A0598
Title: A discrete-time mover and stayer model with time-varying covariates for studying Italian student mobility
Authors: Martina Vittorietti - Delft University of Technology (Netherlands) [presenting]
Eni Musta - University of Amsterdam (Netherlands)
Abstract: The mover-stayer model, also known as the "cure model", is a model to study social change over time in a heterogeneous population. It extends the first-order Markov process by recognizing a subgroup of individuals called "stayers," who will not experience the target event. Traditionally, the probability of being a stayer is the same for all individuals and corresponds to the proportion of stayers in any given state. However, this probability can be influenced by fixed-time covariates and time-varying covariates. The inclusion of the latter is not trivial. When subjects are assessed periodically over a certain period, panel data is dealt with. Panel data are common in the education field. A new dynamic version of the discrete mover-stayer model is presented using multinomial logistic regression with time-varying covariates, specifically focusing on panel data on the mobility of Italian Master's students. Factors such as the students' undergraduate degrees and the rankings of their universities are considered time-varying covariates in modeling their probability of moving; sex and age at enrollment as fixed-time covariates. The model employs a maximum likelihood estimation approach, integrating both the multinomial responses and the stayer status. Its identifiability is evaluated, simulation studies are conducted, and its performance is compared against established models in the literature.
