Abstract

Microsimulation is a technique to answer many what-if-questions using microdata that, otherwise, can hardly be analysed. Li and O’Donoghue (2013) emphasized that microsimulations cover two areas, the policy support per se as well as the generation of synthetic data generation as an important base for performing microsimulations. Especially in regional dynamic microsimulations a sound data base will have to be generated using many variables preferably on all microunits in the target population.

More and more methods such as data integration of different surveys and data types, prediction methods, as well as modern ML approaches are applied. However, modelling strategies need to be adjusted accordingly, in particular depending on cross-sectional or longitudinal applications. Further, the increasing attention is laid on the granularity of the modelling.

All in all, little attention is paid to the accuracy of the generated data as well to assumptions and implicit decisions of developers of microsimulation models. The presentation will focus on different aspects of synthetic data generation and so-called digital twins. Special attention will be laid on the impact of using multiple data sets including survey, nonprobability, and other data in the entire modelling process.