MIGRATION, EARNINGS AND SELECTIVITY: FROM THEORY TO FACT -- EVIDENCE FROM TURKEY, 1963-73 (GEOGRAPHIC MOBILITY)

This study examines the links between geographic mobility and earnings from a theoretical and empirical perspective. A dynamic programming formulation of the migration decision that allows for multiple moves constitutes the theoretical model. Individuals compare location-specific earnings trajectories and choose that alternative with the best life-time prospects. The statistical model takes this into account by positing a joint structure incorporating the migration decision and the earnings determination process. The discrete choice formulation distinguishes between stayers and movers in the simple model, and between stayers, one-time movers and frequent movers in the extended model. Measures facilitating earnings comparisons between the various categories in a manner consistent with the optimizing framework of a theoretical model are proposed.;Although support for the theoretical premises is not uniform across the various earnings equation specifications, the proposed measures of the migration effect provide new insights in reconciling theory with fact. It is seen that selectivity figures in an important way in the earnings comparisons. Differences between individuals with identical values of the observed variables in the earnings equations are shown to exist, due to heterogeneity in the costs of migration. Both the average migrant and the average non-migrant are found to realize positive returns from their decision, at least over some range of the moving costs. Another significant finding--which is likely to be of interest in applications in other fields--is that the migration effect definition used in the earlier literature (neglecting selectivity) does indeed capture the earnings differential in a conceptually sound manner, albeit at a single point.;The empirical investigation is carred out using a rich micro data set from Turkey. In the simple model, strong evidence of selectivity in the earnings of movers is found. In the extended model, the findings do not support the simultaneous formulation of the re-migration decision very well. This suggests that information acquired after the initial move might be more important in setting frequent movers apart from one-time movers, a conjecture which could not be directly tested here.