Policy Motivation

Numbers, as Szdkely (2005) observed, can move the world. Thus, the third and equally central motivation for multidimensional poverty measurement is to inform policy, and thereby join the struggle to confront and overcome the pressing hazards and disadvantages that blight so many lives.

Thus far we have discussed the ethical or normative reasons to consider the many faces of poverty. These are echoed in the policy fields. Scouring many empirical studies, we have concluded that it does not seem possible to proxy multidimensional poverty levels or trends using a single indicator. Many important and informative measurement methodologies have been developed, and will continue to be used and advanced in appropriate contexts, and Chapters 3 and 4 discuss these in depth. Further, this area of study is advancing rapidly. Still, in this last section, we mention why the AF methodology may add value empirically and theoretically, and in so doing open a window onto policy.

The building blocks of counting measures, including the AF class, are individual deprivation profiles. These show what deprivations one particular person or household experiences. For example, we might find that someone called Miriam is deprived in nutrition, in housing, in sanitation, and clean fuel, and literacy. This is called Miriam's joint distribution of deprivations—the deprivations she experiences at that particular point in time. These are summed, with weights, to create Miriam's deprivation score.

But would any measure do? A salient feature of the AF methodology is its properties—as described in Chapters 2, 3, and 5—which make it an attractive option for informing policy transparently. Among AF measures, the so-called Adjusted Headcount Ratio or M₀ measure is particularly suitable due to three properties: (a) its ability to use ordinal or binary data rigorously, given that poverty indicators regularly have such data; (b) its ability to be decomposed by population subgroups like states or ethnic groups, to understand disparities and address the poorest; and (c) its ability to be broken down by dimensions and indicators, to show the composition of poverty on aggregate and for each subgroup. To this we might add a non-formal feature, which is the intuition of the measure and its partial and consistent sub-indices, which include a familiar headcount ratio, and also a novel feature reflecting the intensity or average share of deprivations poor people experience.

Because of these properties an M0 measure has been described as a high-resolution lens. The single index value gives an overview of poverty levels and how these rise or fall over time. But it can (and should) be unfolded in different ways—by groups and by dimensions; at a single point in time or across time—to inform various policy purposes. It can therefore been used:

• to produce the official measures of multidimensional poverty;

• to identify overall patterns of deprivation;

• to compare subnational groups, such as regions, urban/rural, or ethnic groups;

• to compare the composition of poverty in different regions or social groups;

• to report poverty trends over time, both on aggregate and by population subgroups;

• to monitor the changes in particular indicators;

• to evaluate the impact of programmes on multiple outcomes;

• to target geographical regions or households for particular purposes;

• to communicate poverty analyses broadly.

Initial applications of multidimensional measurement methods used individual- or household-level data. More recently, the methodology is being applied to different units of analysis and with respect to different focal areas such as women's empowerment, targeting, child poverty, governance, fair trade, energy, and gender, with other applications, including using mixed methods and participatory work, in progress. The policy avenues for these alternative applications are a bit different from those outlined above, but continue to draw on the policy-salient features of the methodology.

1.3 Content and Structure

This book aims (a) to introduce the AF methodology as one approach among a wider set of multidimensional techniques; (b) to provide a clear and systematic introduction to multidimensional poverty measurement in the counting and axiomatic tradition, with a specific focus on the AF Adjusted Headcount Ratio (M₀); and (c) to address empirical and normative issues, as well as recent methodological extensions in distribution and dynamics.

The book may be divided into four parts, each containing two or three chapters. The first part introduces the framework for multidimensional measurement and system­atically presents and critically evaluates different multidimensional methods that are frequently used for assessment of multidimensional poverty. The second part presents the counting-based measures that have been widely used in policy, and the Alkire-Foster methodology which joins together the axiomatic and counting approaches. The third part addresses pre-estimation issues in poverty measurement—the normative and empirical aspects of constructing a poverty measure. The fourth and final part of the book deals with the post-estimation issues—analysis after the poverty measure is constructed.

In the first part, Chapter 2 presents the framework for the whole book, outlining the basics of unidimensional and multidimensional poverty measurement, introducing the terminology and notation to be used throughout the book, discussing the scales of measurement of indicators and comparability across dimensions, and describing with illustrations the properties of multidimensional poverty measures.

In the second part, Chapter 4 reviews the counting approaches to multidimensional poverty measurement that have been widely applied and used for policy. Then Chapter 5 provides an in-depth account of one particular axiomatic and counting-type multidimen­sional poverty measurement methodology: the AF counting methodology. Specifically, the chapter presents the AF methodology of identification and presents the Adjusted Headcount Ratio or M₀ measure and its partial and consistent sub-indices.

In the third part, Chapter 6 clarifies and outlines the normative choices in measurement design, drawing on Sen's capability approach and related applied literature; and Chapter 7 provides a synthetic overview of distinctive practical issues in multidimensional poverty measurement design and analysis. In the final part, Chapter 8 presents avenues for performing robustness analysis and statistical inference, and Chapter 9 discusses different methodologies for conducting distributional and dynamic analysis. Chapter 10 presents some relevant regression techniques for analysing the M₀ measures.

1.4 How to Use this Book

This book was written with academic researchers, technical staff in governments and international agencies, and graduate students in quantitative social sciences in mind. Readers are likely to have a quantitative interest or training and come from different disciplines, ranging from economics to statistics, sociology, social policy, gender studies, education, public health, development studies, and area studies.

Naturally, some sections will be of more relevance to each reader than others.

Also, readers will come to this book with varying degrees of familiarity with terms and operations. Some will have a deep familiarity with axiomatic approaches to poverty meas­urement; others with empirical operations such as bootstrapping, regression analysis, and robustness checks; whereas others might have greater familiarity with the choice of indicators and cutoffs. Still other readers will have knowledge of tests of indicator validity and reliability or may focus more on categorical and ordinal data analysis or on the link between measurement and policy processes. We have sought to explain key operations or to point researchers to background reading. Some content may seem rather basic but is included in order to be intelligible to others from different backgrounds. In addition, a substantial body of more intuitive and less technical materials that could not fit in the book are available on our associated website,. These online resources also include relevant software codes, training videos, and problem sets. The book, together with the online resources, thus provides a systematic introduction to the field for those learning these techniques and a set of reference materials for those implementing multidimensional measures.

₂