New way to measure poverty may transform how international aid and development work operate

How can researchers measure poverty and understand local development needs in places where conventional data collection is difficult or impossible?

To overcome this obstacle, Rutgers researchers have for the first time used georeferenced content from Twitter (now known as X) as a poverty measurement tool, offering a new way to understand development from the ground up.

“This approach could transform how international aid and development work operates,” said Woojin Jung, an assistant professor at the Rutgers School of Social Work and the principal investigator of the study. “Instead of waiting years between expensive surveys, organizations could get real-time insights into community needs by analyzing what people are actually talking about online.”

The paper, “Digital Pulse of Development: Constructing Poverty Metrics from Social Media Discourse,” co-written by Jung and Tawfiq Ammari, an assistant professor of library and information science at the Rutgers School of Communication and Information, with Andrew H. Kim, Charles Chear, Vatsal Shah, and Ying Hung, received an honorable mention in the Applied and Quantitative Modeling Category from the organization Equity and Access in Algorithms, Mechanisms, and Optimization.

Jung will present the paper and accept the award at the fifth Association for Computing Machinery Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (EAAMO 2025) being held from Nov. 5 to Nov. 7 at the University of Pittsburgh.

“In many developing countries,” Jung said, “surveys are expensive, infrequent and often miss remote areas. We wondered if social media posts—specifically posts on Twitter—could serve as a real-time window into what communities are actually experiencing and discussing.”

Three main discoveries emerged from their study of Twitter data in Zambia:

• Twitter reflects real poverty patterns: The topics people post on Twitter strongly correlate with their village’s wealth level. Poor villages tend to discuss immediate, local concerns such as food shortages and corrupt politicians, while wealthier villages engage with broader policy issues.
• Social media can predict wealth: Using just seven development-related Twitter topics (such as discussions of food systems, mining and social inequality), their model could explain more than 60% of the variation in village-level wealth—performing as well as satellite imagery but with much clearer explanations.
• Missing data can be filled in: They developed methods to estimate poverty levels in areas with little or no activity on Twitter by using spatial patterns from nearby areas.

To conduct the study, the researchers combined official poverty data from Zambia’s 2018 Demographic and Health Surveys with more than 20,000 geotagged posts on Twitter from 2019 to 2021. They used artificial intelligence to identify 103 different topics in the posts, then worked with local experts to identify seven topics most relevant to development issues. They tested whether these Twitter topics could predict village wealth levels and compared their approach to conventional methods such as satellite imagery and building footprint analysis.

“Our findings are especially valuable because we capture citizens’ own perspectives on their problems—what researchers call ‘development conceived, measured and planned by citizens’ rather than outsiders,” Ammari said.