Date of Graduation


Document Type


Degree Type



Chambers College of Business and Economics



Committee Chair

Brian J. Cushing.


This dissertation consists of three essays that examine two important roles of migration, promoting regional wage convergence and increasing economic efficiency. The first essay is a survey of the migration and migration-related literatures, examining why migration and regional wage differentials can co-exist over time. The direct reason is that the classical assumptions leading to regional wage convergence are rarely fully applicable in the real world. This essay reveals various real-world conditions violating those assumptions and explains how they can either create or magnify regional wage differentials. The essay also reveals how the difficulty in distinguishing wage differentials from true wage disequilibrium has posed serious measurement problems. The second essay examines how individuals' occupational choice affects their destination choice in a simultaneous equation framework. The results show that individuals are more likely to choose destinations where they can maximize the returns to their occupational skills. This essay also develops the application of a two-step maximum likelihood method to examine two simultaneous choices. Considering that individuals frequently face simultaneous choices, this methodology fills an increasingly important need. The third essay investigates whether violating the Independence from Irrelevant Alternatives (IIA) assumption, as standard discrete choice models do, poses serious estimation problems. It compares the results from nested logit and mixed logit models, which relax IIA, with the standard conditional logit model, which assumes IIA. The essay shows that their results, while statistically different, are qualitatively similar. Their parameter estimates are of the same sign, of comparable statistical significance, and of comparable magnitudes. Given the substantial computational cost of the more complex nested and mixed logit models, the finding that a well-specified, but computationally much simpler, conditional logit model may suffice is encouraging. The results also suggest that the growing body of conditional logit migration models may not be inherently flawed.