The counting approach to multidimensional poverty. The case of four African countries

• Published date: 01 June 2019
• Author: Valérie Berenger
• Date: 01 June 2019
• DOI: http://doi.org/10.1111/saje.12217

South African Journal of Economics Vol. 87:2 June 2019

doi : 10.1111/ saje .12217

200

THE COUNTING APPROACH TO MULTIDIMENSIONAL

POVERTY. THE CASE OF FOUR AFRICAN COUNTRIES

VALÉRIE BERENGER†,*

Abstract

This paper investigates the levels and evolution of poverty in Malawi, Mozambique, Tanzania and

Zimbabwe using the decomposability properties of poverty measures based on a counting

approach. We compare poverty measures such as the Alkire and Foster index with alternative

poverty indices that are sensitive to inequality. Poverty is estimated using Demographic and Health

Surveys for different years for Malawi (2004, 2010 and 2015), for Mozambique (2003 and 2011),

Tanzania (2004, 2010 and 2015) and for Zimbabwe (2005, 2010 and 2015). Our findings show

that one obtains insightful information when looking at the breadth and inequality components

of multidimensional poverty.

JEL Classification: I32, D63

Keywords: Countin g approach, multidimensional poverty measur ement, Malawi, Mozambique,

Tanzania, Zimbabwe

1. INTRODUCTION

In September 2015, the adoption by the international community of Sustainable

Development Goal (SDG) 1.2 of ending poverty in a ll its forms everywhere incorporates

an explicit multidimensional focus. Simultaneously, the SDGs encompass implicit or

explicit goals addressing inequality.

The recent adoption by the United Nations Development Programme (UNDP) of the

so-called Multidimensional Poverty Index (MPI) illustrates the importance of taking

multiple dimensions of poverty into account and addressing them. The MPI draws on

the counting approach developed by Alkire a nd Foster (2011) and assesses poverty along

the same dimensions as the Human Development Index (HDI). It not only counts the

share of the population deprived in at least 30% of the dimensions, but it also provides a

snapshot of the breadth of poverty. This new measure, computed for over 100 countries

using global standa rds such as the international monetar y poverty line, is somehow sim-

ilar to the calculation of monetary poverty.

However, recent literature stresses some conceptual and empirical weaknesses of the

MPI. Among its conceptual shortcomings is its lack of sensitivity to inequality in the

distribution of deprivations. The MPI does not provide any information on whether

* Corresponding author: University of Toulon, LEAD, France, a nd University Côte d’Azur,

CNRS, GR EDEG, France. E-ma il: berenger@univ-tln.fr

† University Côte d’Azur, CNRS, GR EDEG

This is an ope n access arti cle under the terms of t he Creative Commons A ttribution-NonComm ercial-Share Alike Licen se, which

permits use a nd distribution in a ny medium, provided t he original work is pr operly cited, the use i s non-commercial a nd the content

is offered und er identical terms .

© 2019 UNU-WIDER. South A frican Journa l of Economics published by John W iley & Sons Ltd on behal f of

Economic Societ y of South Africa.

South African Journal

of Economics

201South African Journal of Economics Vol. 87:2 June 2019

© 2019 UNU-WIDER. South A frican Journa l of Economics published by John W iley & Sons Ltd on behal f of

Economic Societ y of South Africa.

a decrease in poverty af fects the poorest of the poor. Before Alkire and Foster (2011)

suggested their measures, independent studies proposed several poverty measures using

a counting approach. These papers suggested alternative ways of defining multidimen-

sional poverty indices that comply with the basic axiomatic properties of multidimen-

sional poverty indices based on continuous variables (Bossert et al. , 2013).

Recently, Silber and Yalonetzky (2013) proposed a general framework that integrates

ordinal variables into the measurement of multidimensional poverty. Drawing on this

framework, Bérenger (2017) put special emphasis on the decomposability properties of

the three “I”s of poverty of the main inequality sensitive poverty measures based on a

counting approach. She showed the complementarities between several measures when

analysing trends in poverty levels in Egypt and Jordan and in three South East Asian

countries (Bérenger, 2016).

Drawing on Bérenger (2017), the main goal of th is paper is to make use of the decom-

posability properties of four main counting-based poverty measures found in the liter-

ature. We show how to use them in a way that complements the information provided

by the MPI to assess the levels and trends in multidimensional poverty in Mozambique,

Malawi, Tanzania and Zimbabwe in the 2000s. We compare the results obtained when

using poverty measures based on Alkire and Foster (2011), such as the UNDP’s MPI,

and counting-based poverty measures sensitive to inequality. These include measures

proposed by Chakravarty and D’Ambrosio (2006), by Rippin (2010), and the one sug-

gested by Silber and Yalonetzky (2013). As these measu res cover all deprived individuals,

using a union approach, it is possible to manipulate them to adopt a more flexible ap-

proach that fits the identification approach of the MPI.

Section 2 presents four main multidimensional povert y indices based on the counting

approach. Section 3 shows the results obtained when applying these multidimensional

poverty indices to data from the Demographic and Health Surveys (DHS) of Malawi,

Mozambique, Tanzania and Zimbabwe. Section 4 gives concluding comments.

2. MULTIDIMENSIONA L POVERTY USING A COU NTING-BASED APPROACH

WITH OR DINAL VAR IABLES

In the counting approach, the way to summarise information amounts to determining a

poverty line in each dimension and aggregating the dimensions for each individual, and

then across individuals, in order to derive a summa ry measure of deprivations in multiple

attributes. This order of aggregation preserves the essence of the multiple facets of pov-

erty as it embeds the association between simultaneous deprivations experienced by the

individual. Moreover, the central features of the counting approach are the use of binary

variables and the comparison of individual achievements in a given dimension with a

dimension-specific poverty line. Using solely binar y indicators of well-being, the MPI

summarises the information by counting the number of dimensions in which an indi-

vidual suffers from deprivations. The ma in innovative feature of Alkire and Foster (2011)

is that they combined the counting approach that has been widely used at the empirical

level in social sciences to construct indices of deprivation or social exclusion measures

with the axiomatic approach to multidimensional poverty. Based on the extension of the

Foster et al. (1984) indices, they suggested a class of poverty mea sures “the adjusted FGT”

measures that are able to capture the three “I”s of poverty (Jenkins and Lambert, 1997)