ÿþZastosowanie modelu mover-stayer dla danych dotyczcych kredytobiorców indywidualnych
[Application of a discrete time mover-stayer model for the credit data]
Anna Matuszyk (ZakBad Systemu Finansowego, SGH, Warszawa)
A discrete time mover-stayer model is considered in which stayer probability is
specified as the logistic function of the time fixed covariates. The transition matrix
of the movers is not parametrized. The model is estimated from panel data using the
EM algorithm. The proposed model is illustrated with the data on installment loans.
For a lending institution, it may be of interest to not only estimate the number of
stayers in a given state, but also to predict if a new loan applicant is going to be
a stayer or a mover based on her/his time fixed covariates. Using the proposed M-S model,
it is explored how such predictions can be made for loans data with the covariates.