The many faces of vulnerability: Utilizing non-income data to reveal multidimensional poverty in urban Bangladesh

Nuzhat Fatima, Research Assistant, Policy and Strategic Advisory Unit, UNDP

In a world that is increasingly faced with crises such as the COVID-19 pandemic, the Ukraine-Russia war, and rising costs of living, accurately identifying the vulnerabilities of deprived groups requires more than just consumption-expenditure-centered measures of poverty and commensurate data on income.

A worrying trend is emerging on top of this. Groups hovering just around monetary poverty thresholds – thereby not categorically marginalized ‐ are becoming increasingly vulnerable. Policy measures therefore need to address vulnerabilities that go beyond income and affect different communities. The UN World Data Forum 2023's second thematic area focuses on the value of data for improving lives and building a pathway to better data for sustainable development. Specifically, TA2.16 "Improving decision-making by integrating multidimensional data" aims to address a key limitation of traditional monetary measures of poverty by utilizing multidimensional, non-income data for shock-resilient social protection.

One tool that can complement income- and other monetary-based measures of poverty is the Multidimensional Poverty Index (MPI), developed by the United Nations Development Program (UNDP) and the Oxford Poverty and Human Development Initiative (OPHI). The MPI "score" captures deprivations in non-monetary dimensions of wellbeing by using non-income data on a range of indicators in education, health, and living standards to calculate poverty levels for a particular population. Typically, the higher the score, the greater the level of multidimensional poverty.

To assess the potential effectiveness of multidimensional approaches in addressing vulnerability during crises, the Research Facility at the UNDP Bangladesh Country Office conducted an analysis of data from its Livelihoods Improvement of Urban Poor Communities (LIUPC) project. This program, which covers four million urban poor across 19 cities in Bangladesh, utilizes the MPI to identify deprivation levels of potential recipient households as part of its poverty reduction efforts.

‘MPI-poor’ households – those with high levels of deprivation across these indicators–received conditional cash grants through the LIUPC project to start or expand their businesses. In addition, these households received COVID-19 relief in the form of cash, food, or preventive materials as unconditional emergency support, separate from the grants that were already part of the project.

This study was presented in a recent UNDP Development Futures Series brief which compared the MPI figures of recipient and non-recipient households across a time period spanning from just before the pandemic, i.e. around November 2019, to mid-pandemic, January 2022. Households were split into three groups:

  1. Recipient households- those with high MPI scores who received conditional cash grants from the project,
  2. MPI-poor non-grantee households- those with high MPI scores who did not receive cash grants, and
  3. Vulnerable MPI non-poor households- those with MPI scores not sufficiently high for cash-grant eligibility.
  4. (Not all MPI-poor households ultimately received grants. The study’s methodology explains why here.)

Some of the study's findings were intuitive, such as the fact that the business grants provided by the project helped poor households reduce their multidimensional poverty levels despite the pandemic's impact. However, the analysis also yielded some less obvious policy insights.

One policy insight from the analysis is to emphasize the collection of non-income data. As MPIs are constructed using non-income data, adopting MPI-based programmatic support during crises can identify vulnerabilities in households beyond income, especially as national-level poverty measurements often overlook regional disparities in living, education, and health standards in a country. MPI indicators can be further disaggregated to determine which dimensions contribute the most to poverty. This is particularly useful during crisis periods, where deprivations often occur due to various factors beyond income.

Our understanding of multidimensional poverty could also be greatly improved with dynamic data. Insights from the study were based on the analysis of static data collected by the project, which could not capture real-time changes as they occurred. If the data had been dynamic, reflecting changes in households' conditions during the pandemic, the project could have targeted recipients more effectively and determined the nature of relief needed more accurately.

Additionally, vulnerable non-poor groups in development programming need to be included. The study's findings highlight the emergence of the 'new poor': vulnerable MPI non-poor households who experienced on average an increase in their multidimensional poverty levels during the pandemic. These households typically remain outside the purview of policy and project support measures, as they fall just above the 'poor' threshold under normal circumstances. However, in the absence of support during crises, their vulnerabilities become more apparent and need to be addressed in development programming.

Lastly, using a context-specific MPI can better complement income-based poverty measures. Aggregate changes in MPI scores may obscure specific vulnerabilities in households, such as members with disabilities or belonging to a susceptible age group (such as youth or elderly), or differences due to regional characteristics. For example, despite an overall drop in MPI scores among recipient households, multidimensional poverty did not improve for households with disabled members. Thus, the use of a uniform MPI metric regardless of variations in local contexts risks overlooking specific needs of underserved communities.

Bangladesh's experience with COVID-19 has highlighted the impact of non-income factors on impoverished groups in a crisis, particularly in resource-poor settings. In such contexts, the use of MPIs as an analytical tool can help identify the most vulnerable individuals and reveal patterns of poverty. However, the efficacy of MPIs can be further improved by collecting dynamic data that can reflect real-time changes in households' conditions during crises. This would enable development assistance providers to more accurately target program recipients and identify specific vulnerabilities in crisis settings, which is crucial in ensuring that no groups are left behind.