This paper uses the random-coefficients logit methodology that controls for potential endogeneity of prices and allows for general substitution patterns to estimate various demand systems. The estimation takes advantage of an original ticket-level revealed preference data set on travels from the New York City area to Toronto that contains prices and characteristics of not only flight choices but also of all non-booked alternative flights. Consistent with having higher valuations, our results show that travelers buying closer to departure have a higher utility of flying. Moreover, consumers' heterogeneity decreases as the flight date nears. At the carrier level, we identify which carriers have more price-sensitive consumers and which carriers face greater competition. In addition, the results suggest that our multi-airport metropolitan area can be considered as a single market and that JFK and Newark are relatively closer substitutes. Overall, consumers are more willing to switch to alternative carriers than between airports or departure times.