Duration Analysis provides a qualified class of statistical models for examining the influential factors on unemployment spells.
This thesis is an attempt to introduce the field of duration analysis by a non-parametric, a parametric and a semi-parametric model approach. To demonstrate the importance of competing risks for unemployment duration, cumulative incidence functions are calculated besides differentiated Kaplan-Meier estimations. A log-logistic model is chosen for the demonstration of parametric covariate effects on the unemployment duration of individuals who became reemployed. Another model approach to detect unemployment determinants is given by calculation of Cox cause-specific hazards as well as subdistributional hazards in competing risks regressions. An application of the models is presented for a sample set from the Austrian micro-census database of the year 2005.