Evaluating County Population Projections Using a Century of Census Data

Stefan Rayer, University of Florida

This study evaluates summary measures of population projection error for counties in the continental United States over the period 1900–2000. The paper compares 125 projection horizon/base period combinations for seven primary trend extrapolation techniques and three averaging methods. The study investigates the relationship between the length of the projection horizon and the length of the base period with respect to accuracy and bias. It then focuses on county characteristics, and their impact on forecast error and bias. Throughout, the paper highlights how the use of historical data covering a wide range of census years, a broad sample of geographical units, and a rich cross-section of projection techniques can assist the development of population projection models for the future.