We propose a new method of using the public geocodes in the Demographic and Health Surveys for integrating environmental context into analysis. Currently, the public geocodes in the DHS are produced by randomly displacing rural cluster locations by up to 5km (10km for 1 percent of points). DHS currently advises researchers using these geocodes to average environmental characteristics over a 5–10km buffer. We hypothesize that a smaller buffer around a neighboring settlement—even an incorrect settlement—is a better measure of true environmental context in rural areas than the large buffer around the published point. This hypothesis is motivated by the rapid growth of literature pointing to the importance of connecting geographic measures of context to the lived experiences of individuals and by our assumption that smaller buffers more adequately measure those lived experiences. We test our hypothesis on an analysis that combines DHS data with fine-scale measures of human agricultural activity, with the truth provided by access to the confidential true cluster location.
Published in a peer-reviewed journal of the Population Council.