Mexican immigrants to the United States have better reproductive outcomes than do U.S.-born non-Latina whites. Explanations offered for this "epidemiologic paradox" include (1) poor outcomes among Mexican women may be hidden by their return to Mexico; (2) Mexican women may have a higher fetal death rate that alters the pattern of live birth outcomes; (3) Mexican women may have socioeconomic characteristics which, if properly measured, would explain the outcome; (4) Mexican women may have personal characteristics which would explain the outcome, if properly measured; (5) there may be ameliorative or salutogenic "protective" effects of culture; and (6) migration may be selective of healthier women who are thus more prone to positive outcomes. We test these explanations, with an emphasis on the last one, using a data set that combines reproductive histories and birth outcomes for Mexico-born women delivering in San Diego, California and Mexican women delivering in Tijuana, Mexico. These data are compared with U.S.-born Latinas and U.S.-born non-Latina Whites. Multivariate logistic regression analysis suggests that when controlling for birth history (stillbirths and miscarriages), socioeconomic characteristics (education and prenatal visits), personal characteristics (age, parity, time in area, history of family problems), and health characteristics (history of smoking, alcohol use, drug use, anemia, vaginal bleeding, urinary infection), the adjusted odds of a positive birth outcome (measured as a live birth of 2500 grams or more) is highest for women delivering in Tijuana, implying that migrants may not be so selective when compared to the country of origin. The number of prenatal visits was an important explanatory variable.