b rb b b rb rb Ur Ur Ur Ur Ur Ur rA iU tU tU rn n al lta o n nt te es as af er r tr te he ea Vo ha rW rE Ah st n es rt Ce Gr As Ea pe No W g pe on Up Up Br 2005 2012 Source: Calculations based on GLSS 5–6. Drivers of Poverty Reduction 27 provide greater access to secondary and tertiary extent to which each of these factors has ­contributed educational institutions and higher‐quality teach- to poverty reduction in Ghana and, thereby, to ing and support services. Urban centers in Ghana identify the main drivers. have thereby improved human capital, roughly The analysis proceeds in two stages. First, we doubling and tripling the shares of their popula- analyze the determinants of poverty status using tions with secondary and tertiary education, a limited dependent variable model (probit) in which, in 2000–10, grew from 9.3  percent to which the dependent variable is binary—with a 19.3 percent and from 2.3 percent to 7.6 percent, value of 0 for the non-poor and 1 for the poor— respectively. The two regions in which Ghana’s and the explanatory variables are standard cor- two largest cities are located, Ashanti (Kumasi) relates of household consumption grouped in five and Greater Accra (Accra), experienced the great- broad categories: household characteristics, est improvement in secondary educational attain- location, education, and type of job of the house- ­ ment in the country over the period. Thus, hold head, and an index of access to infrastruc- urbanization may have advantages beyond the ture2 (see appendix B for a detailed list and the increased welfare among migrants by boosting detailed results). We then analyze the determi- the prospects for inclusive growth and the reduc- nants of changes in consumption at the 20th, tion of inequalities of opportunity. 40th, and 60th  percentiles of the consumption Although the recent urbanization has likely distribution using unconditional quintile regres- contributed to the reduction in poverty, it has sions and the same explanatory variables (Fortin, also raised challenges. Often, rapid population Lemieux, and Firpo 2011).3 In 1991 households shifts from rural to urban centers are associated in the 60th  percentile were around the poverty with the uncontrolled expansion of the urban line, but by 2012 their consumption level was centers (Agyei-Mensah and Owusu 2010), includ- well above the poverty threshold. Households in ing in slums, if the provision of housing and basic the 40th  percentile were poor in 1991 but had services is inadequate. If the supply of services progressively moved out of poverty by 1998, cannot meet the growing urban demand and if although they remained vulnerable to falling urban economies do not generate sufficient job back into it. Finally, the 20th percentile includes opportunities, slums rise up, leading to declining those households who remained poor over the health outcomes, growing poverty, and greater whole period. insecurity. These and other challenges will emerge The educational attainment of the house- if urbanization continues apace without changes hold head and the jobs she does are a main in the country’s current policies and institutional determinant of poverty status.4 For example, structures. in  2012, as detailed in table  B.1, households headed  by somebody with tertiary (secondary) education were 20 (9)  percent less likely to be The Main Drivers of Poverty poor than identical households with an unedu- Reduction and Consumption cated head. Likewise, having a head that works outside agriculture reduces the probability of Growth the  households being poor significantly and, if Which of the above developments were the main the head works in the public sector, the house- drivers of Ghana’s recent success? Above, we hold is 13 percent less likely to be poor than an describe three major developments of the last two identical one headed by a farmer. Urban house- decades that may have played a key role in reduc- holds are 22 percent less likely to be poor than ing poverty and increasing consumption: struc- their rural counterparts and the probability of tural transformation, the educational achievements being poor is lower for smaller households of the labor force, and effective urbanization. (4  percent) and those with better access to Below, we use econometric analysis to quantify the ­infrastructures (9 percent). 28 Drivers of Poverty Reduction Ghana’s recent success with poverty reduc- The drivers of poverty reduction were also tion is driven by a combination of improve- the engines of consumption growth at different ments in the characteristics of the households points in the distribution. The results of the and higher returns to these characteristics. unconditional quantile regressions (given in Using a standard Oaxaca-Blinder (OB) decompo- detail in tables A.2–A.5) confirm that the drivers sition, we can decompose the change in poverty of poverty reduction were the same as those that that occurred between different periods into the led to higher consumption across all points in the coefficient effect (changes in returns on vari- distribution. ables), the endowment effect (changes in mean Three interesting trends emerge regarding characteristics), and a residual. The results, pre- the role of the education attainment of house- sented in ­figure 2.10, show that the rapid improve- hold heads (figure 2.11). First, during the entire ment in the characteristics of households are as period, and for all  percentiles analyzed, the important in explaining the poverty reduction returns to education were positive and increasing that occurred between 1991 and 2012 as the with the level of education. Second, for the 40th increase in the returns to those characteristics. and 60th  ­ percentiles, returns on primary and However, there were significant differences across secondary education flattened over time (or subperiods. Increases in returns were particularly ­ declined as for primary in the 60th  percentile), important between 1991 and 2005, as the growing but they grew very fast for the 20th  percentile economy was putting a high premium on skills between 2005 and 2012. Finally, the gap between and the returns to having access to infrastructure returns to higher education and those to the and to being in nonagriculture employment were other levels grew substantially for the 40th and high. By contrast, between 2005 and 2012 poverty 60th  percentiles. These findings suggest that reduction appears to have been driven mainly by investment in primary and secondary education investment in improved endowments while the remains key to increasing consumption of the returns to those characteristics declined.5 This poor (bottom 20th in 2012), who still have com- was probably due to the limited capacity of the paratively low educational attainments. By con- modern urban sectors to absorb the increasing trast, for the better-off only investments in higher educated labor force and to take full advantage of education continue to pay as the poor quality of the improved endowments. education, and the inability of the economy to Figure 2.10  Decomposition of Poverty Changes, 1991–2012 1991–2012 1991–2005 2005–12 –0.20 –0.15 –0.10 –0.05 0 0.05 0.10 Variation in poverty rate Endowments Coefficients Interaction Source: Calculations based on GLSS 3–6. Drivers of Poverty Reduction 29 Figure 2.11  Education Coefficients, by Percentile and Year 20th 40th 60th 0.6 0.6 0.6 0.4 0.4 0.4 Coefficient 0.2 0.2 0.2 0 0 0 1991 1998 2005 2012 1991 1998 2005 2012 1991 1998 2005 2012 Survey year Primary education Secondary education Higher education Source: Calculations based on GLSS 3–6. absorb the fast-growing supply of more educated, diversification in the household portfolio of have muted the return to lower educational activities may affect the household head only in attainment. its final stage. Employment outside agriculture had a posi- The role of urbanization as a driver of pov- tive impact on consumption at all levels of the erty reduction is confirmed by the high returns distribution. In 1991, when household heads to living in urban areas. These returns are sig- were predominantly working in agriculture, the nificant across all  percentiles (see tables B.2 returns to a nonagriculture job were high across through B.5). For the 40th and 60th percentiles, all percentiles (figure 2.12). The returns started to the coefficients follow a similar pattern: they decline between 2005 and 2012, and became declined between 1991 and 1998, surged rapidly mostly insignificant, possibly as a reflection of the between 1998 and 2005, and eventually declined inability of the economy of absorb the workers again in the last decade. Again, these trends may moving out of agriculture in higher productivity be closely linked to the rise in the number of activities. The reduction in returns was particu- households migrating from rural areas between larly noticeable for the employment in the public 2005 and 2012, which progressively raised the sector, which declined sharply after 2005 for both opportunity costs of moving to urban areas while the bottom and the top 20  percent. Non- reducing the net returns. This confirms that agricultural self-employment was the only activ- structural transformation, changes in the labor ity for which returns remained unchanged over market, and urbanization are inextricably linked. the entire period, especially for the bottom 20. The results of the infrastructure index and However, the focus of our analysis on household the regional dummies confirm that living in heads only may underestimate the full extent of dynamic areas and having access to basic infra- the returns to sectoral transformation, as the structures significantly improve consumption 30 Drivers of Poverty Reduction Figure 2.12  Employment Category Coefficients, by Percentile and Year 20th 40th 60th 0.6 0.6 0.6 0.4 0.4 0.4 Coefficient 0.2 0.2 0.2 0 0 0 1991 1998 2005 2012 1991 1998 2005 2012 1991 1998 2005 2012 Survey year Private employee Public employee Nonagricultural self-employed Source: Calculations based on GLSS 3–6. Figure 2.13  Decomposition of Household Consumption 1998 over 1991 2005 over 1991 2012 over 1991 0.8 0.8 0.8 0.6 0.6 0.6 Total difference 0.4 0.4 0.4 0.2 0.2 0.2 0 0 0 20th 40th 60th 20th 40th 60th 20th 40th 60th Percentile Endowments Coefficients Interaction Source: Calculations based on GLSS 3–6. Drivers of Poverty Reduction 31 for all percentiles. Using as the baseline one of Figure 2.14  Decomposition of Changes in the poorest regions in the country, the Upper Characteristics, 1991–2012 West Region, we find that all other regions tended 0.30 to perform consistently better during the entire period (see tables B.2 through B.5). Overall, the 0.25 coastal regions and Ashanti fared better than the regions of the north. The coastal regions are also 0.20 Tot l diff r nc the most highly urbanized, and the urban area 0.15 coefficient became positive and significant in 2005 across all percentiles and remained positive 0.10 and (often) significant in the following survey round. 0.05 The combination of improved characteristics 0 and higher returns to these characteristics were 20th 40th 60th also the main drivers of increases in consump- Percentile tion for all Ghanaians. The results of the OB Household composition Location decomposition for the entire consumption distri- Infrastructure index Employment category Educational attainment bution, reported in figure 2.13, confirm the find- ings of the analysis based on growth incidence Source: Calculations based on GLSS 3–6. curves (reported in chapter 2) that consumption grew more quickly for the highest  percentiles.6 those that drove the changes. As illustrated by They also suggested that the difference in growth figure 2.14, the key factor in changes in the con- across percentiles was due mostly to the higher sumption of the bottom 20th is the movement rates at which the better-off accumulated endow- toward richer areas. For the 40th percentile loca- ments during the period. While the returns to tion, it remains important, but access to infra- characteristics changed in very similar ways for all structure becomes more important and change Ghanaians, the households’ characteristics in the in household composition also matters. The driv- 60th percentile increased faster. ers of change for the 60th percentile are similar to Among the characteristics, ­ infrastructures, those of the 40th but changes in educational location, and household composition7 were attainment play a bigger role. Notes 1. Given variations in the structure and implementa- as cultivation, raising livestock, and so on. (See tion of the GLSS, efforts have been made to con- Bardasi et al. 2010). struct comparable labor market indicators across 2. The infrastructure index is obtained combining surveys using the approach developed by the four variables of principal component analysis: International Labour Organization. Essentially, access to protected water, access to electricity, the employed are defined as those people who access to protected sanitation, and access to safe worked for pay or profit during the seven days sources of cooking. prior to the survey(s), while the unemployed are 3. The unconditional quantile regression methodol- people who did not work, though they wanted ogy is based on regressions in which the depen- to and were actively trying to find a job or start dent variable is a transformation (the “recentered a business. The remaining population is defined influence function”) of the outcome variable. as inactive. According to the approach of the While conditional quantile regression allows to International Labour Organization, unpaid house- recover the impact of a small location shift in the hold duties are not counted as employment. As distribution of a variable of interest at quantiles shown, for instance, in Tanzania, this may lead to (different points) of the conditional distribution underreporting of economic activities, especially of the dependent variable—that is, given the dis- with respect to women who do unpaid work such tribution of the variable of interest—unconditional 32 Drivers of Poverty Reduction quantile regressions allow estimating the same dummies, although with the right sign, become impact for the entire (unconditional) distribution insignificant. of the dependent variable. 6. Results are reported in table  B6. The interaction 4. Negative coefficients indicate less poverty. term is not reported. 5. When comparing the marginal effects (table B.1) 7. The variables estimated in the quantile regres- of education, labor, infrastructure, and locality sion have been grouped in five groups: household between 2005 and 2012, those of 2012 are sys- characteristics, location, infrastructure  index, tematically smaller with the notable exception of employment category, and educational attain- secondary education. Private and public sector ment. See tables B2–B5 for details. Drivers of Poverty Reduction 33 Chapter 3 Remaining Challenges Despite Ghana’s success in reducing poverty and inequality, we decompose the Gini index into promoting shared prosperity, challenges remain. three components for each household charac- This chapter describes three of these challenges: teristic: one reflecting cross-household varia- the growing inequality and polarization in tions in the characteristics (the between household consumption, the large spatial dispar- component, ­ represented in figure 3.1 by the ities, and the deteriorating macroeconomic envi- orange blocks), one measuring differences in ronment. Consolidating the progress made in consumption across groups of households with poverty reduction and shared prosperity over the same characteristics (the within component, recent decades will require that these challenges represented in figure 3.1 by the yellow blocks), be addressed promptly and effectively. and an interaction or overlap term (represented in figure 3.1 by the red blocks.1 In panel a in ­figure 3.1, for example, the orange component of Growing Inequality and the bars (the between component) represents Polarization in Household the share of inequality that is accounted for by differences in educational attainment among Consumption household heads. The yellow component of the Inequality in household consumption wid- bars (the within component) shows the share of ened substantially between 1998 and 2005. inequality arising because of differences in con- The picture that emerges is consistent across dif- sumption across households the heads of which ferent indicators of inequality (table 3.1). have the same educational attainment. The red Inequality in household consumption was ini- component of the bars shows the residual of the tially constant, but widened considerably decomposition. between 1998 and 2005—a jump of about 9 Differences in household characteristics percent in the Gini coefficient and 20 percent in generally account for only a small part of the Theil index. The consumption share of the inequality. The relative large orange (between) poorest quintile of the population (the bottom block in the relevant panels in figure 3.1 shows 20) also declined steadily between 1991 and that household differences in the region of resi- 2005 (from 6.8 percent to 5.7 percent) while the dency, the education attainment of house- share of the top 20 increased slightly (from 44.8 hold  heads, and the type of job she does percent to 46.6 percent). Inequality has remained (­nonagriculture) are important determinants of constant at the higher level after 2005 but the consumption inequality. However, the fact that trends in the share of consumption of the bot- the yellow (within) component is also large sug- tom and top quintile have continued in the same gests that a large degree of inequality persisted direction. across groups of households. Inequalities among What is behind the changes in inequality? regions played a particularly important role, Our econometric analysis in chapter 2 points to accounting for an average of 40 percent of the four main drivers of household consumption: total Gini over the period, while within regions educational attainment, structural transforma- inequalities were responsible for only 10 percent tion, spatial transformation, and demographic of total inequality. By contrast, the heterogeneity changes. To quantify the relative importance of in consumption was much greater within the these factors in determining the overall level of rural and urban  areas and among households Remaining Challenges 35 Table 3.1  Measures of Per Capita Household Consumption Expenditure, 1991–2012 Measure 1991 1998 2005 2012 Gini 0.38 0.38 0.41 0.41 Theil 0.25 0.25 0.30 0.29 Mean 459.91 568.45 736.80 883.48 Median 352.66 438.04 559.44 655.60 Consumption shares Bottom 5 percent 1.11 1.00 0.79 0.82 Bottom 10 percent 2.71 2.42 2.08 2.13 Bottom 20 percent 6.82 6.21 5.65 5.63 Top 20 percent 44.78 44.47 46.59 46.94 Top 10 percent 29.16 28.17 30.75 30.43 Top 5 percent 18.52 17.41 19.95 19.17 Source: Calculations based on GLSS 3–6. Figure 3.1  Decomposition of the Gini Index, by Household Characteristic, 1991–2012 a. Agriculture/nonagriculture b. Education level 100 100 80 80 Percent Percent 60 60 40 40 20 20 0 0 1991 1998 2005 2012 1991 1998 2005 2012 c. Region d. Rural/urban 100 100 80 80 Percent Percent 60 60 40 40 20 20 0 0 1991 1998 2005 2012 1991 1998 2005 2012 Overlap Between Within Source: Calculations based on GLSS 3–6. headed by nonagricultural workers. A compari- Finally, despite the increase in the 1990s, son of the components across periods also shows inequality remains narrow in Ghana compared that the link between inequality, on the one hand, with other African countries. Figure 3.2 ranks and education and nonagricultural employment, several African countries according to the aver- on the other, increased (the relative size of the age Gini index over the last 20 years. In 1991, orange block has grown), while the link of Ghana was in the bottom 20 percent of the Gini inequality with location declined slightly. distribution in Africa, and, despite some 36 Remaining Challenges deterioration, in 2012, it was still below the continued to diverge at a steady rate so that the median and among the lowest among the rapidly gap expanded by 30 percent over the full period. growing African economies. The divergence was widening because the bottom However, the narrative about inequality is 10 was being left behind, rather than because the more nuanced than the summary measures sug- top 10 was gaining disproportionately compared gest. The summary measures of inequality ana- with the rest of the population. The average con- lyzed above only partially capture the changes at sumption of the 90th percentile rose little relative various points of the consumption distribution. to the median, the 50th percentile, while the aver- Analysis of the incidence curve provides more age consumption of the bottom 10 had deterio- detailed information on the changes occurring at rated by nearly 20 percent by 2005. The bottom 10 all points of the distribution (see chapter 2). The appears to be losing ground also compared with results of a simple interquantile analysis can com- other households in the bottom 25, who are also plement the analysis of the incidence curve losing ground to the median but only half as (table 3.2). They show that the ratio of average con- quickly. sumption among the top 10 percent of the distribu- These findings are in line with the clear evi- tion (the top 10) to the average consumption among dence indicating a rise in polarization. A detailed the bottom 10 percent (the bottom 10) had risen analysis of polarization allows us to zoom in on considerably even before 1998, suggesting that the the dynamic changes occurring at different points better off had benefited more than the poorest of the consumption distribution.2 The results of decile from the economic growth in 1991–98. this analysis suggest that the distributional changes Over the years, the consumption levels of observed in 1991–2012 hollowed out the middle the  top and the bottom of the distribution of the Ghanaian household consumption distribu- tion and increased the concentration of house- holds around the highest and lowest deciles. Like Figure 3.2  Gini Indexes in Sub-Saharan inequality, polarization began to increase in the Africa late 1990s, but continued to grow even after 70 inequality stabilized, although at a slower pace. We measure the degree of polarization of 60 Ghanaian society using the relative distribution method.3 This method involves the creation of a distribution that captures the share of households GHA 12 in the comparison year (2012) that falls within GHA 91 each income decile of the reference year (1991). This distribution is then decomposed into a loca- 20 tion effect, which tells us if there is a change in the 0 20 40 60 80 100 median (or mean) of the income distribution, and Percentiles of Gini in SSA a shape effect, which, representing the relative Source: Calculations based on PovcalNet. distribution, net of the location change, is useful Table 3.2  Interquartile Consumption Ratios, by GLSS Wave, 1991–2012 Year p90/p10 p90/p50 p10/p50 p75/p25 p75/p50 p25/p50 1991 5.23 2.42 0.46 2.37 1.56 0.66 1998 6.00 2.48 0.41 2.60 1.64 0.63 2005 6.36 2.46 0.39 2.63 1.62 0.61 2012 6.73 2.65 0.39 2.68 1.66 0.62 Source: Calculations based on GLSS 3–6. Remaining Challenges 37 in isolating movements (redistribution) between the relative density into the changes arising the reference population and the comparison because of the increase in median income (the population. location effects) and the changes arising because of The generalized increase in consumption changes in the distribution (the shape effect). The meant that, in 2012, households were crowded latter are those changes produced by the changes around the top deciles of the 1991 distribu- in distribution, net of the impact of the increase in tion. Figure 3.3 shows the share of households median consumption (figure 3.4). The polariza- in 2012 that fall into each decile of the 1991 dis- tion of households around the bottom 10 and the tribution. It is obvious from the figure that the top 10 is striking. Without the rise in median share of households with consumption levels income, the greater dispersion of consumption equivalent to the consumption levels of the expenditures would have led to relatively more upper deciles in 1991 rose dramatically through- low-consumption households in 2012, and this out the two decades, while the shares with con- effect was mainly concentrated among the bottom sumption levels equivalent to the consumption 10. By contrast, at the top of the distribution, the levels of the bottom and around the middle in effect of higher median income reinforced the 1991 declined. polarization in the distribution, which, of itself, However, the crowding of the population in would have boosted the share of households in the the top 1991 deciles masks growing polariza- top 10 by nearly 160 percent. In sum, once we net tion. To gauge this, we decompose the changes in out changes in real median expenditure, we Figure 3.3  Relative Consumption Distribution, 1991–2012 4 Relative density 3 2 1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Reference proportion Source: Clementi, Molini, and Schettino 2015a. Figure 3.4  Median-Adjusted Relative Consumption Distribution, 1991–2012 4 3 Relative density 2 1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Reference proportion Source: Clementi, Molini, and Schettino 2015a. 38 Remaining Challenges Figure 3.5  Median-Adjusted Relative Consumption Distribution Series, Ghana, 1991–2012 Re la tiv ed en sit y Referenc e e proportio av n W Source: Clementi, Molini, and Schettino 2015a. observe a U-shaped relative density, indicating Through the period 1991–2012, 80 percent of the that polarization was hollowing out the middle of poor lived in rural areas. The increasing concen- the household consumption distribution. tration of the poor in the north is also quite clear; The level of polarization has been growing. the highest poverty rate in the country is in the For each wave of the GLSS between 1991 and Upper West Region (in the north). Moreover, 2012, figure 3.5 shows the shape effect of the while in 1991, 25 percent of the poor were living household consumption relative density; 1991, as in the north, the share had increased  to  about the reference sample, is constructed to be flat.4 40  percent by 2012, despite the stabilization in Following the plot through each successive wave, the population share of the north at around we find that the shares of households at both the 17–18 percent. However, great heterogeneity top and the bottom tails of the consumption dis- exists even within regions. tribution rose consistently over the period, while The better performance of urban areas is the shares in the middle declined. Compared driven mostly by the striking progress of the with 1991, more households had dispersed from Ashanti and Greater Accra regions. Greater the central deciles toward the lowest and the Accra has enjoyed the lowest poverty rate in the highest tails by 2005. In 2012, the hollowing out country since at least 1991, and the rate has con- of the middle continued, but at a slower rate: tinued to decline there. Although the number of both the 1st and the 10th percentiles were above the poor rose considerably in 2005 as a result of their respective values in 2005 and, of course, far mass migration from surrounding poorer areas, above the 1991 reference levels. the poverty rate fell by half in Greater Accra between 2005 and 2012.5 Among the coastal and forest regions, Greater Accra is the only one that had reduced both the poverty headcount and Persistent Spatial Disparities the number of the poor by 2012. Ashanti, the Our descriptive profile and econometric analy- most heavily populated and the second most sis suggest that geography has been an impor- urbanized region in the country, has experi- tant correlate of the patterns of poverty and enced a more stable pattern of poverty reduc- inequality in Ghana (see above and chapters 2 tion over the last 20 years. There, the poverty and 3). Two spatial trends are evident. As in 1991, rate continued to decline, and the decline accel- poverty is mainly a rural phenomenon, and the erated between 1998 and 2005, although the lion’s share of the poor are located in the north. number of the poor began to grow between Remaining Challenges 39 2005 and 2012. Ashanti was the only region in increasing to 5 in 10 in the Northern Region which both the urban and rural populations (50.4 percent), and 7 in 10 in the Upper West rapidly expanded. Unlike Greater Accra, half Region (70.7  percent). the population of Ashanti lives in areas classi- In 2005–12, the pace of poverty reduction fied as rural. slowed in those regions that had previously Between 1991 and 2012, the Brong Ahafo witnessed considerable poverty reduction. In and Western regions enjoyed the most rapid 1998–2005, poverty reduction was driven by the poverty reduction in the country. The poverty progress made in the southern and central rate dropped from about 60 percent to less than regions. By contrast, between 2005 and 2012, 25 percent. A key driver of this outcome was the Greater Accra and the three regions in the north boom in cocoa production, which is highly con- drove most of the change. In the southern and centrated in these regions. However, the urban- central regions, poverty rates declined slightly, rural divide remains large: poverty rates in the but the influx of people from poorer areas offset urban areas of the two regions were around this reduction, and the net effect was a slight 10  percent, compared with 20 percent (the increase in the number of the poor. Western Region) and 30 percent (the Brong To understand fully the extent of the spatial Ahafo Region) in rural areas. Moreover, after a inequalities in Ghana, one should deepen the long positive trend, the fall in the poverty head- within-region analysis because heterogeneities count halted in 2012, and the absolute number of exist also within regions. Poverty maps allow a the poor started to rise in the Western Region, better focus on the spatial distribution of poverty despite the expected positive spillovers from the at the local level (Elbers, Lanjouw, and Lanjouw discovery of oil there. 2003). Although the number of districts (110 and All regions bordering Greater Accra, but 216) differ across 2000 and 2012, the two years especially the Central and Eastern regions, shown in map 3.1, the visual interpretation is still benefited from the growth of the capital city, meaningful. Map 3.2 adds two more indicators where a large share of the population in these relevant to welfare. regions work. In the Central and Eastern regions, In 2000, poverty rates above 20 percent were the poverty rate and the absolute number of the the norm in Ghana. Only a few districts in the poor fell by more than half between 1998 and Ashanti and Greater Accra regions showed pov- 2005. In the subsequent period, poverty reduc- erty rates below 20 percent. Districts in the tion stagnated, and, in the Eastern Region, the Ashanti, Greater Accra, and Western regions and absolute number of the poor began to rise, albeit in the Eastern Region neighboring Accra, typi- marginally. cally had poverty rates at between 21 and Both the poverty rate and the number of 40 percent. Rates in the regions in the center of the poor were stubbornly high in the north the country mostly exceeded 40 percent. There until 2005, but have declined somewhat since. was a rather homogenous central belt stretching Between 1998 and 2005 poverty rates fell in the across almost all districts in the eastern Ashanti, north too. Nevertheless, because of the rapid Brong Ahafo, Eastern, and Volta regions in which growth  in population, the absolute number of poverty rates never fell below 40 percent. Part of the poor rose in all three regions in the north the Northern Region bordering the Brong Ahafo (see figure 1.8). In 2012, both the poverty rate and Volta regions showed similar characteristics, and the absolute number of the poor declined while practically all districts in the Upper East across these three regions. In the Upper East and the Upper West regions recorded poverty Region, the share of the poor was cut by rates well above 40 percent. At the extreme, a almost half. Nonetheless, the poverty rate was group of Upper East and Upper West districts still high. More than 4 people in 10 were poor bordering, respectively, Togo and Burkina Faso in  the  Upper East Region (44.4 percent), showed poverty rates above 80 percent. 40 Remaining Challenges Map 3.1  Poverty Maps, 2000 and 2012 Source: Calculations based on GLSS 4 and 6. Map 3.2  Additional Indicators of Welfare, 2011 and 2012 Source: Calculations based on GLSS 6. Remaining Challenges 41 BOX 3.1  Poverty Maps in Ghana The Ghana Statistical Service (GSS) has produced two comparable poverty maps. One, issued in May 2005, is based on data of the 2000 Population and Housing Census and the 1998–99 GLSS (round 4). It shows 110 districts. The more recent one has been computed based on data of the 2010 Population and Housing Census and the 2012–13 GLSS (round 6). It illustrates 215 districts. Both poverty maps have been created based on the small area estimation methodology that has been developed to allow accurate estimates of consumption-based poverty and inequality at lower levels of ­ disaggregation by combining information from censuses and household consumption surveys. ­ The country was radically different 13 years Adaklu District (89.7 percent) in the Volta later. In the corner formed by the Ashanti, south- Region is more than two and half times the ern Brong Ahafo, Eastern, Greater Accra, regional average (33.8  percent). Greater Accra Western, and coastal Volta regions, poverty rates has the lowest regional poverty rate in the coun- had typically declined to below 20 percent in try, and poverty there is concentrated in two dis- 2012. There were clusters of districts with higher tricts, Ningo Prampram District (31.2 percent) poverty rates, particularly in the inland parts of and Shai Osudoku District (55.1 percent). In the the Central Region, but, overall, the improve- Northern Region, the poverty rate is around 50 ment is striking. The central belt is now much percent, but two districts have poverty rates more heterogeneous, albeit most districts have below 30 percent: Tamale Metropolitan District poverty rates at less than 40 percent, mainly in (24.6  percent) and Sagnarigu District (29.3 the Brong Ahafo and Volta regions. In the north, percent). the difference in performance between the east- The incidence of poverty appears highly ern and western portions of the area is apparent. correlated with proximity to roads in hours In the east, the districts belonging to the Northern and the yield of the maize crop. A comparison Region reduced poverty to less than 40 percent, of the 2012 poverty map (map 3.1, panel b) and while most of the districts belonging to the Upper the map of the proximity to roads in hours East region, which were among the poorest in (map  3.2, panel a)—a good proxy for market 2000, reduced the poverty rate to below connectivity and ease of access to basic facilities 60 percent. By contrast, in the west, particularly such as hospitals and schools—shows that prox- in the Upper West Region, poverty rates remained imity to roads is strongly associated with low mostly above 60 percent, although with large poverty rates. The south and west of the coun- within-region disparities, from a low of about try,where poverty is below 20 percent in most 36 percent in Wa Municipal District to approxi- districts, also have the best road networks; there, mately 84 percent in Wa East District and over 90 the average distance to the nearest road is less percent in Wa West District. It is noteworthy that than an hour. By contrast, in the north, the aver- the two poorest districts in the Upper West age distance to the nearest road is greater, as are Region border the least poor district, a fact the poverty rates. that  would not have been apparent without a The map of the yield of the maize crop, the poverty map. most common staple in the country, has some The poverty maps also reveal islands of similarities with the poverty map, too (see poverty and prosperity. The poverty rate in ­ map 3.2, panel b). In areas of the Eastern Region 42 Remaining Challenges and the inland parts of the Central Region where Expansionary fiscal policies increased infla- poverty was reduced rapidly, maize yields are the tionary pressures. Overall inflation rose sharply, highest in the country. By contrast, in the high- from 8.8 percent in 2012 to 17.0 percent in 2014, poverty north, maize yields are low. These are and nonfood inflation shot up from 11.6 percent areas that are only modestly urbanized and in to 23.9 percent. The producer price index climbed which cash crops are not so diffused. There, low by around 17.1 percent in 2012 before jumping levels of production seem to be associated to 35.8 percent in 2014. The inflationary impact with low productivity in the cultivation of maize of central bank financing was aggravated by ris- and, consequently, low levels of consumption. ing public sector wages, the pass-through effects Meanwhile, in southern Ghana, high maize yields of rising fuel and utility prices, and the deprecia- are not always associated with low levels of tion of the cedi, which effectively boosted import ­ poverty. Often, the contrary is true. Along the prices. coast, poverty shrank, but this is not an area suit- In an effort to stabilize the economy, the able for maize cultivation. Likewise, highly urban- government prepared and adopted a multiyear ized areas such as the Ashanti Region registered plan aimed at reducing the fiscal deficit. It took rapid poverty reduction, but not because of higher important steps to address fiscal imbalances, crop productivity. In the Western Region, there including the elimination of subsidies on fuel was rapid poverty reduction and higher produc- products and utilities. In 2013, as crude oil prices tivity in crops other than maize. rose and the cedi depreciated, it began to pass a larger share of the higher cost of energy produc- tion on to consumers, which severed the link The Deteriorating between commodity price volatility and fiscal accounts. In February 2013, administratively set Macroeconomic Environment gasoline and diesel prices increased by 20 percent, Ghana has suffered a number of serious exter- and, by December 2014, gasoline and diesel nal and internal macroeconomic shocks since prices had risen by 100 percent, while liquefied 2012. Major external shocks included the rup- petroleum gas prices rose by 128 percent over the ture of the West African natural gas pipeline in same period. Median electricity prices skyrock- 2012 and the highly volatile export price of gold. eted by 160 percent between October 2013 and The pipeline supplies Benin, Ghana, and Togo December 2014, and water prices climbed by with natural gas from Nigeria. Ghana was thus 80 percent. forced to increase oil imports for the generation This adjustment, combined with the rapid of electricity, causing the import bill to rise increase in inflation and the depreciation of ­ dramatically, to approximately US$27 million a the cedi, had a strong impact on household month.6 consumption in urban areas other than Accra The increase in the cost of oil imports was and in the rapidly growing south. Households partially offset by the rising export price for faced a sharp fuel price rise and, as a conse- gold. However, by 2014, global gold prices had quence of high inflation and capital outflows, a tumbled and could no longer offset the higher strong cedi devaluation (Clementi, Molini, and cost of oil imports. Furthermore, the rising oil Schettino 2015b). The fuel price rise affected the imports pushed the current account deficit to an cost of many other items, including the prices average 11.5 percent of GDP in 2012–13, and it of  nonfood items, which increased the most remained high, at 9.2 percent, in 2014. Currency figure 3.6). These developments hit urban (­ exchange rates depreciated the Ghanaian cedi by households particularly hard because they 35 percent against the U.S. dollar on the inter- spend  a larger share of their budgets on fuel bank market and by 43 percent on the foreign and  imported items and are typically net food exchange market. consumers. Probably because of the increase in Remaining Challenges 43 Figure 3.6  Regional Price Indexes: Total and Nonfood, 2013 . Tot l b. Nonfood 115 115 109.89 Accr S pt. Pric ind x Pric ind x Accr S pt. Accr J n.=100 Accr J n.=100 95 95 90 90 80 80 Ce rn Vo l Ea ta As ern i o pe ern rW t t Ce rn Vo l Ea ta Br sha n i o pe ern rW t t on ant nt ra ra pe as es pe as es No haf No af A er l l te te nt nt Up r E Up r E Ah st Up rth st Up rth h es es A W W g g on Br January 2013 September 2013 Source: Calculations based on the consumer price index. transport costs, the impact was especially severe self-production and the nonfood component in urban areas other than Accra. During the first represents around 60 percent of consumption— nine months of 2013, prices grew relatively more were more resilient to these short-term macro- in the Ashanti and Western regions than in economic cycles (Fox 2015). The spike in prices Accra, and the price differential was appreciably in the regions that have been the backbone of larger for nonfood items. By contrast, the living Ghana’s economic success raises concerns about standards in rural areas and in the north—where the country’s future growth prospects. about 20 percent of food consumption relies on Notes 1. For more details on the decomposition, see of compositionally adjusted distributions. Relative Bhattacharya and Mahalanobis (1967) and Rao distribution methods can be applied whenever the (1969). distribution of some quantity across two popula- 2. Whereas inequality is a measure of the overall dis- tions is compared either cross-sectionally or over persion of the distribution and refers to the dis- time. For our purposes, the relative distribution tance of every individual from the median or the is defined as the ratio of the density in the com- mean income, polarization captures the combina- parison year to the density in the reference year tion of divergence from the global mean income evaluated at each decile of the consumption distri- and convergence toward local mean incomes. For bution; it can be interpreted as the share of house- a detailed explanation of the concept of polar- holds in the comparison year’s population that fall ization and the techniques used to measure the into each decile of the reference year’s distribution. extent of polarization, see Clementi et al. (2015) This allows us to identify and locate changes that and Clementi, Molini, and Schettino (2015a). have occurred along the entire Ghanaian house- 3. Techniques based on the relative distribution hold consumption distribution. method developed by Handcock and Morris 4. This analysis uses relative distribution methods (1998, 1999) enable the counterfactual comparison that allow us also to analyze how redistribution 44 Remaining Challenges occurred across households over the entire period. the Eastern Region and the Greater Accra Region For more details, see Clementi, Molini, and who, in 2006, were wrongly counted in the Eastern Schettino (2015a). The relative distribution and, Region and who, in 2012, were reassigned to therefore, the corresponding shape effect are, by Greater Accra. Additional research is needed to definition, flat in the reference year (see Morris, establish the exact number of people affected. Bernhardt, and Handcock 1994). 6. Oil imports rose in value from US$2.9 billion in 5. We must exclude from this group a large num- 2012 to US$3.7 billion in 2014; see World Bank ber of residents living close to the border between (2015a). Remaining Challenges 45 Chapter 4 A Roadmap for Policy Action The economic development of Ghana over the it was only 10 percent lower. The extreme ­poverty last two decades has been a tale of success. The rate declined even more quickly and, by 2012, country has cut the poverty rate by half, consid- was only around a quarter of the 1991 rate. erably reduced vulnerability, and substantially However, the speed of poverty reduction slowed improved nonmonetary indicators of living stan- after 1998, from 13.0 percent to 7.1 percent in dards, such as measures of health care, educa- 2005–12 despite the rising GDP growth rate. tion, and access to basic services. The driver The increase in household consumption behind this success has been sustained economic was accompanied by dramatic improvements growth, which has led to a generalized increase in nonmonetary indicators. Today, over the life- in consumption. However, the position of the span, newborns are expected to live two years bottom 40 relative to the rest of the population longer than newborns in 2005; children are more changed little and inequalities in both outcomes than twice as likely to be enrolled in secondary and opportunities remain substantial. school, and households are more than twice as The development objective of Ghana is now likely to have electricity and improved sanitation to consolidate the success in the face of internal facilities. In Ghana in 2012, the level of child and external challenges and a rapidly changing malnutrition was among the lowest in Africa, economic and social environment. The deterio- and the literacy rate and rate of access to basic rating macroeconomic outlook, the nature and services were among the highest. The striking speed of structural and spatial transformation, progress in providing better opportunities for all and the persistent inequalities threaten future is more than an achievement in its own right. It progress in poverty reduction and the prospects also strengthens the prospects for strong and for growth. inclusive growth in the future. After highlighting the main achievements The reduction of poverty and the expansion in  recent years and the challenges ahead, this in shared prosperity were achieved during a chapter outlines a roadmap for policy action. period of rapid changes in the economic and The task ahead is complex, and the sequencing sociodemographic structure of the economy. and prioritization of policies need to be carefully Four factors are especially relevant in this out- considered. The key to success will be the identi- come: (1) the decline in the average household fication and implementation of a fiscally sustain- size and the drop in the dependency ratio, able policy package that balances the acute needs (2) structural transformation away from agricul- of the poorest regions and the requirement to tap ture, (3)  the increased skills among the labor into the most dynamic segments of the economy. force, and (4) successful urbanization. The changes in household composition led to the lower dependency ratio (Falco et al. 2014). The changes in demographics played an impor- A Tale of Success tant role in securing the position of vulnerable The poverty rate has been reduced by half over households. We observe important changes in the last two decades. It fell from 52.7 percent in household structure starting with the 60th per- 1991 to 21.4 percent in 2012 (chapter 2). It is now centile (chapter 3). The other three factors helped less than half the average in Africa, while in 1991, raise the earnings potential of the typical adult in A Roadmap for Policy Action 47 the household. In a country in which employ- Ghana is urbanizing quickly. Between 1991 ment rates are high, the increase in earnings is and 2012, over 8 million Ghanaians migrated strongly associated with growing consumption to  urban areas, mostly into the metropolitan and a movement out of poverty—for example, it areas  of Accra and Ashanti, which gained over explains nearly half of the poverty reduction 2.4 ­million inhabitants each, around half in the between 1998 and 2005 (Azevedo et al. 2013). last decade. As a result of the shift, by 2012, the In 2011 industry had a larger share of value population was equally split between urban and added than agriculture. The recent economic rural areas, while in 1991, 70 percent had lived in growth has been associated with a shift of the rural areas. economy out of agriculture and a substantial rise in agricultural productivity. The share of agricul- ture in GDP declined by nearly 50 percent, and, by 2011, agriculture was the smallest sector in A Less Positive Outlook the economy in terms of value added. Meanwhile, Despite the success in reducing poverty the service sector expanded to nearly half of and  promoting shared prosperity, challenges GDP, from an initial 34.4 percent in 1991. remain. We analyze three development chal- Moreover, in 2011, for the first time since inde- lenges in detail in this report: growing inequality pendence, the share of industry in GDP sur- and polarization in household consumption, passed that of agriculture. large spatial disparities, and the deteriorating Workers are now less concentrated in agri- macroeconomic environment. Consolidating the culture. The sectoral distribution of employment progress made over recent decades requires that adjusted in line with the changes in the struc- these challenges be addressed promptly and ture  of the economy, albeit at a lower rate. effectively. Agriculture remains the main sector of employ- There are still large inequalities in opportu- ment, at 43.2  percent of total employment. nities and outcomes. The recent success with However, over the last decade, workers have been poverty reduction and shared prosperity has shifting into services and, to a lesser extent, been driven by the generalized growth in living industry. Employment in the service sector ­ standards and has been achieved despite a small expanded from 28.8 percent in 1991 to 42.0 percent increase in inequality, which was particularly in 2012. The share of construction more than noticeable between 1998 and 2005. Our analysis doubled. The share of industrial employment also points to differences in educational attainment increased, from 10.7 percent to 14.9 percent. among household heads, their jobs, and their The labor force has become better educated region of residency as important determinants of over the past two decades. Between 1991 and consumption inequality, although inequality also 2012, the share of the labor force without school- persisted across households with similar charac- ing shrank almost by half, from 41 percent to teristics (chapter 4). Inequalities between regions 24 percent. By 2012, the majority (52 percent) of played a particularly significant role; they explain, workers had completed at least lower-secondary on average, 40 percent of the total change in the education, compared with 39 percent in 1991. Gini. Of special concern are inequalities in the However, large inequalities remain in access to quality of education across schools and geo- education, especially secondary education and graphical areas—for example, the top 10 percent above. Moreover, the quality of learning in basic of schools produce 90 percent of the students education and beyond appears to be declining, entering university (Darvas et al., forthcoming). and the capacity of the education system to foster The crucial role of the spatial dimension in relevant skills to boost competitiveness and pro- explaining income inequality is evident in ductivity is limited (Darvas, Favara, and Arbnold, the  magnitude of disparities across regions. forthcoming). The incidence of poverty has historically been 48 A Roadmap for Policy Action higher in the north than in the rest of the country A Roadmap for Policy Action because of the less favorable climate, the distance from the sea, and the lack of infrastructure. This The most immediate policy priority is to pre- disadvantage has increased over the last two vent further deterioration in the macroeco- decades. Poverty rates have fallen below 20 nomic environment. Our decomposition points percent in the large area that includes the Ashanti, to economic growth as the main driver of the suc- southern Brong Ahafo, Eastern, Greater Accra, cess in reducing poverty (chapter 2). Thus, coastal Volta, and Western regions, but they are addressing the macroeconomic imbalances that still around 50 percent in the north. In 2012, as a have muted the country’s economic growth pros- result of this gap, nearly 40 percent of the poor pects is a priority in the effort to insure that prog- were living in the north, which accounted for only ress in poverty reduction is sustained. The 17 percent of the country’s population and despite long-term growth prospects are positive, but to the large outflows to the richer south. The num- realize the full potential of growth, it is crucial to ber of the poor was rising in the north, while it continue the stabilization program started in was declining everywhere else. 2014. Its success hinges on maintaining the com- The urbanization process is at a crossroads. mitment to fiscal discipline and rapid structural The population is moving steadily into urban reforms. Assuming that the issue of energy ration- areas. It has grown exponentially in Accra and ing is resolved and that the planned fiscal adjust- Kumasi, but outside these metropolitan areas, it ment remains on track, we expect the GDP has grown more quickly in towns and smaller cit- growth rate to rebound to 5.9 percent in 2016 and ies than in larger cities. However, recent improve- 8.2 percent in 2017. Provided that the growth ments in jobs opportunities and in access to elasticity of poverty remains at the levels of the services have been highly concentrated in the last decade, this can have a strong impact on addi- larger metropolitan areas. Combined with high tional poverty reduction. inflation, this trend is likely the reason poverty Tackling the inequalities in outcomes and rates in urban areas other than Accra are rising. opportunities is a longer-term development Moreover, all urban areas have started to see the challenge. With its rise to middle-income-­country side effects of rapid urbanization, including con- status in 2011, Ghana entered a new stage of gestion, unregulated expansion, and a decline in development, one in which it will be difficult to access to services and affordable housing. achieve progress in poverty reduction and shared The macroeconomic environment is deterio- prosperity without broadening the reach of the rating. Since 2012, Ghana has suffered a num- development process to those people who have so ber  of serious macroeconomic shocks, both far been left behind. The main development chal- external­ ­ —the rupture of the West African natural lenge in this new phase is to enhance the access to gas pipeline and the high volatility in the price of opportunities for all without stifling the energy of gold exports—and domestic. GDP growth fell by the economy. This calls for a multifaceted, half in 2014 and is projected to slow to 3.4 percent well-targeted, and fiscally sustainable policy in 2015. Inflation has risen sharply, from 8.8 package that balances the needs of the poor with percent in 2012 to 17.0 percent in 2014, and non- the needs of the most dynamic sectors. food inflation has increased from 11.6 percent to Reducing the inequality of opportunity 23.9  percent. The Ghanaian cedi depreciated by requires a continued commitment to invest- 35–45 percent against the U.S. dollar, and the mix ment in improving nonmonetary indicators. of the escalating public sector wage bill, the Recent years have seen considerable advances in ­ energy-rationing regime adopted in response to human development indicators and in access to the disruption in natural gas supplies, and the services. However, service quality is still low, and liquidity constraints associated with the purchase the inequalities are significant, particularly across of oil, has exacerbated the imbalance. geographical areas. The key to fostering shared A Roadmap for Policy Action 49 prosperity in the new phase of development will well-targeted transfer program (World Bank be the capacity to reduce these disparities. This is forthcoming). Thus there is scope to continue to because the inequality of opportunity—such as expand programs such as Livelihood unequal access to education, health care, and Empowerment against Poverty. Given the extent credit—can persist across generations by exacer- to which multiple disadvantages overlap in some bating income inequalities, which limit the reali- geographical areas or across some types of indi- zation of the potential of individuals based only viduals, school feeding programs or cash transfer on inherent characteristics (such as gender or programs conditional on a particular behavior ethnicity) or accidents of birth (such as parental among recipients in health care, education, or income and educational attainment). It can also the work environment may be a viable alternative have a severely negative effect on productivity. to current transfer programs. A scaling-up of the Thus, to insure the sustainability of growth and public works program in the north might also be despite the tight fiscal space, it is essential for the considered after a rigorous evaluation of the government to maintain its commitment to impact of the program so far. investment in reducing the gaps in service deliv- However, the main component of an effec- ery and in improving the provision of quality tive program for sustainable poverty reduction services outside the metropolitan areas, but espe- are policies designed to increase earnings cially in the poorest regions. among the poor. Labor is the principal asset and An important element of a successful pol- main source of income among the poor. Changes icy package to reduce inequality is the imple- in labor income accounted for nearly half the mentation of more and better targeted social reduction in poverty in Ghana between 1998 and ­protection. The expansion of social protection 2005 (Azevedo et al. 2013). Boosting earnings can have a positive impact in a country such as requires that the process of structural transfor- Ghana in which (1) poverty is concentrated mation of the country into a modern and diversi- among households with particular characteris- fied economy be accelerated, while increasing the tics in relatively deprived areas, (2)  the various productivity of traditional economic sectors. dimensions of poverty overlap so that the poor Achieving this objective will require three sets of are at a disadvantage in more than one dimen- policy interventions designed to (1) increase pro- sion, and (3) vulnerability to poverty is substan- ductivity in agriculture and in low-­ productivity tial. If well targeted, social protection can help sectors, (2) raise the number of jobs in the reduce the inefficiencies in the allocation of modern private sector (in wage employment and ­ resources and boost the productive potential of in higher productivity self-employment), and individuals and communities by breaking the (3)  facilitate the occupational and geographical vicious circle that links income inequality and mobility of workers. inequality of opportunity across generations. The barriers to improving agricultural pro- Experimenting with innovative ways to expand ductivity are many and complex. Increasing social insurance to cover self-employed and agricultural productivity and transforming agri- informal workers can also lower the vulnerability culture from a subsistence base to a market base of households to shocks. have been government priorities since indepen- The government expenditure on social dence. Various policies and interventions have protection—1.4 percent of GDP—is low com- ­ been implemented to boost cash crop production pared  with the expenditures among Ghana’s for internal and especially international trade, middle-income peers in Africa. The evidence but the results have been mixed (Molini et al. also suggests that current programs are poorly 2010). Cocoa beans are the main export crop, targeted in Ghana and simulations indicate that and Ghana’s increasing share of the global cocoa extreme poverty can be eradicated with a mini- market has been strongly associated with poverty mum investment of 0.5 percent of GDP in a reduction: the poverty rate among cocoa farmers 50 A Roadmap for Policy Action declined from 60 percent in 1991 to about opportunities to the growing nonagricultural 24  percent in 2005. However, despite targeted labor force. Although in 2011, the share of the government interventions, the average yield per industrial sectors in GDP exceeded the corre- hectare—431.0 kilograms in 2005–12—is low sponding share of agriculture for the first time compared with the yields among the country’s since independence, the industrial share of main competitors, Côte d’Ivoire and Indonesia. employment is still below 15 percent. Most newly The growth in the production of staple crops created jobs are concentrated in services. In such as rice, maize, and millet, and in ­higher-value Accra, a key element in the expansion of the ser- vegetables and fruits for domestic and export vice sector was the rapid growth of high–value markets is also encouraging (Breisinger et al. added services such as information and commu- 2008). The average output of staple crops grew nication technology, finance and insurance, and much more quickly than the population, and per real estate. However, elsewhere, the bulk of the capita production was more than 80 percent increase was in low–value added activities that higher in 2005–07 than in 1981–83. However, are borderline between the formal and informal the potential for gains in agricultural productiv- sector and that characterize West African towns: ity is limited by traditional farming methods, retail activities, construction, transport, and so rainfall volatility, and poor access to the expand- on. Progressing to the next stage of structural ing internal market. transformation will hinge on the creation of a Interventions to raise productivity in agri- dynamic, but labor-­ intensive private sector capa- culture should become focused on reducing ble of absorbing the relatively low-skilled work- the reliance of crops on rainfall, increasing the ers released from agriculture. access to larger markets through better infra- Policies to boost the creation of modern structure, and scaling up production. However, jobs are essential. These will include policies the most effective ways of reducing poverty aimed at creating an enabling business environ- among farming households remain (1) reducing ment, promoting investments in skills and the dependence of incomes on volatile yields by innovation, and enhancing the rule of law and ­ expanding agricultural insurance and encourag- property rights. However, a more detailed analy- ing income diversification into nonfarm activi- sis of the binding constraints to modernization ties and (2) fostering the shift of underemployed of the economy needs to be carried out to iden- workers out of the sector. The capacity of individ- tify the policy priorities within these broad cate- uals to diversify their incomes and undertake gories. An initial analysis indicates that poor more productive activities hinges crucially on the infrastructure and poor skills among labor force availability of jobs in more productive sectors, outside Accra are major potential culprits. The the extent to which the agricultural labor force lack of adequate skills is an especially important possesses the skills required by more productive constraint and is exacerbated by the considerable jobs, the degree of connectivity with markets and wage premium for public employment, which is with the areas where nonagricultural jobs are crowding out private sector employment among located, and the existence of well-defined land the more well-educated. title rights and a well-functioning land market. The occupational and geographical mobil- Reforms in these areas are critical. ity of workers must be fostered. The ongoing Further progress in poverty reduction and structural transformation requires a more highly shared prosperity requires a concerted effort to skilled labor force clustered around the urban boost the development of the modern sector. and peri-urban areas where job opportunities are In recent years, the economy has undergone a increasingly concentrated. The increase in edu- profound structural transformation (chapter 3). cational attainment among workers experienced However, Ghana still lacks a vibrant modern since 1991 has been a major factor in the suc- sector capable of offering good earnings ­ cesses so far. The concentration of skilled A Roadmap for Policy Action 51 workers in the areas of the country that were credit market. On the supply side, it will be cru- expanding economically was facilitated by the cial to ensure that workers have the skills required mass movement of workers to urban areas. by the new jobs and that the costs of migrating to Investment in infrastructure and in the provision urban areas are not prohibitively high. The suc- of public services to the growing urban popula- cess of the interventions will hinge on ensuring tion was promoted by the spatial and economic that economic transformation and migration transformation. Despite the inflow of 8 million expand beyond the large metropolitan areas into people into urban areas, average household con- smaller cities. sumption continued to increase, and poverty A small set of win-win policy areas are declined from about 30  percent in 1991 to 9 emerging as priorities in the effort to extend percent in 2012. Nonetheless, that much of the Ghana’s successes in poverty reduction and progress has been concentrated in the metropol- shared prosperity. The most immediate priority itan areas of Accra and Kumasi where most of the is to restore a sound macroeconomic environ- jobs are also clustered means that efforts must ment so as to enhance the prospects for growth now be undertaken to broaden the reach of the and foster additional economic transformation. transformation process to other areas to ensure Ensuring adequate investment in infrastructure that workers can shift into higher-productivity and skill development will be key to increasing jobs and enjoy higher earnings elsewhere, too. productivity in agriculture, creating modern sec- Several policies can foster structural trans- tor jobs, and ensuring that workers have the skills formation and help maximize the benefits of they need to take advantage of the new employ- urbanization. On the demand side, effective ment opportunities. Greater and better connec- interventions to unleash the potential of urban- tivity between rural and urban areas, combined ization include better land use management and with clearly defined land title rights and more planning in municipal and metropolitan areas, efficient land markets, will facilitate structural improvement in transport to connect markets transformation by allowing Ghana to benefit and boost factor mobility, and a more efficient from geographical agglomeration. 52 A Roadmap for Policy Action Appendix A Computing Poverty This appendix discusses the methodology to lines and the methods used to construct robust compute poverty using the Ghana Living price deflators. Standard Survey round 6 data. The World Bank provided technical assistance (TA) to the General Statistical Service of Ghana (GSS). The Rebasing the Poverty Line GSS and World Bank worked together to con- struct the household consumption expenditure The poverty line and extreme poverty line aggregates, adjust this measure of total expendi- (­measured in Ghanaian cedis) are designed to ture into real terms for comparison across space measure whether a household is poor or and time (back to 2005/2006 to compare with the extremely poor, by comparing the total house- previous survey, GLSS 5), and construct two hold consumption and food consumption to poverty lines to measure the poor and the each line. The basis for extreme poverty line is extreme poor. In a second stage of collaboration, the cost of buying a basic bundle of food which is a poverty report was produced. sufficient to give adequate calories and based on This effort faced two major challenges: “typical” consumption of the poor/near poor (in rebasing the poverty lines and computing reli- terms of food types and their quantities). The able price deflators. This brief note describes the poverty line is computed by adding an amount to methodology used to calculate the new poverty cover nonfood expenditures to the extreme pov- erty line. Two key decisions are made: the bundle Map A.1  Administrative Map of Ghana of food (types/quantities) to reach sufficient cal- ories and the nonfood amount (usually a fixed share of the food poverty line). The basis for both the basket of food items and the nonfood share had been selected in 1999 and had not been updated since then. Several new items had entered Ghanaian household consumption since then and general changes in consumption pat- terns had occurred (items such as DVD/VCD, MP3/MP4 players, vacuum cleaners, rice ­cookers, mobile phones, tablet PCs, etc.) In this regard, these aspects were deemed to be potentially out- dated and warranted revision. The GSS decided to ­re-compute the poverty lines based on the GLSS 6, to reflect changes in the food basket consumed by Ghanaian households. ­ In line with international practice, GSS cal- culated the average expenditure of the food consumption basket for the bottom 50 percent of individuals ranked by consumption per adult equivalent, and derived the amount of calories in this basket. The calorie price is then Computing Poverty 53 calculated by dividing the adult equivalent is January 2013 prices of the Greater Accra Region expenditure of the food basket by the amount of (to match the base for the poverty lines). adult equivalent calories provided by the basket. The Consumer Price Index (CPI) is one This calorie price is representative of the price source for converting nominal values into a paid by a typical household in the bottom real value. The current CPI, introduced in 50 percent. This price is then multiplied by 2,900 January 2012, is different from the previous in calories, which was used to compute the extreme terms of the basket used but also in terms of data poverty line for the GLSS 6. collection. The new CPI was not rescaled back to Following common practices in other devel- previous years. It is worth mentioning two inno- oping countries, expenditure on nonfood con- vations in the current CPI among others. First, sumption is added to the extreme poverty line data are collected separately in urban and rural calculated above. This nonfood basket is deter- areas, whereas previously this data were just mined by those whose total food expenditure is averaged; and second, Upper West and Upper about the level of the extreme poverty line East regions are now considered two different (10 percent individuals below and above the line). regions while before they were collected under This is based on Engel’s law, which states that the only one set of data. share of food expenditure decreases as household The new CPI covers only the last two years; income/expenditure increases. By  selecting the before January 2012 the only available data are population whose food consumption is around those from the old CPI. This posed a problem of the extreme poverty line, their nonfood expendi- comparability between the GLSS 5 (2005/2006) ture is used as the benchmark for estimating the and GLSS 6. Using the old CPI (for which data absolute poverty line. collection continued until the end of 2013) was The methodology used produced revised not a viable option. The old CPI indicated that poverty lines: an extreme poverty line of 792.05 between 2005 and 2012, prices varied by Ghanaian cedis and a poverty line of 1,314.00 160  percent (see table 1); food prices varied by Ghanaian cedis per equivalent adult per year in 100 percent. In terms of prices faced by the typi- the January 2013 prices of Greater Accra cal Ghanaian households, other evidence sug- Region. In U.S. dollar terms,1 the revised poverty gested that these were substantially lower than line is equivalent to about US$1.83 per day what households were experiencing. For exam- (US$1.10 for the extreme poverty line). The pov- ple, taking the old poverty line and using the old erty line indicates the minimum living standard CPI to inflate it to 2012/2013 prices resulted in a in Ghana while the extreme poverty line indicates much lower poverty rate for 2012/2013 (13.3 that even if a household spends their entire bud- percent, a 15 percentage point drop) than seems get on food, they still would not meet the mini- credible (table 2). Both GSS and World Bank mum calorie requirement. judged this result not credible and decided to look at alternative price deflators. GSS proceeded with another type of check, since both GLSS 5 and 6 have price modules. Price Deflators During household interviews, enumerators went Once one has defined poverty lines, to compare to local markets and collected price information the survey expenditure data, one needs to adjust on a set of food and nonfood items; this was a the nominal expenditures in the date into real rich source of information and potentially more values. Since the GLSS 6 was conducted over a accurate in describing the prices households period of 16 months and across the ­ country, the really face. Nonetheless, this dataset showed price levels facing households vary across both some limitations; the kilogram conversion of time and space. The base for the real expenditure some food items was sometimes inaccurate and 54 Computing Poverty Table A.1  Price Deflators (2005/06–2012/13) Mixed deflator Old CPI Old CPI—Food GLSS prices—Food Mixed deflator (food) (food+nonfood) Price Index (2005/2006 = 100) 260 200 460 290 330 Table A.2  Poverty Rates Using the Revised and the Old Poverty Lines with Different Price Deflators Source Old CPI Mixed deflator Year GLSS 5 GLSS 6 GLSS 5 GLSS 6 2005–06 2012–13 2005–06 2012–13 Revised poverty line 44.7 24.2 31.9 24.2 Old poverty line 28.5 13.3 28.5 21.4 Note: The final poverty rates reported by the GSS for the GLSS 6 analysis are 31.9 percent and 24.2 percent (using revised poverty line and mixed deflator). nonfood prices turned out to be incomparable The  combination of food and nonfood indexes between the two surveys. Nonetheless, using the yielded the mixed deflator, which showed an food price data in the household surveys, food overall price increase of 230 percent from price inflation looked quite different than from 2005/2006 to 2012/2013 (table A.1). the food CPI (table 1). Table A.2 presents poverty figures from Given the high data heterogeneity and GLSS 5 and GLSS 6, by revised and old poverty the need to reproduce, as much as possible, the lines and by CPI and new mixed deflator. For conditions households faced, GSS constructed temporal deflation of poverty lines different defla- a food price index that uses survey weights and tors are used. For spatial deflation the new CPI raw prices from CPI, henceforth called the food in 2012–13 and the old CPI in 2005–06 are used. mixed deflator. This method uses survey infor- Overall old and new CPIs spatially do not differ mation to reflect the food weights in household much. consumption more accurately combined with the Two points are important to mention. First, more systematic price data collected under the the revised poverty lines produce minimal CPI effort. The food mixed deflator shows, in ­ differences in the poverty rates for GLSS 6 the  period considered, an increase of prices of (table  A.2, 31.9 percent and 28.5 percent and 190 percent. 24.2 percent vs. 21.4 percent). By contrast, the The construction of the nonfood compo- use of different deflators changes the poverty nent was more intricate. Whereas for food rates significantly. When using the old CPI (and prices there was an almost prefect correspon- the old poverty line), poverty more than halved dence in coverage between the GLSS (used for from 2005–06 to 2012–13 (from 28.5 percent to weights) and CPI price data, this was not 13.3 percent). The new deflator’s estimate is the case for nonfood prices. Items asked in the much higher than the old CPI and, as such, GLSS were not the same as those collected in mechanically the fall in poverty is reduced. This the  CPI. To  overcome this problem, GSS esti- is because deflating consumption in 2012/2013 mated the nonfood index using the same Engel’s to 2005/2006 levels will produce a smaller level method used for the nonfood poverty line. consumption when there is more inflation. Note 1. Market exchange rate. Computing Poverty 55 Figure A.1  Distribution of Highest Education Level of Household Head, by Percentile and Year 60 50 40 P rc nt 30 20 10 0 1991 1998 2005 2012 1991 1998 2005 2012 1991 1998 2005 2012 20th percentile 40th percentile 60th percentile Primary Secondary Higher Source: Calculations based on GLSS 3–6. Figure A.2  Distribution of Type of Employment, by Percentile and Year 30 25 20 P rc nt 15 10 5 0 1991 1998 2005 2012 1991 1998 2005 2012 1991 1998 2005 2012 20th percentile 40th percentile 60th percentile Private employee Public employee Self-employed, nonagriculture Source: Calculations based on GLSS 3–6. 56 Computing Poverty Figure A.3  Distribution of Household Characteristics, by Percentile and Year 60 50 40 P rc nt 30 20 10 0 1991 1998 2005 2012 1991 1998 2005 2012 1991 1998 2005 2012 20th percentile 40th percentile 60th percentile Urban resident Household size Share of adult female Source: Calculations based on GLSS 3–6. Computing Poverty 57 Appendix B Regression Tables Table B.1  Probit Results Year 1991 Year 1998 Year 2005 Year 2012 Number of obs 4523 5998 8687 16772         Pseudo R2 0.2200 0.2540 0.2791 0.2610         Marg. Marg. Marg. Marg. Coef. Coef. Coef. Coef. ef. ef. ef. ef. Educational Up to primary –0.045 –0.018 –0.144** –0.053** –0.080 –0.026 –0.099** –0.025** attainment school Up to secondary –0.184*** –0.073*** –0.281*** –0.104*** –0.225*** –0.073*** –0.331*** –0.085*** school Higher than –0.724*** –0.288*** –0.490*** –0.182*** –1.067*** –0.348*** –0.825*** –0.211*** secondary school Employment Private workers –0.188* –0.075* –0.190* –0.071* –0.182** –0.059** –0.046 –0.012 category Public workers –0.327*** –0.130*** –0.352*** –0.131*** –0.366*** –0.119*** –0.130 –0.033 Nonagricultural –0.294*** –0.117*** –0.174*** –0.065*** –0.241*** –0.079*** –0.315*** –0.081*** self-employed Other –0.181* –0.072* 0.071 0.026 0.155** 0.051** –0.033 –0.009 Infrastructure Infrastructures –0.236*** –0.094*** –0.333*** –0.124*** –0.237*** –0.077*** –0.292*** –0.075*** index Location Western 0.196 0.078 –1.814*** –0.673*** –1.518*** –0.494*** –0.947*** –0.242*** Central –0.283* –0.113* –1.043*** –0.387*** –1.477*** –0.481*** –1.045*** –0.267*** Greater Accra 0.047 0.019 –2.057*** –0.764*** –1.115*** –0.363*** –1.193*** –0.305*** Volta –0.135 –0.054 –1.270*** –0.471*** –1.157*** –0.377*** –0.796*** –0.203*** Eastern –0.042 –0.017 –1.383*** –0.513*** –1.670*** –0.544*** –0.958*** –0.245*** Ashanti –0.237 –0.094 –1.441*** –0.535*** –1.338*** –0.436*** –1.019*** –0.261*** Brong Ahafo 0.148 0.059 –1.480*** –0.549*** –1.188*** –0.387*** –0.936*** –0.239*** Northern –0.103 –0.041 –1.003*** –0.372*** –1.067*** –0.347*** –0.533*** –0.136*** Upper East 0.772*** 0.307*** –0.202 –0.075 –0.420*** –0.137*** –0.616*** –0.158*** Urban area –0.554*** –0.220*** –0.109* –0.041* –0.405*** –0.132*** –0.333*** –0.085*** residence Household Household size 0.090*** 0.036*** 0.076*** 0.028*** 0.104*** 0.034*** 0.080*** 0.021*** composition Share of children 0.196 0.078 0.345** 0.128** 0.009 0.003 0.147 0.038 Share of –0.016 –0.006 0.073 0.027 –0.378** –0.123** 0.304** 0.078** care-dependent persons Household head 0.001 0.001 0.001 0.001 0.003 0.001 0.002 0.000 age Sex of household 0.034 0.014 0.009 0.003 –0.036 –0.012 –0.190*** –0.049*** head Share of adult −0.693*** −0.276*** −0.528*** −0.196*** −0.273* −0.089* −0.316*** −0.081*** males Share of adult −1.341*** −0.534*** −1.253*** −0.465*** −0.879*** −0.286*** −1.091*** −0.279*** females Constant 0.194 1.115*** 0.588*** 0.503*** Regression Tables 59 Table B.2  Quantile Regression, 1991 20th percentile 40th percentile 60th percentile R-squared   0.1734 0.2354 0.2822 Adj R-squared   0.1688 0.2311 0.2782 Root MSE   0.7561 0.7139 0.7165 Coef. P>t Coef. P>t Coef. P>t Educational attainment Up to primary school 0.027 0.603 –0.005 0.914 0.062 0.161 Up to secondary school 0.047 0.231 0.043 0.244 0.110 0.002 Higher than secondary school 0.206 0.002 0.203 0.021 0.321 0.001 Employment category Private workers 0.129 0.017 0.124 0.026 0.113 0.047 Public workers 0.146 0.003 0.154 0.001 0.204 0.000 Nonagricultural self-employed 0.114 0.006 0.141 0.000 0.182 0.000 Other 0.061 0.250 0.081 0.114 0.087 0.059 Infrastructure index Infrastructures 0.035 0.058 0.103 0.000 0.129 0.000 Location Western 0.096 0.343 –0.011 0.898 –0.149 0.052 Central 0.251 0.011 0.298 0.000 0.140 0.086 Greater Accra 0.116 0.230 0.067 0.431 –0.005 0.955 Volta 0.193 0.047 0.114 0.176 0.013 0.870 Eastern 0.151 0.139 0.089 0.311 –0.015 0.851 Ashanti 0.217 0.022 0.192 0.017 0.102 0.187 Brong Ahafo 0.122 0.226 0.006 0.941 –0.086 0.261 Northern –0.108 0.315 –0.017 0.842 0.075 0.355 Upper East –0.465 0.001 –0.302 0.003 –0.310 0.000 Urban area residence 0.280 0.000 0.309 0.000 0.314 0.000 Household composition Household size –0.030 0.000 –0.034 0.000 –0.047 0.000 Share of children –0.193 0.048 –0.114 0.194 –0.208 0.015 Share of care-dependent persons –0.016 0.866 –0.030 0.748 0.046 0.618 Household head age 0.000 0.905 –0.001 0.359 –0.001 0.606 Sex of household head –0.092 0.014 –0.038 0.309 –0.028 0.395 Share of adult males 0.268 0.002 0.438 0.000 0.407 0.000 Share of adult females 0.263 0.007 0.604 0.000 0.675 0.000 Table B.3  Quantile Regression, 1998 20th percentile 40th percentile 60th percentile R-squared   0.2470 0.2836 0.3103 Adj R-squared   0.2439 0.2806 0.3074 Root MSE   0.8597 0.7444 0.7687 Coef. P>t Coef. P>t Coef. P>t Educational attainment Up to primary school 0.151 0.001 0.091 0.018 0.005 0.885 Up to secondary school 0.192 0.000 0.160 0.000 0.132 0.000 Higher than secondary school 0.216 0.000 0.238 0.000 0.298 0.000 Employment category Private workers 0.127 0.008 0.126 0.008 0.153 0.005 Public workers 0.258 0.000 0.178 0.000 0.180 0.000 Nonagricultural self-employed 0.143 0.000 0.122 0.000 0.152 0.000 Other –0.017 0.714 0.003 0.945 0.039 0.296 Infrastructure index Infrastructures 0.109 0.000 0.158 0.000 0.194 0.000 table continues next page 60 Regression Tables Table B.3  continued 20th percentile 40th percentile 60th percentile R-squared   0.2470 0.2836 0.3103 Adj R-squared   0.2439 0.2806 0.3074 Root MSE   0.8597 0.7444 0.7687 Coef. P>t Coef. P>t Coef. P>t Location Western 1.429 0.000 0.938 0.000 0.625 0.000 Central 1.071 0.000 0.491 0.000 0.260 0.000 Greater Accra 1.292 0.000 0.919 0.000 0.764 0.000 Volta 1.035 0.000 0.622 0.000 0.400 0.000 Eastern 1.300 0.000 0.711 0.000 0.401 0.000 Ashanti 1.222 0.000 0.728 0.000 0.484 0.000 Brong Ahafo 1.265 0.000 0.727 0.000 0.472 0.000 Northern 0.722 0.000 0.406 0.000 0.327 0.000 Upper East 0.291 0.017 0.079 0.159 0.074 0.108 Urban area residence 0.068 0.078 0.053 0.118 0.077 0.020 Household composition Household size –0.041 0.000 –0.044 0.000 –0.053 0.000 Share of children –0.089 0.318 –0.161 0.035 –0.106 0.172 Share of care-dependent persons –0.012 0.891 –0.001 0.988 0.083 0.295 Household head age 0.000 0.991 0.000 0.757 –0.003 0.011 Sex of household head 0.027 0.456 –0.009 0.758 0.029 0.338 Share of adult males 0.084 0.325 0.227 0.002 0.372 0.000 Share of adult females 0.406 0.000 0.550 0.000 0.739 0.000 Table B.4  Quantile Regression, 2005 20th percentile 40th percentile 60th percentile R-squared   0.2835 0.3104 0.3171 Adj R-squared   0.2814 0.3084 0.3151 Root MSE   0.8953 0.7719 0.7290 Coef. P>t Coef. P>t Coef. P>t Educational attainment Up to primary school 0.041 0.416 0.072 0.065 0.086 0.008 Up to secondary school 0.144 0.000 0.162 0.000 0.140 0.000 Higher than secondary school 0.268 0.000 0.398 0.000 0.524 0.000 Employment category Private workers 0.053 0.253 0.152 0.000 0.184 0.000 Public workers 0.137 0.002 0.154 0.001 0.157 0.002 Nonagricultural self-employed 0.103 0.007 0.201 0.000 0.202 0.000 Other –0.096 0.036 –0.009 0.811 0.007 0.829 Infrastructure index Infrastructures 0.088 0.000 0.115 0.000 0.144 0.000 Location Western 1.472 0.000 0.690 0.000 0.280 0.000 Central 1.397 0.000 0.693 0.000 0.272 0.000 Greater Accra 1.235 0.000 0.473 0.000 0.061 0.177 Volta 1.334 0.000 0.480 0.000 0.121 0.007 Eastern 1.506 0.000 0.783 0.000 0.293 0.000 Ashanti 1.329 0.000 0.611 0.000 0.253 0.000 Brong Ahafo 1.304 0.000 0.534 0.000 0.161 0.000 Northern 0.914 0.000 0.409 0.000 0.201 0.000 Upper East 0.317 0.001 0.080 0.139 –0.070 0.083 Urban area residence 0.154 0.000 0.259 0.000 0.276 0.000 table continues next page Regression Tables 61 Table B.4  continued 20th percentile 40th percentile 60th percentile R-squared   0.2835 0.3104 0.3171 Adj R-squared   0.2814 0.3084 0.3151 Root MSE   0.8953 0.7719 0.7290 Coef. P>t Coef. P>t Coef. P>t Household composition Household size –0.052 0.000 –0.054 0.000 –0.044 0.000 Share of children 0.054 0.552 –0.049 0.505 0.036 0.602 Share of care-dependent persons 0.223 0.005 0.125 0.087 0.174 0.011 Household head age –0.002 0.137 –0.001 0.447 –0.003 0.004 Sex of household head 0.017 0.670 0.054 0.124 0.048 0.139 Share of adult males 0.060 0.454 0.126 0.055 0.342 0.000 Share of adult females 0.295 0.001 0.479 0.000 0.718 0.000 Table B.5  Quantile Regression, 2012 20th percentile 40th percentile 60th percentile R-squared   0.2469 0.3065 0.3106 Adj R-squared   0.2458   0.3055   0.3096 Root MSE   0.9049   0.7731   0.7826 Coef. P>t Coef. P>t Coef. P>t Educational attainment Up to primary school 0.177 0.000 0.076 0.009 0.044 0.087 Up to secondary school 0.263 0.000 0.172 0.000 0.175 0.000 Higher than secondary school 0.315 0.000 0.381 0.000 0.473 0.000 Employment category Private workers 0.010 0.819 0.088 0.008 0.109 0.001 Public workers 0.079 0.048 0.101 0.020 0.117 0.016 Nonagricultural self-employed 0.147 0.000 0.199 0.000 0.189 0.000 Other 0.032 0.364 0.017 0.577 0.020 0.527 Infrastructure index Infrastructures 0.173 0.000 0.174 0.000 0.186 0.000 Location Western 0.868 0.000 0.445 0.000 0.219 0.000 Central 0.929 0.000 0.393 0.000 0.093 0.009 Greater Accra 0.869 0.000 0.558 0.000 0.414 0.000 Volta 0.771 0.000 0.349 0.000 0.159 0.000 Eastern 0.892 0.000 0.401 0.000 0.126 0.000 Ashanti 0.913 0.000 0.433 0.000 0.186 0.000 Brong Ahafo 0.857 0.000 0.412 0.000 0.167 0.000 Northern 0.505 0.000 0.169 0.000 0.037 0.173 Upper East 0.623 0.000 0.238 0.000 0.062 0.061 Urban area residence 0.197 0.000 0.168 0.000 0.118 0.000 Household composition Household size –0.052 0.000 –0.053 0.000 –0.047 0.000 Share of children 0.009 0.898 0.063 0.308 –0.049 0.435 Share of care-dependent persons 0.019 0.779 –0.057 0.364 –0.007 0.921 Household head age –0.001 0.246 –0.002 0.012 –0.002 0.047 Sex of household head 0.086 0.006 0.088 0.001 0.066 0.025 Share of adult males 0.040 0.522 0.269 0.000 0.456 0.000 Share of adult females 0.380 0.000 0.625 0.000 0.746 0.000 62 Regression Tables Table B.6  Oaxaca-Blinder* Poverty Decomposition by 40th and 60th Percentiles, Variation between 1991 and 2012 40th percentile variations 60th percentile variations Endowments Coefficients Endowments Coefficients Educational attainment Up to primary school 0.000 0.009 0.006 –0.002 Up to secondary school 0.002 0.052 0.005 0.026 Higher than secondary school 0.009 0.005 0.013 0.004 Employment category Private workers 0.011 –0.002 0.010 0.000 Public workers –0.010 –0.007 –0.013 –0.011 Nonagricultural self-employed 0.008 0.011 0.010 0.001 Other 0.001 –0.006 0.001 –0.007 Infrastructure index Infrastructures 0.075 –0.029 0.095 –0.023 Location Western 0.000 0.045 0.001 0.037 Central –0.004 0.010 –0.002 –0.005 Greater Accra 0.003 0.058 0.000 0.049 Volta 0.001 0.019 0.000 0.012 Eastern –0.003 0.044 0.001 0.020 Ashanti 0.007 0.038 0.004 0.013 Brong Ahafo 0.000 0.047 0.002 0.029 Northern 0.000 0.018 0.000 –0.004 Upper East 0.004 0.029 0.004 0.020 Urban area residence 0.053 –0.046 0.054 –0.064 Household composition Household size 0.018 –0.120 0.026 –0.001 Share of children 0.001 0.037 0.003 0.033 Share of care-dependent persons 0.000 –0.001 0.000 –0.002 Household head age –0.001 –0.041 –0.001 –0.050 Sex of household head –0.004 0.080 –0.003 0.060 Share of adult males 0.016 –0.036 0.015 0.010 Share of adult females 0.022 0.005 0.025 0.018 * The change in consumption between 1991 and 2012 by percentile is decomposed as follows. For example, looking at the 60th percentile, being Y91 = X91 b91 + e91 the linear model for 1991 and Y12 = X12 b12 + e12 the linear model for 2012 as long as E( b91) = E(e12) = 0, the mean consumption difference between the two years can be decomposed as DY = (X12 – X91) b91 + (b12 – b91)X91 + (X12 – X91) (b12 – b91) = Endowments Effect + Coefficients Effect + Interaction. Regression Tables 63 References Adams, Arvil V., Sara Johansson de Silva, and Setareh Duclos, Jean-Yves, Joan Esteban, and Debraj Ray. 2004. Razmara. 2013. Improving Skills Development in the “Polarization: Concepts, Measurement, Estimation.” Informal Sector: Strategies for Sub-Saharan Africa. Econometrica 72 (6): 1737–72. Directions in Development: Human Development. Elbers, Chris, Jean O. Lanjouw, and Peter F. Lanjouw. Washington, DC: World Bank. 2003. “Micro-Level Estimation of Poverty and Agyei-Mensah, S., and G. Owusu. 2010. “Segregated Inequality.” Econometrica 71 (1): 355–64. by Neighbourhoods? A Portrait of Ethnic Diversity Esteban, Joan, and Debraj Ray. 1999. “Conflict and in the Neighbourhoods of the Accra Metropolitan Distribution.” Journal of Economic Theory 87 (2): Area, Ghana.” Population, Space and Place 16: 379–415. 499–516. ———. 2008. “Polarization, Fractionalization and Azevedo, João Pedro, Gabriela Inchauste, Sergio Conflict.” Journal of Peace Research 45 (2): 163–82. Olivieri, Jaime Saavedra, and Hernan Winkler. ———. 2011. “Linking Conflict to Inequality and 2013. “Is Labor Income Responsible for Poverty Polarization.” American Economic Review 101 (4): Reduction? A Decomposition Approach.” Policy 1345–74. Research Working Paper 6414, World Bank, Falco, Paolo, Andrew Kerr, Pierella Paci, and Bob Washington, DC. Rijkers. 2014. Working toward Better Pay: Earning Bardasi, Elena, Kathleen Beegle, Andrew Dillon, and Dynamics in Ghana and Tanzania. World Bank Pieter Serneels. 2010. “Do Labor Statistics Depend Study 88483. Washington, DC: World Bank. on How and to Whom the Questions Are Asked: Filmer, Deon, and Louise Fox. 2014. Youth Employment Results from a Survey Experiment in Tanzania.” in Sub-Saharan Africa. With Karen Brooks, Policy Research Working Paper 5192, World Bank, Aparajita Goyal, Taye Mengistae, Patrick Premand, Washington, DC. Dena Ringold, Siddharth Sharma, and Sergiy Bhattacharya, N. and B. Mahalanobis. 1967. “Regional Zorya. Africa Development Forum Series. Disparities in Household Consumption in India.” Washington, DC: World Bank and Agence Française Journal of the American Statistical Association 62 de Développement. (317): 143–61. Filmer, Deon, and Lant H. Pritchett. 2001. “Estimating Breisinger, Clemens, Xinshen Diao, Shashidhara Wealth Effects without Expenditure Data—Or Kolavalli, and James Thurlow. 2008. “The Role of Tears: An Application to Educational Enrollments Cocoa in Ghana’s Future Development.” GSSP in States of India.” Demography 38 (1): 115–32. Background Paper 0011, Ghana Strategy Support Fortin, Nicole, Thomas Lemieux, and Sergio Firpo. Program, International Food Policy Research 2011. “Decomposition Methods in Economics.” In Institute, Accra, Ghana. Handbook of Labor Economics, vol. 4A, edited by Clementi, Fabio, Andrew L. Dabalen, Vasco Molini, Orley Ashenfelter and David Card, 1–102. San and Francesco Schettino. 2015. “When the Centre Diego, CA: Elsevier. Cannot Hold: Patterns of Polarization in Nigeria.” Fox, Louise. 2015. “Are African Households Review of Income and Wealth. September. Heterogeneous Agents? Stylized Facts on Patterns Clementi, Fabio, Vasco Molini, and Francesco of Consumption, Employment, Income, and Schettino. 2015a. “Polarization and Its Discontents.” Earnings for Macroeconomic Modelers.” IMF Unpublished Manuscript. Working Paper WP/15/102, International ———. 2015b. “Poverty and Social Impact Analysis Monetary Fund, Washington, DC. (PSIA): The Impact of Reductions in Fuel and GSS (Ghana Statistical Service). 2014. “Poverty Profile Electricity Subsidies.” Unpublished Manuscript. in Ghana (2005–2013).” August, GSS, Accra, Darvas, Peter, Marta Favara, and Tamara Arbnold. Ghana. Forthcoming. World Bank, Washington, DC. Handcock, Mark S., and Martina Morris. 1998. “Relative Deaton, Angus S., and Salman Zaidi. 2002. Guidelines Distribution Methods.” Sociological Methodology for Constructing Consumption to Aggregates for 28 (1): 53–97. Welfare Analysis. LSMS Working Paper 135. ———. 1999. Relative Distribution Methods in the Living Standards Measurement Study. Washington, Social Sciences. Statistics for Social Science and DC: World Bank. Public Policy Series. New York: Springer-Verlag. References 65 Haughton, Jonathan, and Shahidur R. Khandker. Ranzani, M., and Michele Tuccio. Forthcoming. “Is 2009. Handbook on Poverty and Inequality. Government Employment Attractive? Public Wage Washington, DC: World Bank. Premiums and Job Queues.” Mimeo. Kolenikov, Stanislav, and Anthony Shorrocks. 2003. Rao, V. M. 1969. “Two Decompositions of Concen­ “A Decomposition Analysis of Regional Poverty in tration Ratio.” Journal of the Royal Statistical Society, Russia.” UNU-WIDER, Helsinki. Series A (General) 132 (3): 418–25. Ministry of Agriculture. 2012. “Agriculture in Ghana, Sandel, Michael J. 2012. What Money Can’t Buy: The Facts and Figures.” http://mofa.gov.gh/site​ Moral Limits of Markets. New York: Farrar, Straus /­w p-content/uploads/2011/04​ /­m ofa_facts_and​ and Giroux. _­figures.pdf. Teal, F., A. Zeitlin, and H. Maamah. 2006. “Ghana Mohan, G., and H. Matsuda. 2013. “Regional Level Cocoa Farmers Survey 2004: Report to Ghana Total Factor Productivity Growth in Ghana Cocoa Board.” Centre for the Study of African Agriculture.” Journal of Economics and Sustainable Economies, Oxford University. http://www.gprg​ Development 4 (5). .org/pubs/reports/pdfs/2006-04-teal​-­zeitlinmaamah​ Molini, Vasco, Michiel Keyzer, Bart van den Boom, .pdf. Wouter Zant, and Nicholas Nsowah-Nuamah. 2010. Vigneri, M. 2005. “Trade Liberalization and Agricultural “Safety Nets and Index-Based Insurance: Historical Performance: Micro and Macro Evidence on Cash Assessment and Semiparametric Simulation for Crop Production in Sub-Saharan Africa.” PhD Northern Ghana.” Economic Development and thesis, Oxford University, Oxford, UK. ­ Cultural Change 58 (4): 671–712. WDI (World Development Indicators) (database), Molini, Vasco, D. Pavelesku, and M. Ranzani. “Internal World Bank, Washington, DC. http://data.­worldbank​ migration in Ghana in the last decade.” Mimeo. .org/data-catalog/world​-development-indicators. Morris, Martina, Annette D. Bernhardt, and Mark S. World Bank. 2015a. “First Macroeconomic Stability Handcock. 1994. “Economic Inequality: New for Competitiveness and Growth Development Methods for New Trends.” American Sociological Policy Financing.” Report number 955284- Review 59 (2): 205–19. GH2015, World Bank, Washington, DC. Oaxaca, R. 1973. “Male-Female Wage Differentials in ———. 2015b. “Rising through Cities in Ghana: Ghana Urban Labor Markets”. International Economic Urbanization Review, Overview Report.” April, Review 14 (3): 693–709. World Bank, Washington, DC. PovcalNet (online analysis tool), World Bank, ———. Forthcoming. “Ghana: Social Protection Washington, DC, http://iresearch.worldbank.org​ Assessment and Expenditure Review.” World Bank, /­PovcalNet/. Washington, DC. 66 References SKU K8485