| One objective in the analysis of longitudinal data is the
estimation of state-to-state transitions over time. An important advantage
of using panel data is the ability to directly observe individual-level
transitions. The basic approach is the turnover table that tabulates responses
at one wave against the responses at another wave. A limitation of panel
surveys is that they often encounter selective attrition and that it is
generally difficult to update the panel to ensure representativeness. Moreover,
for many research issues panel data are unavailable. In the last several
years, various econometrics methods have been proposed to estimate dynamic
models with repeated cross section data. The strength of a repeated cross-sectional
survey is that it selects a new sample at each time point, so that each
survey is based on a probability sample of the population. The lecture focuses
on a dynamic Markov model for the estimation of state-to-state transition
probabilities from independent cross-sectional samples. It discusses model
specification, parameter estimation and an empirical application.
powerpoint presentation Eisinga
|