Bayesian Population Projections for Every City in the World
Leontine Alkema, National University of Singapore
Jennifer L. Chunn, University of Washington
Patrick Gerland, United Nations Population Division
Danan Gu, United Nations Population Division
Gerhard K. Heilig, United Nations
The United Nations Population Division publishes the World Urbanization Prospects, which includes projections for all capital and major cities in the world. A deterministic model is used to construct the city projections, based on the most recent observed growth differential between the city population and the urban population in the country. An uncertainty assessment of future outcomes is lacking. In this paper, we propose a Bayesian time series model to capture changes in city-urban growth differentials over time, and to construct probabilistic city projections. We evaluated the Bayesian city projection model by an out-of-sample exercise, and found that the proposed model is reasonably well calibrated and has smaller projection errors than the current projection method.
Presented in Poster Session 2