A Method for Indirectly Estimating Historical U.S. Migration Patterns: 1850-1880

Jani S. Little, University of Colorado at Boulder
Andrei Rogers, University of Colorado at Boulder

How can standard census variables provide inter-regional migration flows in the absence of a direct migration question? Three different methods are developed and compared. The first is a log-linear model and the second is a regression model. Both employ commonly available census variables as predictors. The preferred method is the third method, which uses a mixture of log-linear modeling, the method of offsets, and basic regression modeling. This method requires fewer assumptions than other methods available at present. Applied to a period in U.S. history that faced great demographic upheaval, the method is validated by the migration data recently released in the North Atlantic Population Project (NAPP) linked dataset, which tracked a representative sample of the U.S. 1880 population and found their places of residence ten years earlier, in 1870.