This study examines factors predicting participation in study abroad programs using a bivariate probit selection model where the probabilities of attending university and studying abroad are jointly estimated. Given that unobserved variables influencing these two sequential events are likely to be correlated, this model provides estimates that are less contaminated by misspecification bias compared to those from a conventional univariate probit model of study abroad participation. Using data from an Italian nationally representative survey, the results emphasise the importance of controlling for unobservables. They indicate that the relationship between type of upper secondary school and study abroad participation disappears once selectivity bias is accounted for.