A multilevel model of U.S. internal migration

Date of Completion

January 2004


American Studies|Geography




The social science literature on internal migration presumes people and place characteristics are interdependent determinants. Unfortunately, single-level multiple regression models are unable to decompose the relative contributions of people and place characteristics on internal migration. A multilevel model of internal migration in the United States between 1985 and 1990 investigates how labor market area characteristics moderate the effect of household determinants of internal migration. The primary source of data for the response variable as well as the micro-level explanatory variables is the Public Use Microdata Sample L (PUMS-L) (Census of Population and Housing, 1990). Specification of a multilevel logistic regression includes determinants of internal migration at the household-level while the labor market area-level includes proxies for push and pull factors of internal migration such as cultural and recreational opportunities, natural amenities, income and cost of living. Two phases of multilevel analysis for destinations and origins reveals that people and place characteristics are interdependent determinants of internal migration. The results also indicate that households do not trade economic gain for greater quality of life. ^