Africa s Growth, Poverty and Inequality Nexus - Fostering Shared Prosperity. Luc Christiaensen, Punam Chuhan-Pole, and Aly Sanoh 1

Transcription

1 1 Africa s Growth, Poverty and Inequality Nexus - Fostering Shared Prosperity Luc Christiaensen, Punam Chuhan-Pole, and Aly Sanoh 1 DRAFT - July 2013 Abstract An uptick in economic growth in Sub-Saharan Africa since the mid-1990s is turning attention to the region s achievements in income poverty reduction and human welfare, especially in resource-rich countries, where growth has been fastest. Reviewing the record confirms the challenge for these countries to harness their resource wealth for the larger benefit of their populations. Income poverty reduction and human development indicators remain generally well below those in resource-poor countries. While data limitations challenge an accurate and refined assessment, high initial inequality and increasing resource dependence suggest a lower conversion rate of Africa s growth into poverty reduction moving forward. Three opportunities to enhance Africa s growth elasticity of poverty are discussed. These include a direct universal transfer of a fixed proportion of countries mineral riches to their citizens, putting a particular attention to staple crop productivity in agriculture investment efforts, and secondary town development (as opposed to metropolitization) to guide urbanization and generate off-farm jobs accessible to the poor and needed to benefit from the emerging demographic dividend. Keywords: Poverty, growth, inequality, resource rents, Sub-Saharan Africa, agriculture, secondary towns, transfers. JEL codes: D31, O10, O12 1 Luc Christiaensen (lchristiaensen@worldbank.org), Punam Chuhan-Pole (pchuhan@worldbank.org), and Aly Sanoh are economists in the Chief Economists Office of the Africa Region at the World Bank. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

2 2 1 Introduction After several years of contraction during the 1980s, economic growth in Africa picked up in the mid- 1990s, with per capita GDP expanding at 2.4 percent per year on average 2 since 1996, increasing GDP per person by about 30 percent. Twenty-one Sub-Saharan African countries are now enjoying middle income status (MIC). Ten more are projected to do so by Vast land resources in an era of high food prices, potential agglomeration economies from rapid urbanization, a demographic dividend around the corner, and last but not least, the prospects of large revenues from mineral exploitation provide many observers with further ground for optimism. African lives have already greatly improved over the past decade. The next 10 years will be even better, The Economist proclaimed in opening its March 2013 Special Report Emerging Africa A Hopeful Continent. Less than a decade ago, pictures of hunger and civil strife were dominating the headlines. Has Sub-Saharan Africa turned the corner? Given improved macro-economic policies and increased political stability, the prospects of sustained growth are considered strong, with six out of the world s ten fastest growers currently in Sub-Saharan Africa 4. Stress tests further indicate that Sub-Saharan Africa stands to weather a sluggish recovery in the US following the budgetary sequester, renewed disruption in the EU economic zone or a hard landing in China reasonably well. Each of these would reduce Africa s projected growth of 5-6 % by one percentage point at most (1, 0.9, and 0.3 percentage points respectively), given their limited projected impact on commodity price, though larger damage would occur if these events were to coincide (World Bank, Global Economic Prospects 2013). The larger risk to sustained growth finds itself on the domestic front, especially drought and political and social tensions (Devarajan et al., 2013). The continued risk of renewed political tensions and civil strife within a number of Africa s countries, and their potential spillover into neighboring countries, has been most recently demonstrated in Mali. At the same time, increasing concerns are expressed that the new-found momentum is not benefiting the population at large. 5 To be sure, substantial strides forward have been recorded in SSA s human development, most pronounced in education, with (net) primary school enrollment increasing from 52 to 70 percent between 1995 and Sixty-three percent of all adults are now literate, up from 53 percent in Progress has also been made in health, with average child mortality declining from 175 to 125 per 1000 births between 1990 and 2010 (Global Monitoring Report, 2012). The prevalence of children underweight has declined from 27 to 22 percent. But, levels of illiteracy and malnutrition remain high and progress has been uneven. Some countries such as Rwanda and Niger have been 2 This is the average growth (unweighted) rate across all countries. The weighted per capita GDP growth is [1.7] percent. 3 Devarajan and Fengler (2012). 4 These are based on estimates and forecast of The Economist and IMF. The countries were Angola, Nigeria, Ethiopia, chad, Mozambique and Rwanda. For the period it is forecasted that 7 out 10 fastest growers will be in Africa. The countries will be Ethiopia, Mozambique, Tanzania, Congo, Ghana, Zambia and Nigeria. 5 McKay (2013). Bob Collymore, chief executive of Safaricom, Kenya s leading telecoms company provides just one expression of such concerns (Financial Times, March 11, 2013): People are now driving bigger and bigger BMWs and bigger and bigger Range Rovers; but [the poor] are in the same place and that I think presents us with a bigger challenge.

3 3 making fast progress 6, while others such as South Africa and Lesotho are seemingly stuck. Especially, and surprisingly, it is the resource-rich countries such as Gabon, Equatorial Guinea, and Nigeria that are disproportionately among the poorer performers. Despite GDP growth rates that have on average been twice as high over the past decade than those in resource-poor countries, their (non-income) human development indicators are on average still well below those of the latter group. There has also been progress in the fight against income poverty. Between 1990 and 2010, the share of people living on less than $1.25-day has been estimated to decline from 56.5 to 48.5 percent and the decline accelerated in the latter years (from 52.3 percent in 2005 to 48.5 percent in 2010). In 2010, the number of poor people was estimated at 413 million compared with 394 million in At the same time, the averages hide a great deal of diversity in performance here as well, even among Africa s faster growers. During the 2 nd half of the 2000s, Ethiopia and Rwanda saw their economies expand at 8-10 percent (or between 5 and 7 percent per capita), resulting in a 1.3 to 2.4 percentage point annual reduction in the national poverty headcount. In contrast, the national poverty headcounts were estimated to have declined only by 1 to 2.5 %point during the whole period in Nigeria, Zambia, and Tanzania, despite robust annual GDP growth of 6 to 7 percent (or about 4 percent per capita). More limited declines in the poverty headcount were also regularly observed in the other countries. While the quality of the national accounts, poverty and price statistics all remain wanting to different degrees, the broad picture emerging from the data is that Africa s economies have been expanding robustly and that poverty is coming down, but that more could also be done to ensure more inclusive growth and faster poverty reduction. The recent discovery of new mineral resources and the emergence of a demographic dividend, two of the defining fault lines in Africa s current development, provide important opportunities in this regard. But translating these into shared prosperity, or, in technical terms increasing the growth elasticity of poverty is by no means automatic. Without abdicating the unrelenting need for maintaining and strengthening economic growth, rather the contrary, this paper goes one step further and discusses some specific program entry points and broader policy directions to make growth in SSA more poverty reducing. First, and likely most controversial, a universal and uniform transfer of part of the mineral wealth directly to its citizens, is proposed for mineral rich countries. This measure has the potential to foster accountability and be more effective in generating inclusive growth in mineral rich countries, than many of the implicit alternatives, including the hugely popular fertilizer and petrol subsidy programs. Second, it is highlighted that in absorbing Africa s fast growing labor force, in effect, the world s work force of the future, informal will remain normal for quite some time. This suggests policy emphasis on increasing productivity in household enterprises and smallholder farms (as opposed to reducing unemployment through formal employment growth alone). In this, the implications for the speed of poverty reduction of different urbanization and agricultural growth patterns are elucidated, with particular emphasis on secondary town development, staple crop productivity, rural public goods and 6 The non-income human development indicator for Rwanda and Niger increased from 0.23 to 0.43 and 0.19 to 0.30 between 1990 and 2011 respectively.

4 4 commercial smallholder farming. Throughout, the need for better statistics to guide these different processes is emphasized. Our paper is related to a strand of literature that tries to explain the dynamic of economic growth, poverty, and income distribution. Dollar and Kray (2002) show that all income groups benefit proportionally from increases in economic growth and that income distribution does not matter for the growth elasticity of poverty. Adams (2004) using a data set of 126 intervals from 60 developing countries found that economic growth reduces poverty only when growth is measured by changes in survey mean income (or consumption) but not by GDP per capita. Ravallion (2007) finds that high inequality dampens the poverty-reducing effects of economic growth. For studies that focus on the evolution of growth and poverty in Sub-Saharan Africa specifically, Ali and Thorbecke (2000) found that poverty has increased where inequality has increased in both rural and urban areas. It also found that the effect of income distribution on poverty is stronger than that of the economic growth. Similarly, Fosu (2009) found that economic growth reduces all three measures of poverty (poverty headcount, poverty gap, and squared poverty gap) and that the growth elasticity is a decreasing function of initial inequality. More recently, Mckay (2013) focusing on 25 of the largest countries in SSA, found a substantial decrease in poverty rates over the past two decades but to a lesser extent than studies reporting elasticities based on national accounts. Likewise, Sala-i-Martin and Pinhovskiy (2010) argue that African poverty and inequality decline has been large and widespread. Other similar studies that draw on asset indices show a much more pronounced positive growth associated with decline in poverty. For example, Young (2012) and Sahn and Stifel (2000) using DHS asset indices found that growth and poverty reduction have been much faster than suggested by official national account statistics. Nonetheless, these asset-based studies have been challenged by Harttgen, Klasen and Vollmer (2013) which finds no evidence supporting the claim of an African miracle in growth and poverty reduction beyond what is reported from national accounts. Our paper is also related to the literature on the intervening factors that explain the low conversion of growth into poverty reduction. For example, Ghani et al. (2013) and Thorbecke (2013) argue that it is the combination of high endemic poverty and inequality that is responsible for low growth elasticity of poverty. These initial conditions remain relevant for Sub-Saharan Africa where countries have stubbornly high level of poverty and inequality by historical standards. High initial poverty and inequality reduces directly the growth rate but also indirectly the poverty-reducing effect of this growth. Other studies emphasized the source and pattern of growth. In particular, growth originating from agriculture has been shown to be more poverty reducing specifically in SSA where the great majority of economies remain agriculture-based and therefore more foundations for large employment (World Bank, 2008; Dercon, 2009; Christianensen et al. 2011). This paper revisits the above issues based on the most current poverty data from the World Bank (PovcalNet 2012). The novelty of our paper lies in the fact that it systematically explores the link between resource rents, growth, inequality and poverty. It proceeds by reviewing Africa s record on growth, inequality and poverty over the past decades and elucidates some of the proximate causes for its lesser performance in translating growth in GDP into poverty reduction, with special attention to the role of countries resource dependence (section 2). Three opportunities to increase the elasticity of

5 5 poverty with respect to GDP in Africa are subsequently discussed, with each corresponding to key deeper running trends in Africa s development today, i.e. the rapid onset of mineral wealth discovery, the high world food price environment combined with the region s comparative advantage in agriculture, and the emergence of a demographic dividend and urbanization. As highlighted in section 3, benefiting from these opportunities will, however, not be automatic. Appropriate policies and institutions will be required to unleash this potential so as to benefit the poorer segments of the population. Section 4 concludes. 2 Has economic growth translated into poverty reduction? 2.1 Data and methodology To examine the link between growth and poverty in Sub-Saharan Africa we look at the evidence from household expenditure surveys for the period [ ]. [Eighty five] surveys for [26] countries are available in the World Bank s PovCal tool. The idea is to see the change in poverty rate between surveys and the interplay of this change with per capita consumption growth and other factors over the same period. To analyze the question of how much change in poverty rate may be expected from a unit percentage change in income/consumption per capita while controlling for changes in initial conditions, we use a country fixed effect regression that includes initial conditions and interaction terms considered to be a standard approach in this literature (Bourguignon (2003), Fosu (2009), Kalwij and Verschoor (2007). P it = α + βg it + λi it + µx it + Ω*Interactions + u it Where P it is the outcome the poverty headcount; α is the intercept; β is the independent effect of annual per capita income growth rates (G) on poverty; λ is the independent effect of annual changes in inequality on poverty; X it includes initial inequality, initial resource rents, and dummy for resource rich countries; the interaction terms include interaction of per capita income growth rates with initial Gini, initial resource rents, and the resource dummy. To examine the link between changes in poverty and growth in average consumption, differences in the log of the former at the beginning and end of the spell are regressed on differences of the logs of the latter as well as differences in the logs of inequality. The differences are each time annualized by division through the length of the spell. In order to correct some unlikely wide fluctuations in the measured variables, we applied the BACON procedure (Weber, 2010) to remove outliers. The regressions are estimated using ordinary least squares and control for unobserved country fixed effects. Poverty,

6 6 consumption growth and inequality are measured by the $1.25 poverty headcount, average monthly consumption observed in the survey (in 2005 purchasing power parity), and the Gini coefficient reported for each survey in the World Bank s Povcal database. Regressions are run separately for SSA and the rest of the world with outlier corrections based on the BACON procedure. In SSA, 9 observations were dropped from the analysis (reducing the sample to 78), in the rest of the world 26 observations were dropped (reducing the sample from 450). To explore how initial consumption levels, inequality and resource dependence affect the elasticity of poverty with respect to consumption, the initial level of consumption, initial inequality and an indicator taking the value of 1 if the share of resource rents in GDP is larger than 5 % at the beginning of the spell (and zero otherwise) are included. These capture their average and independent effects on changes in poverty. In addition, household consumption growth is interacted with the reciprocal of initial consumption, as well as with initial inequality and the resource dependence indicator to explore their marginal effects on the rate of poverty reduction. Our data on natural resource rents is from the World Development Indicators (WDI, 2012). Data on poverty and inequality, which are derived from household based surveys over various years, are from the World Bank online poverty analysis tool PovcalNet. It is a much broader sample than what have been used in previous studies. We exclude high income countries based on the World Bank s classification in What do the data show? Higher, but more volatile growth in Africa s resource-rich countries Between 1995 and 2010, Sub-Saharan Africa s per capita GDP (constant $2000) rose from $500 to $630xx, more than a 25 per cent increase. There are several reasons for this turnaround, including improved macroeconomic policies, increased foreign aid, and the substantial reduction of external debts. Buoyant commodity prices and the expansion of mineral resource exploitation in a number of countries during the 2000s also played an important role, with GDP per capita in resource-rich countries on average growing 2.2 times faster during than in resource-poor countries 7 (Figure 1). How resource-rich countries have fared in terms of economic growth and well-being in Sub-Saharan Africa takes on particular significance, given its impact on Africa s growth over the past decade as well as the 7 A country is defined as resource rich if on average more than 5 percent of its GDP has been derived from oil and non-oil minerals (excluding forests) during The countries included are: Angola, Botswana, Cameroon, Chad, Democratic Republic of Congo, Republic of Congo, Côte D Ivoire, Equatorial Guinea, Gabon, Guinea, Liberia, Mali, Mauritania, Namibia, Nigeria, Sierra Leone, Sudan, and Zambia. Botswana, Namibia and Sierra Leone are also categorized as resource rich because the rent database omits diamonds, which account for a substantial share of GDP in these economies. Cote D Ivoire and Mali were added because of the rapidly increasing share of resources in total export and government revenues over the past 10 years. Ghana has not been included, as resource extraction has begun only a couple of years ago.

7 recent spate of new mineral discoveries. 8 Indeed, 13 of the 21 countries that are middle income today are also resource rich. Looking forward, it is expected that by 2020, only 4 or 5 countries in the region will not be involved in mineral exploitation, pointing to the continued importance of natural resources for the region s growth prospects. Figure 1: Economic growth in Sub-Saharan Africa s resource-rich and resource-poor countries during (unweighted average growth rates) Mean GDP/cap Growth (%) by Resource Type in SSA Res. Rich Res. Poor Source: Authors calculations based on World Development Indicators database. Part of the difference in per capita growth performance between resource-rich and resource-poor countries comes from their lower population growth. Looking at overall GDP growth instead of GDP growth per capita), resource-rich countries grew 1.5 (as opposed to 2.2) times faster 9. Nonetheless, the 8 Resource rents expanded rapidly during the 2000s in the traditional resource exporters (Equatorial Guinea, Republic of Congo, Angola, Gabon, Nigeria, Mauritania) where natural resource rents (from oil, coal, gas, minerals, but not forestry) accounted on average for between 12 and 62 percent during These countries were further joined by a number longstanding resource rich countries whose resource shares in GDP rose again above 5 percent (Zambia, Cameroon, Congo DRC, Guinea) as well a series of newcomers which also saw a rapid expansion of their resource related GDP in the first and second half of the 2000s (Sudan, Chad and Mozambique, Mali respectively, most recently also joined by Ghana). And press reports of new findings, also in new countries, continue. Recent findings of natural gas in Mozambique (estimated at 124 trillion cubic feet) and Tanzania (estimated at 28 trillion cubic feet) could for example yield $10 billion and $2 billion in annual government revenues respectively. Mining of Sierra Leone s Tonkolili iron ore deposits, estimated at 10.5 billion tons, is expected to boost GDP growth to 25 percent in 2012 compared with around 5 percent in 2011 (World Bank, Africa s Pulse, 2012). 9 The contrast is partly driven by the even larger discrepancy in growth performance between resource poor and resource rich fragile states. While GDP per capita in resource poor fragile states shrunk on average by -0.7 percent per year during , it expanded by 3.8 percent in their resource rich counterparts. In non-fragile states, resource poor and resource rich countries grew on average by 2.2 and 3.6 percent respectively. Following countries were considered as fragile: Burundi, CAR,,

8 difference remains important and quite different from the rest of the world, where resource-rich and resource-poor countries during this period grew at similar rates (around 3.2 percent per capita). The growth pattern of Africa s resource-rich countries proved also more volatile, with deeper troughs during the 1980s and early 1990s and stronger upswings over the past decade and a half. When considering the whole period , average annual GDP per capita growth rates across both groups of countries were similar (1.2 in resource-rich and 0.9 percent in resource-poor countries). Moreover, even though resource-rich countries have been growing faster on average during , solid growth performance has also been recorded among some resource-poor countries such as Ethiopia, Rwanda, and Mozambique income from mineral rents has only come on board in Mozambique from 2004 onwards (Figure 2). In those countries growth has been driven by services and agriculture. Contrast this with the growth pattern in Angola, Nigeria, Zambia, three of Africa s longstanding resource-rich and faster growing countries, where services and resource rents make up the lion s share of growth. The difference in the contribution of agriculture to growth is particularly striking (2.5 percentage points per year in the 3 fast growing resource-poor countries versus only 1 percent in the three fast growing resource-poor countries). As observed before, there is also larger volatility in the growth pattern of the resource-rich countries, mainly reflecting the volatility in resource rents. The contribution of manufacturing or other (non-mineral) related industries remained modest in both groups. Figure 2: Composition of growth in selected resource-rich and resource-poor countries in Sub-Saharan Africa Growth in % to to Resource Poor [Ethiopia, Mozambique, Rwanda] Resource Rich [Angola, Nigeria, Zambia Agriculture Resource rent Services Manufacturing & other industries Somalia, Eritrea, Togo and Zimbabwe (resource-poor); Chad, Côte D Ivoire, Democratic Republic of Congo, Liberia, Sudan, and Sierra Leone (resource-rich).

9 Under 5 Mortality (per 1000 births) Sources: Authors calculations. Human development lags in resource-rich countries A recent literature emphasizes that when it comes to human development and poverty reduction, it is not only the amount of growth which matters (Growth Is Good for the Poor, Dollar and Kraay, 2002), but also the composition of this growth, with traditionally more labor intensive sectors (such as agriculture and light manufacturing) usually more poverty reducing than capital intensive ones (such as mining). 10 This is borne out in the stark difference in human development outcomes between Africa s resource-rich and resource-poor countries, despite superior growth over the past 15 years (Figure 3). By way of illustration, while at similar levels of income per capita, inhabitants of Cameroon, live on average 7.7 years less than inhabitants of Senegal. This is largely due to much lower child mortality in the latter (68 per 1000 compared with 127 per 1000 in Cameroon). And there are more extreme examples such as between Ethiopia and Zambia where life expectancy is 59 and 49 years respectively despite de low per capita income of Ethiopia. Similarly large discrepancies (on average 6 % points across the two groups) are observed when it comes to access to sanitation, with a mixed picture in terms of primary school completion rates, where some resource-rich countries are performing well (e.g. Zambia), while primary school completion remains dismally low in others (e.g. Angola, Equatorial Guinea), despite high GNI per capita. Nonetheless, achieving better human development from economic growth remains an important challenge also in resource-poor countries. Controlling for income, many continue to lag in their human development compared to the rest of the world. Figure 3: Human development indicators in Sub-Saharan Africa and the rest of the world Life Expectancy 2011 (Years) Under 5 Child Mortality 2011 (per 1000 births) SLE ZAR MLI TCD CAF GNB AGO ZAR CPV MDG GHA STP NAM ERI COM SDN ETH TZA SEN GMB MRT LBR TGO KEN COG NER RWA CIV UGA MWI MLI NGA BDI TCD AGO MOZ CMR CAF ZMB SLE GNB LSO SWZ MUS GAB ZAF BWA SYC GNQ BFA BDI CMR NER GIN NGA CIV MOZ TGO MRT BEN UGA GMB COG MWI ZMB SDN LBR ETH GHA KEN ERI TZA MDG SEN RWA CPV SWZ NAM GAB ZAF BWA MUS GNQ SYC Log of GNI PC 2011 SSA Res. Rich SSA Res. Poor ROW SSA Res. Rich SSA Res. Poor ROW Log of GNI PC 2011 SSA Res. Rich SSA Res. Poor ROW SSA Res. Rich SSA Res. Poor ROW 10 Ravallion and Chen (2007); Loayza and Raddatz (2010); Christiaensen, Demery, and Kuhl (2011).

10 Primary completion rate (%) ZAR LBR BDI CAF SLE MOZ NER MWI COM UGA RWA ZMB GMB KEN CIV MLI ETH GNB GHA MDG TGO TCDBEN TZA Access to sanitation 2010 (%) SEN CMR NGA MRT CPV COG AGO SWZ NAM ZAF MUS BWA GAB ZAR Primary completion rate (% relevant group) LBR SLE BDI ERI TGO MWI MDG RWA GINGNB ETH MOZ MLI UGA NER CAF ZMB BEN BFA GHA TZA TCD STP CMR LSO GMB SEN CIV NGA MRT COG CPV AGO NAM SWZ MUS GNQ lngni Log of GNI PC 2011 SSA Res. Rich SSA Res. Poor ROW SSA Res. Rich SSA Res. Poor ROW SSA Res. Rich SSA Res. Poor ROW SSA Res. Rich SSA Res. Poor ROW Source: Human Development Indicators (2012); World Development Indicators (2012) Poverty declining at a slower rate in resource-rich than in resource-poor countries Turning to the evolution of income poverty, data limitations become more challenging. For example, from 1995 to 2011, 24 of the 49 countries in the region do not have at least two household expenditure surveys. Data limitations are more pronounced in resource-rich countries and especially evident in Africa s fragile states. 11,12 For that reason, the numbers reported here pertain mainly to the evolution of poverty in non-fragile states, even though survey coverage (and data quality) remains an issue there as well. 13 To begin, poverty levels ($1.25-day headcounts) tend to be lower in resource-rich than in resource-poor states. Given that the former are usually richer (higher GDP per capita), this does not surprise. At the same time however, despite on average 2.2 times faster growth, poverty declined substantially more among (nonfragile) resource-poor than in (non-fragile) resource-rich countries. In the former group, the estimated $1.25 day poverty headcount declined from about 65 percent during to an estimated Eighteen of the world s 35 fragile states (based on the FY13 Harmonized List of Fragile Situations) are in Sub-Saharan Africa. 12 That said, the frequency with which household expenditure surveys have been applied in Sub Saharan Africa is not that different from the frequency with which Demographic and Health Surveys (DHS), which form the basis for many of the fertility and health indicators, have been administered. There are on average 1.7 household expenditure surveys over the (averaged over all 49 countries, including those with no survey at all) compared with 2.2 Demographic and Health Surveys. 13 Statistical information on poverty is especially scarce in the more fragile states 9 out of 13 fragile states do not have the minimum two household expenditure surveys necessary to assess changes in poverty or inequality during the period. Sixty-one percent of the non-fragile states (22 out of 36) have at least 2 household expenditure surveys recorded in the Povcal data (read March, 20, 2013) during However, given that only six of the twelve resource-rich non-fragile states have two poverty estimates directly obtained from the surveys, caution is warranted in interpreting the results. Poverty and inequality estimates of the intermediate years necessary to obtain averages for the three periods were obtained through linear interpolations. For the years in the period lying outside the interval covered by the surveys, extrapolation was applied. To do so, a country-specific elasticity of poverty to GDP per capita was estimated from the nearest spell, which was subsequently applied to the observed GDP per capita growth rate for the respective years. Concerns about coverage, come in addition to ongoing debates about the quality of the poverty data. To be sure, collecting household expenditure data, which form the basis for poverty estimates, is generally challenging, and the challenges only multiply when comparing them over time. For a broader review of the conceptual and practical challenges in measuring poverty over time, see Christiaensen and Shorrocks (2012).

11 11 percent during In the 6 resource-rich (non-fragile) countries covered in the data, it only declined by an estimated 7 percentage points, underscoring that higher economic growth does not always automatically translate into higher poverty reduction. There are further signs of a slow-down in poverty reduction since 2008, especially in resource-rich economies; though on average poverty seems to have continued its downward trend, despite the global triple food, fuel, and financial crises of 2008/9. 14 Given the limited number of actual household survey expenditure observations during , the numbers for that period remain indicative at this stage. 15 Figure 4: Poverty and inequality in resource-rich and resource-poor countries, Source: Authors calculations based on PovcalNet (March 2013). With Gini coefficients hovering around 45 percent, inequality in Sub-Saharan Africa is high. Inequality may even have increased further over the past couple of years. 16 Somewhat surprisingly, inequality in resource-poor countries appears slightly higher than in resource-rich countries. This may partly reflect difficulties in capturing incomes of the very rich by the household survey instrument. 17 Overall, Africa s high inequality raises important questions regarding the poverty-reducing powers of its future growth, as high inequality dampens the poverty-reducing effects of economic growth (Ravallion, 2007). 14 Using self-reported food insecurity information from the Afro Barometer and the Gallup World Poll Verpoorten, Arora and Swinnen (2012) and Headey (2013) report for example a slight increase in food insecurity in Africa between and respectively. 15 The numbers are indicative only. While standard procedures are followed, they are largely informed by extrapolations, i.e. per capita GDP growth rates and past relationships in the country between GDP growth and poverty. The limited availability of more up to date poverty numbers for this period also reflects common 2-3 year lags in the administration of the expenditure surveys and the release of the official poverty numbers. 16 To put it in perspective, a Gini coefficient of 45 percent means that the difference in consumption between two randomly selected people of a country is about almost half the average consumption in that country. While higher Gini coefficients have been recorded elsewhere, especially in Latin America (e.g between 55 and 60 in Brazil), these are typically based on income measures, as opposed to consumption measures as in SSA. Given that part of income is usually redistributed (e.g. via taxation), income-based Ginis are typically higher than consumption-based Ginis. 17 Korinek, Mistiaen, Ravallion (2006). Atkinson and Lugo (2011) explore this in the context of Tanzania.

12 What can the data tell us about the future? Over the past 15 years, Africa experienced robust growth of GDP with uneven, and overall, still relatively slow progress in the reduction of its income poverty, especially among resource-rich countries. Its human development outcomes remain also well below those of the rest of the world, even after controlling for differences in income levels. To better guide Africa s poverty-reducing endeavors moving forward, the observed changes in poverty are now linked to the observed changes in income over the past 30 years to explore which factors would either improve, or deteriorate the conversion of its economic expansion into the reduction of poverty. To do so, correlations are usually estimated in a regression framework linking changes in poverty observed in the household surveys with changes in GDP per capita from the national accounts to estimate what is called the elasticity of poverty with respect to GDP, i.e. the percent change in the poverty measure given 1 percentage point growth in GDP per capita. Poverty to GDP elasticities of -2 are commonly observed, indicating that one percentage point GDP per capita growth (i.e. one percent change in per capita GDP) would reduce the $1.25 poverty headcount by 2 percent (not percentage points!). It is then examined which factors either increase or decrease this elasticity. Application of this standard statistical procedure to the Sub-Saharan Africa sample for the period, suggested however, that changes in poverty and GDP per capita (in 2005 purchasing power parity) were weakly correlated. Strong discrepancies in the economic performance recorded in the national accounts and the households surveys are not unique to Africa they have also been documented elsewhere, most recently for India (Datt and Ravallion, 2011). There are also good conceptual reasons why there should not be a one-to-one correlation between the evolutions of both. Poverty concerns private consumption/income, while GDP comprises private consumption as well as government consumption, private and public expenditures, and exports and imports. Indeed, in SSA, weak correlation was found between changes in average private consumption from the household surveys and changes in GDP per capita from the national accounts. This contrasts with the experience in the rest of the world, where a percentage point increase in GDP from the national accounts yielded on average a one percent increase in average private consumption. 18 And the elasticity of poverty to GDP was also statistically significant in the rest of the world (estimated at -2 for the period). 19 Both statistical as well as structural issues underpin this discrepancy between growth in GDP per capita observed in the national accounts, growth in consumption per capita recorded in the household surveys, and changes in poverty. The statistical foundations for both GDP per capita and poverty estimates in Sub-Saharan Africa remain wanting. GDP accounts often use old methods (Jerven, 2012), more often than not, population censuses are out of date, and poverty estimates are regularly not comparable over time due to changes in survey 18 This is after dropping extreme data points using the BACON procedure (Weber, 2010). Without dropping these points, the coefficient was It was not even close to statistically significant for the Sub-Saharan sample recorded in Povcalnet (as read on March 2013). 19 This is after controlling for country fixed effects and omitting outliers using the BACON procedure.

13 design and inadequate price deflators (Christiaensen, Lanjouw, Luoto and Stifel, 2012). Proximate reasons for this are weak capacity, inadequate funding and lack of coordination of statistical activities (Devarajan, 2013). Deeper reasons may relate to the political sensitivity of statistics and donors tendency to go around countries own National Statistical Development Strategies. In this view, lower household coverage in resource-rich countries may also not be a surprise. The continuing poor quality of Africa s statistics is in need of urgent attention. Cognizant of the limitations of the data, the paper moves beyond the national accounts to assess Africa s prospects for further poverty reduction and directly explores the link between the evolution of poverty and average consumption growth from the household surveys. 20 The data are then internally consistent. The experience from the rest of the world is also drawn in. Measures there are more robust. They also increase the sample size and provide a point of comparison. Statistical artifacts, initial inequality and resource endowments Survey data show a significant link between the change in per capita consumption: a one percent growth in consumption per capita is estimated to reduce poverty by 0.69 percent. This is still substantially less than in the rest of the world, where the elasticity of poverty to consumption is estimated at (Figure 6). Three factors underpin this difference. First, given that poverty levels in Sub-Saharan Africa are higher and incomes lower, equivalent absolute changes in poverty and incomes translate to smaller and larger relative changes respectively, which arithmetically reduce the elasticity of poverty to consumption in Sub-Saharan Africa.. Figure 6: Elasticity of poverty with respect to consumption Poverty to consumption elasticity Poverty to consumption elasticity No controls Sub-Saharan Africa Rest of World With controls Resource Rich Resource Poor Note: Controls include initial consumption, inequality and natural resource share (>5 % of GDP) indicator. All results control for country fixed effects. Full regression results are in appendix 20 A similar procedure was followed in India by Datt and Ravallion (2011).

14 14 Second, and for a structural reason, high initial inequality has been shown to lower the poverty-reducing effects of economic growth (Ravallion, 2007) and as indicated above, initial inequality is on average already high in Sub-Saharan Africa. Finally, beyond the growth rate, the sources of growth also matter for poverty reduction, with growth in labor intensive sectors such as agriculture and manufacturing typically more poverty reducing than growth in capital intensive sectors such as mineral exploitation (Loayza and Raddatz, 2010; Christiaensen, Demery, Kuhl, 2011). Once these three factors are controlled for, the elasticity in Sub-Saharan Africa approaches that of the rest of the world: -3.1 versus -3.8 (Figure 6). The elasticity of poverty to consumption in the region rises from being only 33 (columns 1 and 2) to being about 80 percent (columns 3 and 4) of that of the rest of the world, when controlling for the consumption level, inequality and resource dependence at the beginning of the spell (Table A1). Together they, thus, explain most of the difference. As seen from the positive coefficient on the interaction term between consumption growth and initial inequality, higher inequality is found to dampen the poverty-reducing effect of growth (consistent with the literature). The results further confirm that the effect of growth on poverty is lower when initial inequality at the beginning of the spell are higher (Table A1). Somewhat surprisingly, the sign on the interaction term of growth with resource richness is negative, suggesting that growth is more poverty reducing when resources are at least 5 percent of GDP at the beginning of the growth spell. As changes in inequality are already controlled for, it concerns here changes beyond the inequality channel. However, as can be seen from the indicator on the resource dependence indicator, there is also a strong positive level effect, i.e. poverty reductions during spells with higher initial resource dependence are on average 10 percent less. No effects were found in SSA, possibly also related to the limited number of observations with higher resource dependence. To explore the effect of resource dependence further, the sample was pooled across all regions and split by resource dependence (columns 5 and 6). The larger elasticity of poverty to consumption in resourcepoor countries (by -0.7) confirms the lower poverty-reducing impact of economic growth in countries that are (or become) more resource dependent. Finally, as can be seen from the positive coefficient on changes in the Gini coefficient, growth processes that come along with increases in inequality are less poverty reducing. Signs are that in Sub-Saharan Africa, growth in resource dependence is also associated with increasing inequality. Controlling for initial income and inequality levels, inequality (as measured by the Gini) goes up as the share of mineral rents in the economy increases (Table A2). No similar patterns of rising inequality with rising resource dependence were observed in the rest of the world. This counsels further caution about the expected effects of mineral driven growth on poverty. Not only has growth in more mineral dependent economies less effect on poverty, partly through the inequality channel and partly through other channels, by increasing inequality it also undermines the poverty-reducing effect of future growth. Finally, resource wealth may also undermine the expected positive effects of Africa s upcoming demographic transition on future income growth, by reducing the capacity of the economy to

15 15 productively absorb increases in the work force. The possibly limiting effects of resource wealth on economic growth, through expansion of the non-traded sector at the expense of the traded sector (the so-called Dutch disease), has been widely documented. If the traded sector is more dynamic, this change in sectoral composition may be detrimental to growth. Expansion of the workforce in such a context may further test the employment generating capacity of the resource-rich countries. Recent work by Bjorvatn and Farzanegan (2013) provides empirical support for the thesis that higher natural resource rents limit the beneficial effects on economic growth from lower dependency ratios and the demographic transition. These results, drawn from the cross-country experience, are furthermore not driven by possible differences in institutional quality between resource-rich and resource-poor countries. They hold in both low and high quality institutional settings, and also beyond the oil-rich economies of the Middle Eastern and North African countries. 3 Making growth more poverty reducing 3.1 Managing minerals better the potential of direct transfers Continued demand for Africa s natural resources as well as the recent discoveries of oil, gas and minerals in, among others, Ghana, Kenya, Mozambique, Tanzania and Uganda together with an improved macro-economic environment and better governance sustain prospects for robust economic growth (Africa s Pulse. October 2012). As highlighted by Lederman and Maloney (2009) the empirical data do not support the stylized fact of slow growth in resource-rich countries. Rather, the evidence points to heterogeneity, with some performing poorly compared to resource-poor countries, while others do not. Strong growth remains the first driver of rapid poverty reduction. The more pertinent question is how more of the new-found resource wealth can be converted into fiscal revenues and effective public spending to foster sustainable development, improve human welfare outcomes, and generate more rapid income poverty reduction, so as to avoid another resource curse. To be sure, not developing the resources is not an option. Governments will want to exploit their natural resource base to build their economies and improve the lives of their people, and growing emerging market countries will fuel demand for these resources. But, could things be done differently so as to make faster progress on reducing poverty? Increasingly the debate has turned to how institutions and natural resources interact and how institutions and governance can explain policy failure. A useful framework for better understanding the governance challenges in converting resource wealth in human development is the value chain approach described in Alba (2009) and Barma et al. (2012). Three core components of natural resource management, each embodying their own political dynamics, are highlighted: 1) extraction transparency regarding terms of contracts; 2) taxation efficiency in tax collection; and 3) investment of resource rents careful prioritization of public investment. Some recent initiatives have focused on the first dimension. In particular, under the Publish What You Pay and the Extractive Industries Transparency Initiatives (EITI) mining companies and governments disclose mandatorily or voluntarily

16 16 what they pay and what they earn. Nineteen African countries are now part of the EITI, of which eight are compliant with all requirements. In addition to initiatives to increase accountability of international firms, promising and innovative mechanisms to increase internal, bottom-up accountability and also address components two and three are being proposed, inspired by core insights from behavioral economics such as emotional accounting. One proposal that especially deserves attention calls for directly distributing a small share of the resource wealth (e.g. 10 percent) universally and uniformly to the citizens (Moss, 2011; Devarajan and Giugale, 2013). A key feature distinguishing mineral rents from other forms of fiscal revenues, is that they go directly from the extracting company often a multinational to the government. As a result, citizens often have a hard time to know how much natural rent money should be reaching the public coffers, let alone, how it has been spent. And as it does not immediately concern their hard earned tax dollars, citizens typically also have less incentive to find out, further reducing scrutiny of public spending. This adds to innate human nature to spend easily obtained income less diligently (Christiaensen and Pan, 2012), fostering inefficiency of public spending, or worse, corruption. To break this cycle, some studies propose to transfer a portion of the resource revenues directly to the citizens. Specifically, it is proposed to transfer a fixed proportion of these revenues to everyone (universally) and in the same amount (uniformly), with all three features of essential importance. It then becomes in the citizens immediate interest to scrutinize the amount of mineral wealth received by the government misreporting and leakages would reduce the amount they receive and empowers them directly to overcome financial constraints in expanding their human and physical investments. Theoretical analysis shows that it may even induce governments to become more effective in the provision of public goods as the revenues under their direct control decline. This could in turn even yield an increase in public goods in countries that are already resource-rich and which experience decreasing marginal returns in the production of their public goods, provided the transfer is large enough for citizens to care and the political landscape sufficiently contestable that policymakers take citizen welfare into account. While seemingly revolutionary, and definitely new to Africa, such schemes are already current practice in the U.S. state of Alaska and the Canadian province of Alberta. Clearly, the immediate poverty effects from the transfer alone could be substantial. To fix ideas and illustrate potential orders of magnitude, simple simulations that uniformly and universally distribute ten percent of the expected resource rents across the population suggest that the poverty gap, i.e. the distance at which the poor find themselves under the (national) poverty line, could be more than eliminated in some of the smaller traditionally resource-rich countries (such as Equatorial Guinea and Gabon), or substantially reduced in the larger ones (Angola and Nigeria), with also sizeable reductions in countries with emerging mineral riches (Table 1).

17 17 Table 1: Poverty reduction from uniform and universal direct distribution of 10 percent of resource rents Estimated annual natural resource fiscal revenue (US$ billion) % reduction of poverty gap among the poor following distribution of 10% of natural resource fiscal revenue Equatorial Guinea Gabon Angola Republic of Congo Nigeria Mozambique Uganda Tanzania Note: Calculations are based on national poverty lines. Source: Devarajan and Giugale (2013). The feasibility of such schemes obviously depends on many factors, and many hurdles will have to be overcome. An often heard objection holds that identifying people and transferring money is technically difficult and costly. Over the past years, however, governments have gathered a lot of experience in distributing transfers to large parts of their populations through their social assistance programs, as well as through the seed and fertilizer voucher programs. As the proposed resource transfer is universal and uniform, distributing it would in effect be much easier than the vouchers which are also targeted. Technological advances such as biometric identification (as in India) and phone banking are cutting costs dramatically. While it will take some time for these techniques to become widely available, as soon as they take root, they can spread quickly, as observed in Kenya with the rapid adoption of the mobile money transfer service M-Pesa. On the political front, incumbent rulers, parties or interest groups would have little incentives to give up control over natural resources. But with political competition in Africa expanding, political parties in opposition might find the direct resource transfer powerful to gather political support. Moreover, by making the transfers uniform and universal, nobody will be excluded and everyone will get their fair share. Ethnic, religious or tribal tensions could thus also be relieved and national unity strengthened. In terms of economic efficiency, many governments may need any marginal revenue to supply the many unmet public goods (infrastructure, vaccination programs, primary schooling) before engaging in private transfers. Yet, implicit or explicit private transfers in resource-rich countries are already reality today. For example, over the past year, the government of Zambia spent over 2 percent of its national GDP in supporting maize production through maize purchases at above-market price and subsidizing inputs (over 60 percent of the Ministry of Agriculture s public budget). 21 And subsidizing fuel has been common in oil rich countries like Nigeria as well, where it amounted to 30 percent of the Nigerian government s expenditure or about 4 percent of GDP in 2011 (Moyo and Songwe, 2012). By way of comparison, in 21 Jayne et al., (2011).

18 18 Nigeria, the 10 percent transfer simulated above, would only amount to 2.9 percent of GDP (based on 2010 GDP numbers). Importantly, while poor and rich would equally benefit from the direct dividend transfer, fertilizer and fuel subsidies are regressive, going disproportionally to the richer segments of the population. Fuel consumption per capita is substantially less for the poor, who use primarily public transport, than for the rich. And while the smaller farming households ( ha) received on average 24.1 kg of subsidized fertilizer during the 2010/11 cropping season in Zambia, the largest farmers (10-20 ha) received kg (Mason et al., 2011). Clearly substantial amounts of government revenues, implicitly facilitated by resource rents, are already financing private goods. Direct dividend transfer could do so in more equitable, efficient and welfare enhancing ways, while at the same time possibly also increasing the efficiency of public good provision through increased citizen scrutiny. By fixing the dividend as a proportion of total revenues, the effects of fluctuations in international commodity prices on fiscal revenues are also mitigated. That leaves the moral issue of whether citizens can be trusted to spend resource dividends wisely. For the same reason that governments may apply less scrutiny to spending from resource windfalls, so may citizens spend their transfers more freely on non-basic goods (including non-staples foods), as for example demonstrated in Tanzania by Christiaensen and Pan (2012). While real, the differences remain relatively small in practice. Campaigns to raise awareness may help beneficiaries to better manage the use of these transfers, but more research is needed to demonstrate the effectiveness of such measures. 22 Transfers to the population at large have the added benefit of increasing the demand for locally produced goods and services as opposed to imports, implying that the local multiplier effects can also be large. In sum, direct dividend transfer programs prove promising as an additional instrument to reduce inequality and increase the poverty-reducing powers of economic growth in resource-rich countries. They reduce pressures to nationalize resources, which often weaken incentives for private investment and productive efficiency, while strengthening the accountability links between the state and its citizens. Just like Mexico pioneered conditional cash transfer programs through the introduction of Progressa in 1997, which has been widely replicated across the world, the time is now ripe for an African champion to step forward and lead the way in pioneering the direct transfer programs. 3.2 Advancing Agriculture The Case for Staples A second opportunity to enhance the poverty-reducing powers of Africa s future growth lies in agriculture. World food prices are elevated and expected to stay so in the medium term, and with urban food markets set to quadruple over the next two decades, domestic and regional markets offer attractive opportunities for Africa s producers. Agriculture and agribusiness together are projected to be 22 See World Bank (2013d), Solomon et al. (2012), and Covarrubias et al. (2012).

19 19 a US$ 1 trillion industry in Sub-Saharan Africa by 2030, up from US$ 313 in 2010 (World Bank, 2013a). And last, but not least, growth coming from agriculture has on average been proven to be more poverty reducing than growth coming from other sectors (Diao, Hazell and Thurlow, 2010; Christiaensen, Demery, Kuhl, 2011). But, many of the opportunities have not yet been captured. In the mid-2000s, Africa converted from being a net exporter of agricultural products to becoming a net importer, with many of the mineral dependent economies 23 now being large net importers. Much of the growth in imports concerns staples- -especially, rice, but also wheat and sugar--for its rapidly expanding urban populations, as well as milk products and poultry, whose imports have exceeded US$ 2 billion in recent years. Except for wheat, which is a temperate-zone crop, these are all products in which Africa should have a comparative advantage, given its cheap labor and abundant land. In addition, just as not all nonagricultural growth is equally poverty reducing (natural resource exploitation being just one example), neither is all agricultural growth. Its success in reducing poverty differs across its subsectors as well as the modalities and agrarian structure. Agriculture s GDP growth rate in Brazil, for example, has systematically exceeded that of the rest of the economy over the past 15 years. But given the large scale capital intensive nature of this expansion, with limited use of mainly higher skilled labor, it has not been a major contributor to poverty reduction. Brazil s poverty reduction has been mainly driven by transfers and the generation of rural nonfarm employment (World Bank, 2007; Ferreira, Leite, Ravallion, 2010). Special attention is needed to increase Africa s productivity in staple crop production. While agriculture s growth has picked up during the 1990s and 2000s, this remains largely driven by unsustainable area expansion, which explains two-thirds of the growth in agricultural output, with TFP growth and increased input use accounting for the remainder (Fuglie, 2011). Staple crop yields remain way below potential, with maize yields reaching only 20 percent of their (experimental station) potential and cash crops reaching percent (World Bank, 2007). But more progress appears on the way in some countries. A case in point is Rwanda, where over the past 5 years ( ) cereal yields and the yields of roots/tubers increased by 73 and 52 percent, respectively while the poverty headcount dropped by 12 percentage points (World Bank, 2013b). Recent country specific, economy-wide, recursive dynamic computable general equilibrium simulations in a number of African countries of TFP growth in different crops and agricultural subsectors by Diao et al. (2012) confirm that increasing smallholder staple crop productivity (as opposed to export crops) generates the largest poverty reduction. While export crops typically have higher value and growth potential than food crops, food crops are usually more effective at generating economy-wide growth and reducing national poverty. This follows from their larger multiplier effects and their larger elasticity of poverty to GDP one percent growth in agriculture driven by cereal or root/tuber productivity growth generates a larger decline in national poverty than a one percent growth in agriculture driven by productivity growth in export crops (see Table 2). When smallholders are engaged in growing the export crop, the gaps are usually smaller (such as cotton exports in Zambia and tobacco in Malawi). The results 23 African countries standing out for their strong agricultural export orientation include Côte d Ivoire and Kenya, with Ethiopia, Ghana and Zambia recent African successes in terms of significant increases of their export shares, albeit from a low base.

20 20 also hold in resource-rich countries such as Zambia and Nigeria, underscoring agriculture as another important and often neglected avenue to increase the elasticity of poverty to GDP in resource-rich countries. In sum, while agriculture growth is generally pro-poor, the pro-poorness of the different subsectors should also be at the top of the agenda in developing agricultural strategies. Attention to growing smallholder staple crop productivity, despite the obvious political appeal of the fast growing agricultural niche markets (such as those of export-oriented horticultural products), is especially warranted. Table 2: Multiplier effects and elasticity of poverty to GDP for staple foods versus export crops Staple crops Growth multiplier Elasticity of Poverty to GDP Export crops Growth multiplier Elasticity of Poverty to GDP Ethiopia all cereals all export crops Kenya all cereals all export crops all food crops Malawi maize Tobacco Horticulture Other export crops Mozambique Maize Traditional exports All cereals Biofuel crops Nigeria Maize 1.28 All export crops Pulses and oilseeds All cereals Roots Rwanda Maize Coffee Pulses Tea Other export crops Tanzania sorghum and All export crops millet Livestock Maize Uganda Roots All export crops Horticulture Zambia All cereals All export crops Roots Source: Diao, et al., 2012 Different countries are pursuing different models to increase staple crop productivity, with varying degrees of success. The experience of Zambia and Rwanda are particularly noteworthy. To begin, both Zambia and Rwanda report a doubling of their maize and cereal output between 2006 and 2011 (Mason et al., 2011; World Bank, 2013b), with more than half of the increase coming from yield increases. The models followed to reach these outcomes were, however, starkly different, including in their effects on

21 21 poverty remaining virtually stagnant in Zambia and declining rapidly in Rwanda. Zambia emphasized subsidizing inputs to farmers and purchasing maize at above-market prices, with the bulk of the inputs and benefits going to a smaller group of larger farmers who also produced the bulk of the marketed surplus. The 42 percent of households cultivating less than one hectare of land, who are also among the poorest of the poor, produced only 7 percent of the expansion in production. Also, while maize output per agricultural household virtually doubled, the mean household net value of total crop production (maize and non-maize) increased by only 20 percent due to substitution away from other high value crops and higher input costs. The Rwanda Crop Intensification Program (CIP), wherein subsistence farmers, who traditionally grow an array of crops on very small fields (on average less than 0.3 ha), are invited to pool their land and specialize in one crop depending on the agro-ecological environment, has been the workhorse of the new agricultural strategy of the government. In addition, these farmers get specialized extension services and are provided with fertilizer (initially at no cost; after the first harvest they buy at full price). Decompositions show that almost half (43 percent) of the reduction in poverty in Rwanda between 2001 and most of which happened after the adoption of the CIP in has been accounted for by developments in agricultural production (33 percent) and increased marketing of harvests (10 percent). Non-farm self-employment also played an important role (13 percent of total poverty reduction), as did the decline in the dependency ratio (9%). Non-farm wage employment only accounted for 3 percent of the poverty reduction (see figure 7). Figure 7: Contribution of agriculture to poverty reduction in Rwanda, Other factors and unexplained part, 30% Increased agricultural production 35% non-farm wage employment, 3% Non-farm self employment, 13% Decreased dependency ratio, 9% Increased agricultural commercialisati on 10% Source: World Bank (2013b).

22 22 Yet, no dominant agricultural success model has emerged so far, and adaptation to local circumstances remains key. Rwanda s extreme high population density (416 persons per squared kilometer), for example, is quite peculiar. At the same time, the differential experiences of Zambia and Rwanda are illustrative in highlighting the importance of the right mix of rural public (extension services, coordination) and private good provision by the state. Finally, maximizing poverty reduction in meeting Africa s new urban demands beyond staples, such as the demand for dairy, meat, vegetables and other high value agricultural products,, is also not automatic. Large agribusiness investment should not be assumed to have positive development impacts. The employment generation and poverty-reducing effects will depend in large part on the design of the value chain. In particular, the integration of market-oriented smallholders and rural communities in dynamic value chains through contract farming and strong producer organizations that facilitate aggregation can go a long way in generating employment and reducing poverty. Governments can help overcome coordination costs and strengthen bargaining positions of smallholders. Large scale operations, necessary at times to overcome fixed costs and meet export standards, can also generate substantial benefits for local communities and their poorer segments provided they generate good jobs that also respect social and safety standards, as illustrated in Senegal (Maertens, Colen and Swinnen, 2011). 3.3 Ushering urbanization the role of secondary towns Next to new mineral riches and untapped agricultural potential, Africa s youth bulge and ensuing demographic dividend provides a third opportunity to convert its growth potential into more poverty reduction. After many years of rapid population growth, fueled by a decline in child mortality, fertility has also started to decline, resulting in a falling dependency ratio, which stands at 84% in 2011, compared with 94% at its peak in As Africa s youth bulge is about to enter the labor force, Africa is poised to capture a demographic dividend, which has been estimated to account for about a third of the rapid growth between 1960 and 1990 among East Asia s early growers. But again, productively absorbing the youth bulge into the labor force is far from automatic.

23 23 First, and taking one step back, while fertility has come down substantially in some countries (such as Botswana and South Africa), it has essentially not yet started in others (Niger and Uganda), and it appears to have had mized progress in yet others (Tanzania and Kenya). How to continue and accelerate the fertility transition remains an important policy challenge. Without exception, the most important proximate determinant of fertility remains girls education. Africa s progress in closing the gender gap in primary school enrollment provides hope in this regard. Schooling postpones age at marriage and increases the use of modern contraceptives. On the supply side, family planning programs are often the intervention of choice to reduce fertility, with several studies suggesting that exposure to such programs could reduce completed fertility by up to one child. 24 There are further indications that family planning programs may act as a substitute for formal education. This is especially important, given that they could have immediate effects among women currently of child-bearing age, many of whom have only enjoyed limited formal schooling. They would thus form a timely, complementary intervention, while efforts continue in closing the gender gap in primary and secondary schooling and in reducing child mortality. Second, when it comes to employment, the primary challenge is not unemployment per se, but rather to increase productivity in the informal sector. Formal unemployment rates are only 3 percent in Sub- Saharan Africa s low-income countries and, with the exception of South Africa, 8 percent in lowermiddle-income countries. Meanwhile, the vast majority of the population in low-income countries continues to be employed in the informal sector, both in agriculture (70%) and informal household enterprises (18 percent), with only 5 percent of the population engaged in formal wage employment. In lower-middle-income countries, the shares are 54, 21 and 15 percent, respectively. The very low share of wage employment in total employment dictates that even under optimistic projections of growth in wage jobs, informal is normal will remain so in the years to come as illustrated for Uganda (Figure 8). The recent experience of Rwanda highlighted above points in the same direction. Figure 8: Employment distribution in Uganda 24 Angeles et al. (1998, 2005), Miller (2010), Pörtner, Beegle, and Christiaensen (2012).

24 Density 24 Source: Fox, Louise and Thomas Sohnesen, Note: Estimates using Uganda UNHS 1992/93 and 2005/06. Third, along with the generation of off-farm jobs needed to employ Africa s youth bulge comes a spatial transformation, with people moving out of agriculture into urban settings. Even though urbanization rates remain still well below those observed in the rest of the world, urbanization in Africa has accelerated over the past couple of decades, with much of the new urban population concentrating in the larger cities. Using census data from 42 Sub-Saharan countries 25, Dorosh and Thurlow (2013) find that in 2010 two-fifths of Africa s urban population already lived in big cities (i.e. one million or more), while two-fifths lived in small towns (less than 250,000 people). Moreover, while urbanization expands throughout the distribution, urban growth is much faster in the big cities (6.5 percent per year during versus 2.4 percent in the smaller towns) (Figure 9). Figure 9: Distribution of Sub-Saharan Africa s urban population ,000 10, ,000 1,000,000 10,000,000 Size of urban center (people, log scale) 25 Angola, Comoros, Rwanda, Seychelles and South Africa were not included.

25 25 Note: Probability distribution function using population census data for Sub-Saharan African countries, excluding Angola, Comoros, Rwanda, Seychelles and South Africa. Source: Dorosh and Thurlow (2013). Just as with agricultural growth processes, not all processes of urbanization are equally poverty reducing. Thus far, the debate about the effects of urbanization in development has mainly been between proponents, emphasizing the benefits for economic growth from economies of scale and agglomeration economies, and opponents, underscoring congestion effects and slum formation and seeing these effects primarily as potential new sources of poverty.. It is, however, crucial to move beyond this rural-urban dichotomy to maximize the effects of urbanization on poverty. Indeed, there are increasing indications that the nature of urbanization (secondary town development versus metropolitization) may be more important for poverty and inequality reduction than urbanization itself. There are at least three channels through which these different urbanization patterns may lead to differential poverty outcomes. First, as emphasized in the new economic geography literature, urban concentration or metropolitization may generate faster economic growth and more jobs given larger economies of scale and agglomeration than in secondary towns. Second, the magnitude of the positive spillover effects (for example through remittances and rural nonfarm employment generation) on rural poverty in the hinterlands of metropoles may be larger, though the space and number of people affected may be smaller than those affected by secondary towns. Finally, the poor may find it easier to migrate to and find jobs in secondary towns in their proximity than in distant cities. Lower migration costs, the ability to maintain closer social ties with the areas of origin, and possibly also the higher chance of finding a job commensurate with skill level might all lead the poor to favor migration to nearby towns to find off-farm employment and exit poverty. There is emerging evidence supporting the view that migration out of agriculture into the rural nonfarm economy and secondary towns is conducive to faster poverty reduction. The findings from a unique, representative survey of rural Kagera, a region in northwestern Tanzania, spanning almost two decades (1991/4-2010) illustrate the point most strikingly. Tracking the welfare, occupation and location of 3301 individuals, 82 percent of them in agriculture when first surveyed during 1991/4, researchers found that poverty declined by 28 percentage points, from 58 percent to 30 percent in More strikingly, close to half of the poverty reduction realized in the sample came from farmers moving out of agriculture into rural nonfarm activities and secondary towns, while 32 percent (304 out of 945) comes from farmers who remained in farming. Only 12 percent comes from farmers moving to cities (Figure 10). To be sure, incomes among those moving to the city (Dar es Salaam, Mwanza or Kampala) grew on average much faster than among those moving to the middle (by 233 percent versus 134 percent, respectively). However, the contribution to poverty reduction was much less, because many more found their way to the neighboring towns, where unemployment rates were also slightly lower.

26 26 Figure 10: Contribution of farm migration to poverty reduction, Kagera (Tanzania), 1991/ Source: Christiaensen, De Weerdt, Todo, Looking at the broader historical experience across the world, a similar pattern of faster poverty reduction from rural nonfarm diversification and secondary town development than from metropolitization is observed. Overall, concentration of the urban population in fewer and larger centers (metropolitization) appears associated with faster economic growth (as predicted by the new economic geography), but also seems to come with larger inequality, eroding some of its povertyreducing powers. Rural diversification and secondary town development, on the other hand, are associated with more inclusive growth patterns and faster poverty reduction. While further investigation is needed to better understand the mechanisms, it appears that the ability of the rural poor to connect with growth in nearby towns is important. In sum, Africa s youth bulge provides a unique opportunity for a demographic dividend provided jobs can be generated to productively absorb this new workforce. Most of these jobs will be in the informal sector, requiring sufficient attention to education and skill development, access to credit, and land tenure security to enable farm consolidation. Moreover, where these jobs will be located will be equally important to reduce poverty and foster shared prosperity, calling especially attention to the spatial prioritization infrastructure development across different urban settings. 4 Concluding Remarks Reviewing the record on Africa s human development and poverty reduction over the past fifteen years shows that progress has been made, but that it is uneven and that converting resource wealth into human welfare proves to be particularly challenging. While indications from Africa s household