A dynamic microsimulation model for forecasting educational patterns is presented. At the level of individuals the model simulates lifetime educational behavior, resulting in a long term forecast of the general educational level in Denmark. The model is a light-weight, dynamic, multithreaded and closed microsimulation model using discrete time.
Data on the full Danish population is used as the initial population. Each individual is characterized by age, gender, origin, educational attainment and current educational status. Future demographic events such as births, deaths, immigration and emigration are projected in a separate group-based model and given as input. In the model individuals lives their life?s independently to decrease time-complexity and to utilize the potential of the multithreaded environment.
Transition probabilities are calculated from historical educational behavior using Danish register data. The historical observations are linked to a range of background variables (such as gender, age, origin, current participation in education, study length and educational attainment). Prior to running the model, transition probabilities are computed using conditional inference trees. This data-mining approach groups together observations with similar characteristics and responses based on statistical tests.
This paper describes the features of the model, briefly presents some results and points to the potential of the model in terms of policy analysis and already planned extensions to the model.