Discrepancy between Two Approaches to Global Poverty: What Does It Reveal?

Background

The way in which poverty is conceptualized and measured shapes perceptions of need as well as development aid allocation for poor countries. The purpose of this paper is to inform need assessment and aid allocation by identifying the salient dimensions of poverty for low- and middle-income countries. It does so by quantifying discrepancies of two dominant approaches to poverty: monetary and capabilities. 

This paper first looks at whether poverty measurement according to the capability approach (multidimensional poverty headcount ratio, H) predicts poverty measurement according to the monetary approach ($1.90 a day poverty headcount ratio, P₀) for 135 developing countries by year (Q1).  Next, it explores factors that help explain the discrepancy between income and capability measures (Q2). Finally, the analysis examines to what extent sector aid composition responds to the salient dimension of poverty, described by the discrepancy between the two approaches (Q3).

 Methods

The data in this study are synthesized from four sources: OECD DAC, the World Bank, Oxford Poverty and Human Development Initiative ,and global natural disaster statistics (EM-DAT). The sample is composed of total 213 observations by country (i) and year (t), and contain 79 variables related to country development characteristics such as income, poverty, population, life expectancy, adult literacy, natural disasters, and governance. It covers 135 countries spanning 15 periods from 2000 to 2014.

Q1: The paper analyzes the position of 135 developing countries with respect to the fitted line P_0i =α+βH_i + ε_i of the two measures H and P₀, classifying them into income poor (under the line) and capability poor countries (above the line). Q2: Given the nature of residuals that are not explained by the linear model, the analysis uses non-parametric Random Forest algorithm, one of the tree-based methods for regression. Q3: This paper employs longitudinal models including the Fixed Effect (FE) and the Random Effect (RE) model.

Results

Q1: Although there is a fair correlation between the two measures of poverty in many countries, there are also countries with large discrepancies between H and P₀ or residuals. Out of 213 countries, there are 1.5 more capability poor countries (127) than income poor countries (86).

Q2: The model entered with country characteristics covariates can explain approximately 11.82% of the variation in the “residuals” variable. GNI per capita and life expectancy are the two most important variables, followed by population.

Q3: Income poor countries are likely to receive higher economic sector aid relative to social sector aid, while capability poor countries tend to receive higher rate of social sector aid. Based on the Random Effect Model, one unit increase in residual is associated with a 1.31 decrease in the rate of social to economic sector aid holding control of other variables.

Implications

This paper examines discrepancies between monetary and capabilities approaches to poverty, finding that the two approaches differ in how they capture poverty and how they inform aid allocation. This study illustrates how inconsistencies in distinct measures of international poverty can be used to target, monitor, and evaluate global aid distribution.