Family migration is a joint function of individual-, family-, and contextual-level effects. The first part of this dissertation develops a multilevel theoretical framework for family migration decision-making. This framework emphasizes an integration of individual-, family-, and contextual-level effects, incorporates a longitudinal perspective-human migration history with both economic and non-economic effects, and acknowledges the family as a decision-making unit of migration analysis. The second part of this dissertation introduces multilevel logit models, which deal especially with hierarchical data structures and yield more accurate statistical conclusions, compared to conventional linear logit models, and explores the impact of individual-, family-, and neighborhood-level factors on family migration. The estimation methodology in this dissertation is motivated by the theoretical framework and is new to the study of family migration.;The main data source used is the Panel Study of Income Dynamics (PSID). The PSID is a longitudinal survey that is nationally representative of families in the United States in the civilian noninstitutionalized population.;There are three main empirical conclusions of this dissertation. First, the individual- and family-level effects display patterns consistent with the theoretical hypotheses and play a much more important role in family migration decisions than do the characteristics at the neighborhood-level. Individual-level factors include husband's race, age, and education. Family-level factors include family income, the earnings difference between husband and wife, number of children, home ownership, and migration history. Second, some evidence supports neighborhood-level effects on family migration, but they are of only secondary importance to the individual- and family-level effects. Third, the findings support the nested structure of family migration. Multilevel analysis is an important research approach to generate a more complete understanding of the phenomenon under study. Because this study considers the clustering structure of the data, the explanatory power of the empirical model is improved.