An Analysis of U.S. Domestic Migration via Subset-stable Measures of Administrative Data
63 Pages Posted: 30 Jun 2020 Last revised: 2 Jan 2021
Date Written: June 3, 2020
How does the likelihood of moving across U.S. regions vary with changes in statuses such as income, employment status, local and federal tax payments, number of children, or marital status; and how does the risk of a change in status vary given a move? Statistics for these and other relative risks are calculated via pseudo-experiments run on almost all people who earned formal market income in the U.S., 2001-2015, totaling about 1.6 billion households with 76 million long-distance moves. The Cochran-Mantel-Haenszel (CMH) statistic is adapted to this big data context. The key theoretical result of this paper shows that the CMH statistic is the unique aggregate risk ratio that satisfies some basic desiderata, notably "subset stability": if a statistic has value s₁ for one subset and s₂ for another, then the statistic for the union of the two sets is between s₁ and s₂. Using the CMH statistic and controlling for 13 household characteristics, moving households are more likely to see a drop in income than a rise, relative to the counterfactual of staying. The hypothesis that they are more likely to move to areas with lower local taxes than areas with higher taxes is not supported. Movers are more dynamic than stayers on almost all measures, and these effects persist over a decade past the move. Even federal tax liability shows greater volatility, with or without controls for changes in housing status, income, and so on. With such wide diversity in dynamics given moving, there can be no single population-level explanation for why households move.
Keywords: migration, administrative records, demographic analysis, taxrevenue, relative risk, risk ratios, returns to education, retirement, tax policy
JEL Classification: J61, C14, H24, D19
Suggested Citation: Suggested Citation