The Importance of Reliability and Construct Validity in Multidimensional Poverty Measurement: An Illustration Using the Multidimensional Poverty Index for Latin America (MPI-LA)

Héctor E Nájera Catalán,David Gordon

doi:10.1080/00220388.2019.1663176

Abstract

The empirical properties of a multidimensional poverty index require robust assessment. However, poverty research is yet to systematically implement measurement theories and practices that have been proven to be successful in other fields. Measurement theory has been developed over more than 100 years to produce indexes that are scientific (falsifiable) in that researchers put under scrutiny whether their value judgements and assumptions result in scales that have high internal consistency (reliability) and capture the phenomenon they aim to measure (validity). The paper uses the Multidimensional Poverty Index for Latin America (MPI-LA) to illustrate the importance of adopting sound measurement practices. The MPI-LA aims to be an improvement over the widely applied Unsatisfied Basic Needs (UBN) approach. However, its empirical development was based on ad hoc non-standard methods and principles, making the conclusions of the developer’s analyses unfalsifiable and prone to confirmation bias. This analysis includes six countries and two time periods. The findings suggest that the MPI-LA is an unreliable measure of poverty and that the pre-specified dimensional structure is invalid. The paper illustrates how standard principles like reliability and validity can be used to inform the discussion about the statistical properties of a given poverty index.

Highlights

Poverty has a range of different meanings; such as inadequate material living conditions, social position, and economic circumstances, and so forth (Spicker, Alvarez, & Gordon, 2006)
This section presents the findings of the reliability analysis and the assessment of the presumed dimensional structure of the Multidimensional Poverty Index for Latin America (MPI-LA). It presents the findings using Classical Test Theory (CTT) followed by the results from the Item Response Theory (IRT) analysis
The input data were the raw deprivation variables (0 = not deprived 1 = deprived) but using the Multidimensional Poverty Index (MPI)-LA differentially weighted variables would not have changed the results. α was estimated in Stata to determine the sensitivity to the sampling weights, but they had little impact upon the reliability results

Summary

Introduction

Poverty has a range of different meanings; such as inadequate material living conditions, social position, and economic circumstances, and so forth (Spicker, Alvarez, & Gordon, 2006). Poverty can be defined as the lack of command of resources over time and deprivation is its consequence (Gordon, 2006; Townsend, 1979) This is a multidimensional definition of poverty in that ‘individuals, families and groups in the population can be said to be in poverty when they lack the resources [cash income, capital assets, value of employment benefits in kind, public social services and private income in kind] to obtain the types of diet, participate in the activities and have the living conditions and amenities which are customary, or are at least widely encouraged or approved, in the societies to which they belong’ (Townsend, 1979). Gordon income’ and highlighted the importance of the distribution of non-monetary resources in poverty measurement

Methods

Results

Conclusion