Attachments
By Luis G. Becerra-Valbuena, Benjamin Davis, Ana Paula de la O Campos, Nicholas Sitko and Stefania Veljanoska
Introduction
The last 20 years have seen significant reductions in poverty in sub-Saharan Africa (SSA) as a whole. However, this progress has been highly uneven within and between the various countries of the region (World Bank, 2022). And while the data for monitoring poverty levels in SSA have improved in recent years, there remains a lack of spatial granularity in poverty estimates. This lack can often obscure important geographical disparities in poverty reduction, as generated by the various processes of economic development. Spatially explicit understandings of poverty and welfare dynamics are critical, given the importance of location in shaping household livelihood strategies – including agricultural potential and access to markets. While this is relevant for all regions of the world, it is particularly relevant for SSA, given its largely agrarian economies and the rapid growth of its urban agglomerations (Pesche, Losch and Imbernon, 2016).
The dynamic relationship between physical environment and welfare is widely recognized in the literature (Bigman and Fofack, 2000a; Dixon et al., eds., 2019; Giller et al., 2021; Nguyen and Dizon, 2017; Zhou and Liu, 2022). In particular, the literature on economic geography (Fujita, Krugman and Venables, 1999; Fujita and Thisse, 2002; Lall, Henderson and Venables, 2017; Puga, 1999) suggests that as economies grow, economic development tends to cluster in places that are more favourable for economic activity, for example because such places are endowed with greater natural resources, more suitable agroecological conditions and better market access. As these places experience income growth, they pull in new economic migrants, leading to rapid population growth and synergies created by the economies of agglomeration. The higher concentration of people and economic activities leads to economies of scale, which further sustain and reinforce growth cycles (Fujita and Thisse, 2002; Lall, Henderson and Venables, 2017; Nguyen and Dizon, 2017). At the same time, inequality is created between regions that are more and less endowed (Fujita and Thisse, 2002), with the latter facing difficulties in catching up. This suggests that geographical disparities between places are a fundamental element of the economic development process, and must be addressed through geographically targeted policy actions.
Welfare indicators that are aggregated at the national level mask existing spatial variabilities, giving a false sense of homogeneity within a country and within lower administrative levels (Henninger and Snel, 2002). In fact, poverty varies not only within countries, but also throughout other territorial typologies both within and across country borders. Unfortunately there is a lack of comparable survey data across both locations and time; this has limited analysis of the spatial trajectories of poverty and wealth accumulation (Liu, Liu and Zhou, 2017), particularly for underrepresented and isolated communities (Janz et al., 2023). Despite the significant progress that has been made by national statistical offices to enhance the quality of data collection and survey design, to date, multicountry analysis of African poverty has been hampered by differences in data collection tools, a lack of unified indices, issues in terms of quality, and problems in the calculation of income and expenditure aggregates (Beegle et al., 2016). Cross-country and continent-wide analysis has also been limited to measuring a single welfare indicator – i.e. usually consumption-based data gathered from individual country household surveys, adjusted by purchasing power parity (PPP) (Azzarri and Signorelli, 2020; Chen and Ravallion, 2010; Ferreira et al., 2016; Jolliffe et al., 2022), or deflators computed at irregular intervals of time (Christiaensen et al., 2012).
In this study, we leverage a new dataset of high-resolution welfare indicators, as developed by the technology company, Atlas AI, to examine the spatial distribution and temporal dynamics of asset wealth, per capita expenditures, and poverty levels in SSA over the period from 2003 to 2021. We ask how welfare dynamics in predominantly agrarian societies are affected by the different geographical factors with which they are often associated – namely degrees of urbanization, global agroecological zones and dominant farming systems. These geographical factors, analysed as typologies, reflect a location’s potential to access markets and populated areas, as well as its potential for diversification and growth in agricultural productivity. In providing some answers, we build on the work of past studies (Ratledge et al., 2022; Yeh et al., 2020; Bigman and Fofack, 2000b, 2000a; Giller et al., 2021; Hengsdijk et al., 2014), to analyse the spatial distribution and temporal dynamics of three specific and related welfare indicators – asset wealth, poverty, and per capita expenditures – across different territorial typologies in SSA, from 2003 to 2021. We also address gaps in several previous studies by examining welfare dynamics with a high level of spatial granularity over the last two decades.
Consistent with national World Bank estimates, we find general improvements in all three welfare indicators between 2003 and 2021. However, our findings further demonstrate that most of the progress was concentrated in the more urban areas, and among populations that were already at the top of the wealth distribution in 2003. Moreover, rapid welfare improvements in urban areas and in places with populations at the top of the wealth distribution coincided with sharp increases in welfare inequality. Outside of these more economically dynamic areas, our findings show that welfare progress over the last two decades has been limited. This is particularly the case for places that were in the middle or bottom quintiles for expenditure distribution in 2003, the baseline year for the datasets. Many of these poor-performing areas are located in desert and arid climate zones, and in rural agroecological zones that are classified as tropical lowlands – where nearly three quarters of SSA’s rural population live. Conversely, welfare progress has been significant in areas characterized by high-value, commercially oriented farming systems (e.g. fish-based and humid lowland tree-crop systems).
The analysis paints a worrying picture of the spatially uneven progress being made in improving welfare in SSA over the last 20 years, and makes a strong case for incorporating territorial approaches – along with better spatial targeting of poverty reduction actions – to address the unique needs and challenges of SSA’s diverse geographies. Improving our understanding of the spatial features that influence poverty dynamics and distributions over time can help improve poverty mapping for better, more targeted poverty-reduction interventions in rural areas (FAO, 2021). Similarly, combining geographical targeting with common targeting approaches that are based on socioeconomic analysis can lead to more effective interventions that are better tailored to the geographical context and welfare of the population.
The rest of this paper is organized as follows: The next section presents a literature review and conceptual framework of the different welfare indicators and spatial typologies considered in this study, as well as differences by territorial typology. The section on Methods and data describes the three welfare indicators generated by Atlas AI data, the different territorial typologies over which welfare dynamics are then explored, and the methods used. The final section presents the main Results, along with a discussion of key findings. The paper also includes an Annexes, which provides more detail on the data and some of the main findings.