| Summary: | <p>This thesis moves beyond the limitations of the consumption measure, to exploring the merits of applying a multidimensional framework to the analysis of poverty within the developing world, using Indonesia as a case study. It confirms that how one defines, and measures poverty may significantly affect who and how many are identified as poor and argues for the crucial need for transparency and public discussion on the reasons behind the choices made within poverty measurement.</p>
<p>In support of the above, three existing measures were first examined; i.e. the Consumption measure, the Global MPI and an existing nationally developed multidimensional measure, which I call the Adjusted MPI. Analysis was conducted using one wave of Indonesia’s nationally representative National Socioeconomic Survey (Susenas 2013). Weak overlap was found between these measures, indicating that firstly, using consumption as the sole indicator of poverty may not be enough and secondly, localised measures may be needed when measuring multidimensional poverty within a country as diverse as Indonesia.</p>
<p>The above results in mind, this thesis proposed the use of the Delphi method to address the arbitrariness in the way existing multidimensional measures, including those examined above, selected their components; i.e. the dimensions, indicators and weights to include within measurement. The Delphi exercise was conducted with policy makers within one district in Indonesia, i.e. Bogor City, West Java. This exercise found consensus on the importance of five dimensions; i.e. (1) Education, (2) Employment and Assets, (3) Living Standards, (4) Health and Safety and (5) Child Health and Contraception, alongside eighteen indicators to proxy deprivation within these dimensions. Agreement was also achieved on a range of weights indicating the relative importance and possible trade-offs between each of these components. In order to determine the ‘exact value’ of these weights, opinion data resulting from the Delphi was analysed using an optimisation exercise based on Data Envelopment Analysis and the benefit-of-the-doubt approach to weighting. The weights resulting from this optimisation exercise, dubbed as ‘expert’ weights, were then utilised to aggregate the dimensions and indicators chosen by Delphi participants to form a ‘new’ measure of multidimensional poverty using Susenas 2013 data. Two comparative exercises were subsequently conducted. The first exercise evaluated the effects of using these ‘expert’ weights versus the commonly applied, ‘equal nested’ weights, within the ‘new’ measure. Both weighting schemes assigned similar ranks to households and identified relatively similar types of households as poor. The ‘expert’ scheme, however, identified a significantly higher extent of poverty compared to the ‘equal nested’ scheme.</p>
<p>The second exercise compared the ‘new’ measure using ‘expert’ weights, with the existing measures examined previously within this thesis. Low levels of overlap were also found. The weak overlap found between the Consumption and the ‘new’ multidimensional measure confirmed the importance of applying both uni and multidimensional approaches when measuring poverty; whilst differences between the globally and nationally constructed multidimensional measures vs. the ‘new’ district level Delphi-based measure supported the argument that internationally-generated measures should be supplemented with localised measures to support the formation of sound policy which aims to address the myriad welfare priorities that are relevant for specific areas.</p>
|
|---|