The American Community Survey's Interstate Migration Data: Strategies for Smoothing Irregular Age Patterns

James Raymer, University of Southampton
Andrei Rogers, University of Colorado at Boulder

Because migrations are relatively rare events, age- and origin-destination-specific flows obtained from population samples often contain irregularities. Bias in analyses of migration flows can arise if these irregularities are not corrected for. In this paper, we present some typical examples of age-specific migration flows with irregular patterns, using the recently released American Community Survey (ACS) data. Strategies for smoothing these patterns based on the multiexponential model migration schedule and the categorical log-linear model are presented and compared. Before smoothing the irregularities found in the ACS data, our ideas are first tested on age-specific interstate migration in the U.S. West Region during 1995-2000 using the 5% Public Use Microdata Sample (PUMS). The corresponding full sample Census data are used to assess the smoothed patterns. Our results demonstrate that more accurate migration data can be provided by smoothing the irregularities caused by relatively small samples.