USING LINKED EMPLOYER-EMPLOYEE DATA TO UNDERSTAND LABOR MARKETS AND IMPROVE DATA PRODUCTS

This thesis is comprised of three chapters. The first chapter (joint with John Haltiwanger, Julia Lane, and Kevin McKinney) explores a new way of capturing dynamics: following clusters of workers as they move across administrative entities. Information on firm dynamics is critical to understanding economic activity, yet fundamentally difficult to measure. The worker flow approach is shown to improve linkages across firms in longitudinal business databases. The approach also provides conceptual insights into the changing structure of businesses and employer-employee relationships. Many worker-cluster flows involve changes in industry -- particularly movements into and out of personnel supply firms. Another finding, that a nontrivial fraction of firm entry is associated with such flows, suggests that a path for firm entry is a group of workers at an existing firm starting a new firm. The second chapter makes use of linked employer-employee data from the U.S. Census Bureau's Longitudinal Employer-Household Dynamics (LEHD) Program and matches it to data on business acquisitions from the Federal Trade Commission to examine labor market outcomes of employees at firms undergoing mergers. Earnings and employment can be observed over time for workers at both the acquired firm and the acquiring firm. The findings suggest that while wages tend to be about the same or higher for workers at these restructuring firms, turnover is significantly higher, and the costs of job-loss are large and long lasting. The third chapter (joint with John Abowd and Martha Stinson) provides technical documentation for a project undertaken by the US Census Bureau, the Social Security Administration, and the Internal Revenue Service to explore a potential method of providing the public a valuable new dataset without compromising confidentiality. The underlying database was created by merging the respondents from the Census' own SIPP with administrative data on earnings and benefits from the IRS and SSA. The administrative variables combined with the detailed survey responses from the SIPP offer the potential to do interesting research especially in the areas of retirement, benefits, and lifetime earnings; however, they also add extensive new information for malicious data users to potentially reidentify SIPP respondents. This final chapter develops a cutting edge new technique for providing a micro-dataset that looks, in structure, just like the underlying confidential data. This "partially synthetic" database aims to preserve as many of the complex covariate relationships in the confidential data without posing any significant new risk to disclosure protection.