Existing research examining the self-selection of immigrants suffers from a lack of information on the immigrants' labor force activities in the home country, quotas limiting who is allowed to enter the destination country, and non-economic factors such as internal civil strife in the home country. Using a novel data set from the Federated States of Micronesia (FSM), I analyze a migration flow to the U.S. that suffers from none of these problems. I find that high-skilled workers (relative to the home country skill distribution) are the most likely to migrate from the FSM to the U.S. and that their behavior is explained mainly by the difference in average wages for their skill group. This finding suggests that previous immigration studies have overemphasized the role played by differences in the distributions of countries' wages and skills. Including information on the immigrants' characteristics prior to migration is central to my analysis, which highlights the importance of datasets that contain both home and destination country data on immigrants. Given the home country information, I use weather shocks to predict the probability of outmigration, which overcome the usual endogeneity problems in determining self-selection of immigrants. Second, I conduct nearest neighbor matching for immigrants prior to their leaving the home country using home country wages as the outcome variable to determine the nature of selection on unobservable characteristics.