The first chapter of this dissertation examines the phenomenon of labor market segregation. Using a regression discontinuity (RD) design, I exploit the variation in baseyear minority shares across single-establishment firms to document the dynamics of establishment-level segregation in two five-year intervals: 1995-2000 and 2000-2005. Using the Longitudinal Employer-Household Dynamics (LEHD) infrastructure files, I first show that systematic establishment-level segregation still exists in all industries. Then, I show that the dynamics of segregation among these single-establishment firms are nonlinear and exhibit "tipping" patterns in both five-year intervals, although the magnitude is much larger in the earlier time period. The observed tipping pattern is primarily driven by non-Hispanic whites leaving. The effect due to minorities entering is much smaller. Alternative explanations such as non-linear changes in establishment characteristics or omitted variables do not explain the observed changes in minority shares. Finally, I find that, unlike the 1995-2000 period, during which tipping behavior seems to have been driven equally by blacks and Hispanics, Hispanics are the sole driving force in the 20002005 period. Overall, this chapter provides the first suggestive evidence that the dynamics of establishment-level segregation are highly nonlinear and exhibit a tipping pattern. The second chapter of the dissertation describes the technical linking process and examines the properties and the qualities of the crosswalk files. The crosswalk between the Longitudinal Employer-Household Dynamics (LEHD) infrastructure file system and the Census Business Register (BR) is authorized as part of the LEHD Infrastructure Project. This document describes the LEHD - BR crosswalk and its component inputs: the Business Register, Longitudinal Business Database (LBD), and the LEHD Infrastructure File system. The output files include the LEHD - BR crosswalk at both the establishment and employer levels. These output files can facilitate linking a wide range of contextual variables relating to characteristics of the current and prior employers and co-workers of current employees. Match and non-match rates for various populations are defined and estimated in order to examine the properties and quality of the LEHD - BR crosswalk output files. The third chapter of this dissertation exploits plausibly exogenous changes in family size caused by the initial implementation and subsequent relaxations in China's One Child Policy to estimate the causal effect of family size on educational attainment. I find that the average family size has decreased substantially since the One Child Policy implementation. By employing an Instrumental Variable estimation strategy, I find clear evidence indicating that there is indeed a negative trade-off between child's quantity and quality in urban China. An additional child can lead to a decrease of 1.2 years of schooling. A simple back-of-the-envelope calculation reveals that the implementation of the One Child Policy has significantly increased the average completed years of schooling by approximately 0.68 years in urban China. This effect is in fact larger for women than for men. No negative trade-off effect is found for the rural households in the sample.