A Robust Multi-dimensional Poverty Profile for Uganda

Abstract

We compute a Multi-dimensional Poverty Index (MPI) for Uganda following the specification by Alkire and Santos and the general approach of Alkire and Forster. Using household survey data we show how the incidence of multi-dimensional poverty has fallen in recent years and we use the decomposability features of the index to explain the reduction in multi-dimensional poverty. The robustness of our results is tested in a stochastic dominance framework and using statistical inference. Notably, we extend the one-dimensional analysis of stochastic dominance to include household size as a second dimension, thus taking into account that MPI indicators are collected at both household and individual levels. Moreover, we extend the standard two-stage application of the MPI to include a third stage, which is important given the high degree of multiple deprivations within the standard of living dimension.

Keywords:

• Measurement
• Poverty
• Human development
• Economic development
• Capability approach

Acknowledgements

We pay tribute to Sebastian who cruelly left us in April 2013 during the publishing process. He made this possible and may he rest in peace. Special thought to his wife Cecilia and daughter Nora.

About the Authors

Sebastian Levine worked for UNDP from 1999 to 2013. From 1999 to 2002 he was Junior Professional Officer in Lesotho. He then went on to serve as Senior Economic Advisor for UNDP in Namibia (2003–2009) and Uganda (2009–2011), before being reassigned in February 2011 to serve in New York as Policy Advisor at UNDP's Regional Bureau for Africa. During this last assignment Sebastian was responsible for leading the preparation of the 2012 Africa Human Development Report and for designing UNDP's regional “Extractive Industries for Human Development” programme for Africa. He also authored a number of papers on poverty measurement published in international peer-reviewed journals. Prior to joining UNDP, Sebastian worked as a research economist at the Institut for Konjunktur-Analysea, a think-tank in his home town of Copenhagen, Denmark. He was 43 years old when he tragically died on 2 April 2013. Sebastian served UNDP with the highest distinction. He is survived by his wife Cecilia, and six-year-old daughter Nora.

James Muwonge is working for the Uganda Bureau of Statistics as the Director of Socio-economic Surveys. He is responsible for planning and coordinating the implementation of household-based surveys at the Bureau, a task he has executed for the last 15 years. He has collaborated with local and international organizations to undertake research, including the World Bank, USAID, among others. He holds a master's Degree in Economics of Development (Institute of Social Studies, The Hague). James has also written papers on the role of household surveys in Uganda and on poverty in general. He has also co-authored papers on land productivity and investments and the use of crop diaries in improving agricultural statistics in Uganda

Yélé Maweki Batana is currently a Senior Poverty Economist at the World Bank, Mali office. Prior to that, he served successively as Teaching Assistant at Université de Lomé, Visiting Fellow at Oxford University, Postdoctoral Fellow at Université de Montréal, and more recently as Resource Person at the Partnership for Economic Policy of Université Laval. He holds a PhD in Economics from Université Laval, a DEA-PTCI from Université de Ouagadougou and a master's degree from Université de Montréal. His main areas of research include public economics, welfare economics, econometrics and development economics. A member of the African Economic Research Consortium, he was associated with some thematic and collaborative researches on social and health inequalities among children in Togo, and on women's empowerment in Africa. His is author of several publications in international reviews and edited books.

Notes

1. The MPI satisfies other important axioms such as symmetry, replication invariance, sub-group consistency and decomposability (Alkire and Santos 2010).

2. We also examined the 1995 DHS but decided not to include this in the present analysis because it did not have anthropomorphic data needed to compute the body mass index (BMI) for women, which is one key indicator in the MPI.

3. This is not the case with the current one-dimensional poverty measure used by the World Bank and other international agencies, which is based on the US$1.25 international poverty line expressed in 2005 Purchasing Power Parities. As documented in Levine (2012), a series of methodological issues (e.g. differences in poverty thresholds, adjustments for household composition, and price adjustments) in the computation of the national and international poverty measures generate different results and conclusions about the poverty levels in Uganda.

4. In the earlier surveys, indicators such as energy for cooking and nutritional status of women were not collected.

5. Several Ugandan stakeholders consulted in the preparation of this study have voiced their concern regarding the inclusion of the cooking fuel indicator. A more country-specific approach to developing the MPI could take this and other concerns into account. The type and number of indicators is relatively easily incorporated within the framework presented here.

6. According to the UN Population Division online database (http://esa.un.org/unpd/wpp/ [accessed July 2011]), Uganda's population almost doubled after 1990, reaching 33 million in 2010. Using medium-variant assumptions, the country's total fertility rate for the period 2005–2010 was estimated at 6.38, the sixth highest in the world.

7. Since the deprivation in each dimension does not follow the same trajectory depending on the household size, it is also likely that the weights assigned to these dimensions impact the household poverty. Thus, an increase in the weight of health dimension, for example, will change the deprivation contributions and increase the MPI of large household. It will be sound, when making poverty comparisons, to take account of the differences in size. Another advantage of doing this is the fact that the comparisons performed in this way will be robust to the changes in dimensional weights.

8. The dominance analysis here comes down to a simple one-dimensional dominance since all dimensions are aggregated in a vector c of deprivation counts.