Historically, the United States have been the meeting point for people, ideas and knowledge from all over the world. The first two chapters of my dissertation focus on the composition and impact of the U.S. immigrants; while the last chapter describes a new way for U.S. universities to export education worldwide, analyzing the case of a Mass Open Online Course. In Chapter 1, I study how visa policies may affect the immigrant distribution in terms of both observable skills, as education, and unobservable traits, as ability or motivation. I develop a model of migration choice, estimated via simulated method of moments using the New Immigrant Survey dataset. I find that wage differentials, combined with skill transferability, act as the main force driving self-selection. Positive self-selection on unobservables is stronger among non-college graduates. The dynamic framework gives the opportunity to simulate the immigrant distribution resulting from alternative visa policies. In this work, I apply the visa-point system proposed by the RAISE Act to the model. The results suggest that the scheme is successful in attracting migrants with the desired observable characteristics but it reduces the positive selection on unobservable skills. In Chapter 2, I use a firm-level database on high-skill temporary visa (H-1B) applications to study the effects of H-1B workers on patents. To address endogeneity, I construct an instrumental variable by interacting H-1B quotas with the number of Asian incumbent inventors within the firm. An increase in H-1B applications has a positive impact in terms of firm patent activity, future new inventors and patent activity by incumbents. It also leads to a higher diversity in citations by country and class, as evidence of positive spillovers. Chapter 3 (joint with Rebecca Stein) describes our Mass Open Online Course in microeconomics principles. Using the Cox proportional hazard model, we relate the high attrition rate to demographic data, finding that younger students, U.S. participants and females are less likely to complete the course. The results provide useful insights for future MOOCs: more flexibility may benefit time-constrained students but it is important to keep the structure and sense of community, both instrumental to completion.