In this paper we consider a model for international tourism demand. The point of departure of the analysis is a utility function that is both dynamic and stochastic. In the model the stochastic component is interpreted as random changes in preferences for goods and services, while the dynamic component can be seen as either habit formation or as interdependent preferences. The resulting demand functions are estimated as a multivariate state space model, where the stochastic components enter the model as stochastic seasonal and trend components. An application is constructed for different segments of the Swedish tourism market. The results indicate the importance of including both dynamic and stochastic components in the utility function, and the importance of using disaggregate data to enable investigation of each market segment.