Three applications of propensity score matching in microeconomics and corporate finance: United States internal migration; seasoned equity offerings; attrition in a randomized experiment.

摘要

Propensity score matching is becoming an increasingly popular tool in economics for dealing with selection issues. Recently economists have been evaluating matching methods by trying to mimic the experimental evidence in social experiments where individuals were randomly assigned to job training. There is mixed evidence on whether matching can mimic the experiment results. This dissertation applies the propensity score matching on two non-experimental settings, US internal migration and seasoned equity offerings, and one social experiment outside of job training, the RAND Health Insurance Experiment. These applications generate new empirical evidence on these important issues and also shed light on the practical strengths and weaknesses of matching methods.; The first essay evaluates the effect of migration on wage growth. Using a distance-based measure of migration, the study finds a significant positive effect for college graduates and a marginally significant negative effect for high school dropouts. The study does not find any significant effect for other educational groups or for the overall sample. The results suggest that the better measure of migration matters.; There is mixed evidence on long-run stock performance after Seasoned Equity Offerings (SEOs). Some studies have found that firms issuing SEOs show substantial long-run underperformance, which poses one of the strongest challenges to the Efficient Market Hypothesis. The second essay shows that the long-run abnormal returns after SEOs found in previous studies are an artifact of the inability of traditional matching techniques to control properly for all risk factors. Propensity score matching allows matching on multiple firm characteristics simultaneously. This essay finds that SEO firms do not behave differently from similar firms that are small in size, have low book-to-market ratios and high recent returns.; Attrition has long attracted the attention of econometricians. The third essay uses data from the RAND Health Experiment to explore using matching to address attrition. Specifically, I introduce attrition into the experimental data and then use matching to deal with the problem, creating a new way of using post-baseline data in the matching procedure. I find that matching performs reasonably well, but its performance does depend on the treatment effect estimated.