Abstract: This study demonstrates that migration measures predicated on cross-boundary movement introduce significant misclassification error into both migrant and non-migrant samples. The former is shown to occur when across-boundary moves cover only relatively short distances, while the later results from long distance residential movement occurring within the boundary. Using a unique sample of geocoded data from the National Longitudinal Survey of Youth 1979, I find that over half of all intercounty moves and one-third of all interstate moves cover distances of less than fifty miles. In contrast, more than half of all within-state moves (of at least ten miles) cover distances…of more than fifty miles while over thirty-five percent cover at least one hundred miles. Estimation results from a standard discrete-choice model of migration propensity are shown to be highly sensitive to distance suggesting that inferences based on a particular migration definition will often not generalize across alternative measures.
Show more