|
CQE

Lecture by Ralf Münnich, Trier University

Economic Research Seminar
Wednesday, 18. May 2022 - 16:15, DPL 23.110, Domplatz 23

Regionalized Dynamic Microsimulations - An Introduction to the MikroSim Model

Abstract: Spatial dynamic microsimulations represent a powerful methodological tool for analyzing complex phenomena in populations considering individual and regional dependencies by modelling socioeconomic systems based on micro-level units such as individuals, households, or institutional entities. The basic idea goes back to Orcutt (1957) who criticized that it is not possible to aggregate the heterogeneity of individuals and the complexity of mutual influences on a higher level. Therefore, two statistical components are crucial for conducting a microsimulation study: a microdata set as the base population and appropriate methods for estimating the transitions. In most cases, survey data is used as base population due to its availability and a large number of observed variables. For regional analyses, however, these data can only be used to a very limited extent because of the small regional sample sizes and lack of geographical information. It is, therefore, evident to create and expand a suitable database. Taking into account individual and family relationships, the population is updated as a discrete-time dynamic microsimulation at annual intervals. Model-based transition probabilities regulate the occurrence of certain events, such as births, partnerships, and regional mobility. The given setting urges the need of using many different statistical methods including estimation and prediction of transition probabilities and state changes. The challenges and opportunities of microsimulations are presented on the example of the MikroSim model developed within the framework of the DFG research group FOR2559. The aim of the project is to simulate consistently the approx. 82 million citizens that are living in about 40 million households considering their regional distributions in order to enable the analysis of social developments and their effects at a small-scale level. For this purpose, a base dataset is generated using anonymized distributions of the German register and micro-level survey data. Additionally, persons and households are geographically located within 100 × 100 meter units from census grid cells. With this statistically enriched dataset, socio-demographic changes of the underlying population are simulated by estimating transition probabilities considering empirical data.  This approach leads to the development of analysis models for social science issues and empirically-based, long-term planning. The simulation model is designed to cover a wide range of substantial areas (sectors).