This study aims to identify factors affecting passenger choice of airport ground transportation modes using aggregate data. Incheon International Airport (ICN), known as the gateway of Korea, is connected to several major Korean cities by high-speed trains and intercity buses. The data we received from ICN records the number of passengers on each train and bus, the scheduled and actual arrival times for each flight and the number of passengers on each fight. A limitation of the data is that it does not identify individual passenger choice. The challenge is that, without individual-level data, classical methods used to estimate consumer choice, such as the multinomial logit model, cannot be applied. Therefore, we develop a structural estimation method adaptable to this data. Our method estimates the ridership of each train and bus based on a structural discrete choice model. This approach essentially converts unobserved individuals’ choice probability to the observable ridership. The model coefficients are computed by the nonlinear least-squares estimation which minimizes the distance between estimated and observed ridership. The estimation results suggest that passengers who fly Korean airlines or have experienced longer waiting times are less likely to take public transportation, and passengers who fly low-cost airlines are more likely to take public transport. This approach allows counterfactual analysis to estimate how ridership of airport ground transport services would be if the schedules were adjusted. In current practice, the train schedules are not efficiently matched with the intraday demand nor passengers' choice. Thus, we formulate a nonlinear program with estimated parameters to maximize the train ridership. The proposed schedule could increase the train ridership by 16.67% without adding extra number of trains.