Measuring Poverty and Wellbeing in Developing Countries
Measuring Poverty and Wellbeing in Developing Countries
Motivation and Overview
Abstract and Keywords
Detailed analyses of poverty and wellbeing in developing countries, based on large-scale, nationally representative household surveys, have been ongoing for more than three decades. The large majority of developing countries now conduct on a regular basis a variety of household surveys—income, consumption, health, demographics, labour force, household enterprise, and others. And the information base in developing countries with respect to poverty and wellbeing has improved dramatically. Nevertheless, appropriate measurement of poverty remains complex and controversial; this chapter lays out for the reader the issues and challenges. This is particularly true in developing countries where (i) the stakes with respect to poverty reduction are high; (ii) the determinants of living standards are often volatile; and (iii) related information bases, while much improved, are often characterized by significant non-sample error.
Detailed analyses of poverty and wellbeing in developing countries, based on large-scale, nationally representative household surveys, have been ongoing for more than three decades. The large majority of developing countries now conduct on a regular basis a variety of household surveys—income, consumption, health, demographics, labour force, household enterprise, and others. And the information base in developing countries with respect to poverty and wellbeing has improved dramatically. Nevertheless, appropriate measurement of poverty remains complex and controversial (Ravallion 2016). This is particularly true in developing countries where (i) the stakes with respect to poverty reduction are high; (ii) the determinants of living standards are often volatile; and (iii) related information bases, while much improved, are often characterized by significant non-sample error.
It also remains, to a surprisingly high degree, an activity undertaken by technical assistance personnel and consultants based in developed countries. This book seeks to enhance the transparency, replicability, and comparability of existing practice. In so doing, it also aims to significantly lower the barriers to entry to the conduct of rigorous poverty measurement and increase the participation of analysts from developing countries in their own poverty assessment.
The book focuses on two domains: the measurement of absolute consumption poverty and a specific approach to multidimensional analysis of binary (p.4) poverty indicators. In choosing these two areas of focus, the intent is not to give the impression that these two domains alone are sufficient for rigorous poverty assessment. On the contrary, we highlight that this book is designed to serve as a companion to the recently published volume entitled Growth and Poverty in Sub-Saharan Africa (Arndt, McKay, and Tarp 2016). That volume emphasizes repeatedly the desirability of the application of multiple approaches across multiple datasets, combined with a concerted effort to triangulate results, in order to develop a reasonably complete and coherent picture of living standards and their evolution as one moves across space or through time.
1.2 Facilitating Rigorous Measurement
While a comprehensive assessment of living conditions requires a multi-pronged approach, solid work within each prong encounters a multiplicity of challenges and choices. This is particularly true with respect to estimating absolute poverty lines for the measurement of consumption poverty. The mechanics of estimating multidimensional measures are often somewhat more straightforward. However, the first-order dominance (FOD) approach in focus here is not immediately straightforward to code and requires a considerable amount of data management. In both cases, there is substantial advantage to beginning the analytical process with a series of computer codes that reliably accomplish specific tasks within the overall analytical process.
The editors, in collaboration with many others, have for the last fifteen years gradually developed a unique toolkit (i.e. an analytical code stream referred to as Poverty Line Estimation Analytical Software–PLEASe) for consumption poverty analysis in developing countries based on our experience as advisors, researchers, teachers, and practitioners in a wide variety of contexts (see, for example, Arndt et al. 2016). More recently, we have developed analogous software for estimating multidimensional poverty measures based on FOD. The associated code stream is labelled EFOD.
The existence of these software packages served as an important motivation for the Growth and African Poverty Project (GAPP) initiated in 2011 by UNU-WIDER. GAPP has already resulted in the companion volume mentioned above (Arndt et al. 2016), which sought to analyse trends in poverty and wellbeing in as many as possible of the twenty-four largest countries in sub-Saharan Africa (SSA). These studies were conducted by leading international researchers with expert knowledge of the countries in question, working alongside leading local researchers. The analytical teams returned to the primary datasets used for poverty analysis in each country, with an insistence (p.5) on applying best techniques to at least two comparable surveys over the period studied. GAPP completed studies in sixteen of the twenty-four most populous countries in Africa and nine of the top ten.
With respect to consumption poverty measurement, GAPP successfully applied the PLEASe code stream, appropriately modified for country circumstances, to Ethiopia, Madagascar, Malawi, Mozambique, and Uganda. More recently, PLEASe has been successfully applied to Pakistan. With respect to multidimensional poverty measurement, the FOD approach was applied to the Democratic Republic of the Congo (DRC), Mozambique, Nigeria, Tanzania, and Zambia (all using versions of EFOD). While the companion volume sought to illuminate the story of growth and poverty in SSA since about 1995, the present book enters more into the nitty gritty of how specific estimations were performed. The eleven countries featured in this volume provide a diverse set of examples of the challenges and issues confronted in practical poverty assessment, including both differences in data availability and quality as well as variance in country circumstances.
As noted, a salient observation from GAPP is the extraordinarily high level of dependence of many developing countries on external assistance for the conduct of poverty analysis, particularly the analysis of consumption poverty. Nearly all of the countries included in the GAPP project have relied on substantial technical assistance for extended periods in order to produce official consumption poverty rates. Even in the cases where local analysts are strongly engaged, capacity building leaves much to be desired. Two critical factors appear to be at work: (i) the occasional nature of detailed household consumption surveys; and (ii) the complexity of the analysis. This challenging combination generates a situation whereby, once data from a new survey is available for analysis, the personnel who had worked on the previous survey have often either moved on to new areas of activity or have substantial needs for retraining in order to effectively conduct the analysis.
This book seeks to step into this breach for the analysis of consumption poverty and for multidimensional analysis using the FOD approach. Part I of this volume briefly reviews the conceptual issues involved in estimating absolute poverty lines and determining multidimensional first-order dominance. These conceptual issues are then supplemented by a series of practical country applications in Part II, where emphasis is given to the particular challenges and specificities of each case. The country applications illustrate the imperative of adjusting approaches to reflect country-specific circumstances in order for the analysis to be meaningful. It is our hope that such a scaffolding of the issues and practicalities should enable significant numbers of analysts in developing countries to engage in this type of analysis and more rapidly assimilate the concepts and approaches involved.
(p.6) With respect to estimation of absolute poverty, the case studies illustrate that, in practical terms, there often exists vast swathes of agreement across competing methodologies (see also Arndt et al. 2015). For example, within PLEASe, it is relatively straightforward to implement a large array of approaches to absolute poverty line estimation including (but not limited to):
(i) a single national consumption basket with national average prices;
(ii) a single national basket priced at regional levels;
(iii) rural, urban, or more refined regional baskets with associated price differences;
(iv) different approaches to defining the consumption bundles, such as the iterative procedure by Ravallion and Bidani (1994), or simpler alternatives;
(v) fixed or flexible bundles through time; and
(vi) in the case of multiple flexible bundles, imposition (or not) of the utility consistency requirement of Arndt and Simler (2010).
Turning to multidimensional, often non-monetary, indicators, these are now broadly recognized as important (e.g. Alkire et al. 2015; Alkire and Foster 2011; Foster et al. 2013). Non-monetary measures frequently have the advantage of directly relating to policy agendas and are readily available from censuses and household surveys (e.g. is a child attending school, or does a health post exist within 30 minutes travel time from the household?) (Sonne-Schmidt, Østerdal, and Tarp 2008, 2015). While consensus has emerged on the need to consider the multidimensionality of poverty, methods for incorporating multiple indicators into welfare analysis remain contentious with debate centred on the implications of imposing strong assumptions in terms of weighting schemes, the actual extent of new information provided by generating combined indicators, and the nature of welfare functions.
This book furthers this discussion in its use of the FOD approach. This straightforward method allows multidimensional welfare comparisons across populations over time and space while requiring no more restrictive assumptions than a preference to be non-deprived as opposed to deprived in any dimension. Data requirements—which come in the form of binary indicators—are normally less demanding than detailed consumption surveys. Thus, even while addressing multidimensional poverty, the method is frequently less data-intensive in implementation (as demonstrated in the country applications).
Via this book volume, readers have access to the PLEASe and EFOD code streams. We seek to provide these code streams in a manner that is clearly documented, modularized, and transparent. In providing and documenting standard sets of computer codes that can be used as an initial basis for poverty (p.7) analysis, we take motivation from deep involvement in the initial design and dissemination of the standard computable general equilibrium model made available by the International Food Policy Research Institute (Löfgren et al. 2002); the standard global general equilibrium model developed by the Global Trade Analysis Project (GTAP) at Purdue University (Hertel 1997); as well as contributions to the analysis of stabilization and structural adjustment in Africa (Tarp 1993) relying on a coded merger of widely used models for macroeconomic analysis (Brixen and Tarp 1996a, 1996b).
These standard sets of computer codes are of fairly obvious value to students and analysts seeking to gain skills in economy-wide modelling. They have also proven to be a boon to expert modellers as the standard code sets permit initiation of activities from a known, flexible, and advanced baseline. While any tool can be misused, there are large numbers of examples of imaginative analyses, adapted to specific country circumstances, which were greatly facilitated by the existence of a known and flexible base. We have over the years contributed to this academic literature (e.g. Arndt et al. 2012; Tarp et al. 2002), and believe it is critically important that it is widely disseminated and understood in applied work.
Demand for such products has been notably high. For example, the book volume on the GTAP model, which is the reference to the underlying code, records more than 3,000 citations on Google Scholar. The corresponding publication for IFPRI, a technical paper, was the number one download, by a considerable margin, from the IFPRI website for years; and the Brixen and Tarp volumes have been standard references in both teaching and analysis in Africa and beyond. We hope that the PLEASe and EFOD codes can prove similarly valuable to the community engaged in consumption poverty analysis and in multidimensional measures.
1.3 Structure of the Volume
The remainder of Part I of this volume is dedicated to presenting the theory underlying the PLEASe and EFOD code streams. Chapter 4 provides an overview of the practical application of these code streams.
In Part II, a chapter is allocated to each country application; and they present the data issues encountered, the chosen solution to resolving those issues, the modifications to the code stream necessary to accommodate local conditions, and the implications of alternative decisions for the spatial and temporal distribution of measured welfare/poverty. The overall objectives of the applications are to highlight the formidable advantages to beginning from a standardized and known code stream that has been well documented and (p.8) modularized and to provide concrete examples of the issues encountered in practical poverty estimation and the steps taken to address those issues.
We stress that the intent of making code streams available and understood is not to channel poverty analysis into any one particular approach. Rather, the intent is to lower the barriers to entry to conducting detailed, thoughtful, and locally appropriate poverty analyses by providing analysts with functional tools with a known and reliable starting point.
Part III sums up and highlights lessons learned. Part III also contains an additional chapter addressing estimation of inequality. Because poverty lines are employed to compute real consumption across the full income distribution, alternative poverty line estimates imply differences in measured inequality. This chapter explores these differences, building on the country cases. The last chapter concludes and looks forward.
Finally, two appendices provide documentation of the PLEASe and EFOD code streams. These are intended to be living documents available for download alongside the associated code.
Alkire, S. and J. Foster (2011). ‘Counting and Multidimensional Poverty Measurement’, Journal of Public Economics, 99(7): 476–87.
Alkire, S., J. Foster, S. Seth, M. E. Santos, J. M. Roche, and P. Ballon (2015). Multidimensional Poverty Measurement and Analysis. Oxford: Oxford University Press.
Arndt, C., R. Distante, M. A. Hussain, L. P. Østerdal, Pham L. Huong, and M. Ibraimo (2012). ‘Ordinal Welfare Comparisons with Multiple Discrete Indicators: A First-Order Dominance Approach and Application to Child Poverty’, World Development, 40(11): 2290–301.
Arndt, C., A. M. Hussain, V. Salvucci, F. Tarp, and L. P. Østerdal (2015). ‘Poverty Mapping Based on First-Order Dominance with an Example from Mozambique’, Journal of International Development, 28(1): 3–21.
Arndt, C., A. McKay, and F. Tarp (eds) (2016). Growth and Poverty in Sub-Saharan Africa. Oxford: Oxford University Press.
Arndt, C. and K. Simler (2010). ‘Estimating Utility Consistent Poverty Lines’, Economic Development and Cultural Change, 58: 449–74.
Brixen, P. and F. Tarp (1996a). The South African Economy: Macroeconomic Prospects for the Medium Term. London and New York: Routledge.
Brixen, P. and F. Tarp (1996b). ‘South Africa: Macroeconomic Perspectives for the Medium Term’, World Development, 24(6): 989–1001.
Foster, J., S. Seth, M. Lokshin, and Z. Sajaia (2013). A Unified Approach to Measuring Poverty and Inequality: Theory and Practice. Washington, DC: World Bank.
Hertel, T. W. (1997). Global Trade Analysis. Cambridge: Cambridge University Press.
(p.9) Löfgren, H., R. L. Harris, and S. Robinson (2002). A Standard Computable General Equilibrium (CGE) Model in GAMS. Washington, DC: International Food Policy Research Institute.
Ravallion, M. (2016). The Economics of Poverty: History, Measurement, and Policy. Oxford: Oxford University Press.
Ravallion, M. and B. Bidani (1994). ‘How Robust Is a Poverty Profile?’, World Bank Economic Review, 8: 75–102.
Sonne-Schmidt, C., L. P. Østerdal, and F. Tarp (2008). ‘Ordinal Comparison of Multidimensional Deprivation: Theory and Application’, Discussion Paper 08-33, Department of Economics, University of Copenhagen.
Sonne-Schmidt, C., L. P. Østerdal, and F. Tarp (2015). ‘Ordinal Bivariate Inequality: Concepts and Application to Child Deprivation in Mozambique’, Review of Income and Wealth. Available online: DOI: 10.1111/roiw.12183.
Tarp, F. (1993). Stabilization and Structural Adjustment: Macroeconomic Frameworks for Analysing the Crisis in Sub-Saharan Africa. London and New York: Routledge.
Tarp, F., K. Simler, C. Matusse, R. Heltberg, and G. Dava (2002). ‘The Robustness of Poverty Profiles Reconsidered’, Economic Development and Cultural Change, 51(1): 77–108.