For the alignment of different employment histories the method of optimal matching analysis (OMA) is often used. In contrast to event history analysis, applying OMA allows to analyze different events simultaneously. Furthermore, OMA allows including situations in which the same event appears several times as well as situations in which the event does not appear at all. The aim of OMA is to describe sequences of events over time. In terms of job histories this could include states such as unemployment, low payment, and temporary work.
Combining OMA with clustering techniques provides a number of quite homogeneous clusters. The chosen number of clusters is higher than in traditional applications to increase the internal homogeneity of single clusters. Although the number of clusters may exceed 30 clusters or more, not all single sequences will fit well into one of these clusters. These outlying sequences increase the internal heterogeneity of their best fitting clusters to a relative large amount. Furthermore, not all clusters are meaningful from a substantive point of view. To solve this problem a further classification step will be done according to some substantive criteria resulting in an even higher number of clusters. These clusters can finally be used to analyze different subgroups of the data.
In this paper we discuss the possibilities of using OMA in combination with cluster techniques. As an empirical example we use the German IAB Employment Samples (IABS) of 2004 and analyze a three-year period of young people entering the labor market after completing vocational training.