Using Big Data to Aid Food Insecurity and Poverty

The potential of remotely sensed data and machine learning methods to predict poverty has been demonstrated. Currently available methods, however, perform relatively poorly in identifying the most vulnerable communities.

To remedy this deficiency, Christopher Barrett, Applied Economics and Management; David Matteson, Statistical Science and Social Statistics; and Ying Sun, Soil and Crop Sciences are combining the latest advances in satellite vegetation remote sensing and crowd-sourced food price data, using cutting-edge machine learning and time series estimation techniques. The data will generate a more accurate, timely, and lower cost monitoring tool for Feed the Future (FtF) outcome indicators, such as asset holdings, food consumption, poverty, and nutritional status at the community level in FtF countries’ low-income rural regions.

Accurate, low-cost, and timely prediction of poverty and related nutritional indicators among rural communities is highly desirable for monitoring, targeting, and evaluating FtF programs. The key challenge is to develop a rigorous approach to extract useful information, based on open-access datasets, especially near-real time novel datasets that have only recently become available. Once the model is successfully built, an important component of the work will be to distribute this new approach to FtF country decision makers and technical advisers as well as train them to use the tools to guide local policy.

Cornell Researchers

Funding Received

$1 Million spanning 3 years

Other Research Sponsored by United States Agency for International Development