Constructing Annual Migration Matrices with Census Administrative Data: The Master Address File – Migration Flows (MAF-MIF) Data Set

Nationally representative data sets that support research on internal migration are vital for understanding trends and dynamics driving population redistribution, but they also have limitations. In short, surveys used to estimate migration rates overall and by demographic group are representative of the U.S. population, but do not allow for estimation of migration rates by county in a single year. County-level migration probabilities can be estimated by county for each year using administrative records but are not representative of the US population or decomposable by demographic characteristics. In this technical note, we describe the construction of a county-level migration flow data set from administrative records available in the Census Bureau that can be decomposed into demographic groups by age, sex, race/ethnicity, and nativity. We refer to this as the Master Address File - Migration Flows (MAF-MIF) data set. The MAF-MIF can be used to address a range of research questions about internal migration patterns focusing on the origin and destination counties of specific demographic groups of movers. To assess the validity and reliability of the MAF-MIF, we investigate whether the inter-county migration probabilities estimated from the MAF-MIF match comparable migration probabilities estimated from the public use Current Population Survey (CPS). Several anomalies are identified and treated by weighting the data using national CPS migration probabilities. We then compare the weighted MAF-MIF inter-county migration probabilities stratified by demographic characteristics to the same statistics estimated from the CPS. We conclude that the MAF-MIF is a valuable new source of internal migration data, especially for studies that require finer-scale spatial and temporal units, although the data set is not without limitations. The MAF-MIF provides benefits to the U.S. Census Bureau by evaluating the quality of the Master Address File Auxiliary Reference File (MARF) as a measure of residential mobility and producing a new data set for research on internal migration (Criteria 2). We also prepared novel estimates of annual county-level residential mobility disaggregated by age group, sex, and race/ethnicity (Criteria 11).