Spatio-Temporal Modeling of U.S. State-To-State Migration Flows

• Activity Number: 18
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Trevor Oswald*+
• Keywords: Bayesian ; Census ; Hierarchical ; IRS ; Population ; Prediction

Abstract

Nearly 7 million U.S. citizens move from their state of residence each year. Determining the causes of why individuals and families leave their home has been a topic of debate among scientists beginning with Ravenstein's papers on migration in the 1880's. Recently migration research has experienced an increase of interest due to the readily available amount of data being produced from many governmental agencies including the Internal Revenue Service and the U.S. Census Bureau, as well as the desire of policy makers to have an accurate prediction of population for funding allocation. Historically, models have predicted migration flows without accounting for sampling variability, enforcing the constraint of a system in equilibrium, or making use of the inherent spatio-temporal structure that exists within migration flow data. We develop a Bayesian hierarchical model to fit the 1988-2010 yearly IRS migration flow data. Using covariate information and novel empirical basis functions for areal data with spatially and temporally varying parameters, we are able to predict the state-to-state migration flows one year ahead, while simultaneously accounting for uncertainty in the data.