DEEP Dives | October 2025 | New evidence, insights and policy takeaways on extreme poverty

Welcome to DEEP Dives, the first research bulletin of the Data and Evidence to End Extreme Poverty (DEEP) programme!

We spotlight recent research on innovative methods for measuring extreme poverty, highlighting new insights and evidence into its drivers as well as policies and programmes that take a data-driven approach to ending extreme poverty.

The newsletter is curated by the DEEP team at Oxford Policy Management.

Here’s what we’ve been reading recently:

The last few months has been a busy time for global poverty measurement. Global poverty estimates have been updated, resulting in an increase in the number of people living in extreme poverty globally from 713 to 838 million. A recent blog from Our World in Data helpfully outlines what has changed and why – delving into how the international poverty line is defined and what this change means for our understanding of extreme poverty globally.

Incorporating the costs of disabilities into poverty measurement – Adapting Monetary Poverty Lines to Reflect the Costs of Disabilities: An Empirical Application to Peru by Pinilla-Roncancio, Schwarz, Smith and Stampini adjusts Peru’s monetary poverty line to reflect the extra costs faced by people with disabilities. Incorporating these costs would imply that a person with disabilities requires an income of 1.47 times the poverty line to be considered ‘‘non-poor’’. Further, it suggests that 15% more individuals would be classified as ‘‘poor’’. (See also DEEP research from Bangladesh and Tanzania, which also calculates disability-related expenditure and the implications for poverty measurement.)

Poverty mapping – integrating different sources of data. This Asian Development Bank brief, Data Integration Approaches to Strengthen Asia and the Pacific’s Statistical Capacity to Map Poverty, provides a really useful overview of different approaches and recent research that integrates non-traditional sources of data (including call detail records and satellite data) to provide lower cost and more timely poverty maps (also something DEEP has been grappling with and aiming to make the techniques more accessible).

In Measuring Multidimensional Poverty in Conflict Zones: A Remote Sensing Method for the Tigray Conflict, the authors combine armed conflict location and event data with remote sensing measurements, such as nighttime lights and a normalised difference vegetation index, to develop a multidimensional poverty index – finding that 94.1% of Tigray’s population lived in extreme multidimensional poverty (largely as the result of conflict) by late 2021. They argue that such innovative approaches to poverty measurement are the most viable in data-scarce and conflict environments (which is particularly relevant given the World Bank analysis that extreme poverty is rising fast in economies hit by conflict and Instability, with 60% of the world’s population considered ‘extremely poor’ projected to live in such contexts by 2030).

Can new approaches to poverty mapping improve household-level targeting?

In Evaluating Area-Level Features for Proxy Means Test Models: Evidence from Rural, Semi-Urban and Urban Districts in Poverty Targeting, Mariyah and Wobcke evaluate the use of area-level features in improving household-level expenditure estimates and ultimately reducing inclusion and exclusion errors of the proxy means test models often used to target social assistance programmes. They applied the technique to Indonesia and the answer wasn’t straightforward: though there are indications of improved targeting accuracy in semi-urban districts, incorporating area characteristics in urban and rural districts can actually reduce accuracy since those districts are very diverse and may not be good representations of all the constituent households. In urban and wealthier areas, for instance, poor households are more likely to be excluded (or classified as ‘‘non-poor’’).

And finally – the importance of education to ‘‘pro-poor’’ growth. New analysis concludes that, between 1980 and 2019, education could account for about 45% of global economic growth and 60% of pre-tax income growth among the 20% of people considered the world’s poorest population– see Gethin’s application of a ‘distributional growth framework’ for more in Distributional Growth Accounting:Education and the Reduction of Global Poverty, 1980–2019.