Three Essays on Migration, Occupational Sorting, and Degree Choice: Analyses of Spatial Autocorrelation, Income and the Race Wage Gap
Abstract
This dissertation consists of three essays concerning migration, occupational sorting, and college major choice. They each examine income, Chapter 1 concerning aggregate income at the county level, while Chapters 2 and 3 examine income at the individual level. Chapter 1 investigates the effects of migration on income. Using migration data from the U.S. Census Bureau and income data from the IRS, the chapter examines adjusted gross income in a county and the effect that migration has on it. It is shown that spatial dependence is present in the data, thus spatial models are applied to the data. A pooled panel spatial Durbin model is used and it is shown that this is the more appropriate of the models considered. The conclusion is that migration positively effects the county experiencing the in-migration, but negatively effects the neighboring counties. The negative effect on the neighboring counties requires further investigation. In Chapter 2, the race wage gap for high school graduates is studied to determine the effect of occupational sorting. The American Community Survey 1-year PUMS data over the years 2005 to 2012 are used. Sub-samples of 20,000 high school graduates are drawn without restrictions from each of white and black high school graduates for control groups. Using Bayesian methods, marginal posteriors of the parameters are drawn for a quartic specification of Human Capital Earnings Function (HCEF) for each control group. Then, 20,000 white high school graduates are drawn using the black workers’ occupational probabilities as the treatment group. The treatment group’s HCEF, along with a posterior predictive distribution generated by the marginal posterior draws, are compared to the control groups. It is shown that occupational sorting accounts for approximately 39 percent of the race wage gap seen in the data, giving credence that occupational sorting as groups exacerbates the race wage gap. For Chapter 3, the focus shifts from occupational sorting to college major choice. The race wage gap of college graduates with a bachelor’s degree as their terminal degree and the effect of college major choice for differing races is examined. The data used in this chapter is the ACS 1-year PUMS from the years 2010 to 2016. Differing groupings of college majors are analyzed by first drawing control groups from white, black and Hispanic college graduates, then drawing treatment groups based on probabilities of college majors from another race/ethnicity. Further, once the quartic specification of the HCEF is estimated using Bayesian methods, the Oaxaca-Blinder decomposition is used to evaluate the differences between control and treatment groups. First, all college majors in the data were examined, and no effect on the race wage gap was found. STEM degrees also show no effect on the race wage gap. Business and non-business degrees show competing effects. Non-business degrees actually showed that if white graduates chose college majors like black or Hispanic graduates that their earnings would increase, while in business majors white graduates’ earnings decrease when they choose business majors like black and Hispanic graduates. The OaxacaBlinder decomposition showed that all of this negative effect was due to the treatment effect while the composition effect was positive.