Vacation length choice: A dynamic mixed multinomial logit model

•Panel data used to develop a dynamic mixed logit model for vacation length of stay.•Individual heterogeneity and state dependency are accounted in the model.•Independent variables include lifecycle stages, month and lagged variables.•Long holidays are most strongly affected by trips made in previous years.•Differences between lifecycle reflect flexibility/constraints typical of each group.

This paper uses panel data to develop and estimate a dynamic model of choice of the length of stay of a vacation, controlling for unobserved heterogeneity and state dependency. Length of stay options vary from short (1–3 nights), medium (4–9 nights) to long vacations (10 nights or more) and the decision not to go on vacation in a particular year. Independent variables include family lifecycle stage, income, month and lags of the dependent variable. Results indicate that long holidays are most strongly affected by trips made previously in the same year than medium and short vacations. In contrast, there is an increased need for a vacation when any medium or long trips were not yet made in the current year. Month-specific variables confirm that respondents have preferences for making leisure trips during the main holidays and warm seasons. The observed differences given the various lifecycle stages reflect imposed constraints given age and/or household composition that are typical of each particular group.