Publications - Working Papers IER SAS

The quality of input microdata is one of the key factors influencing the reliability of population-based microsimulation models. Despite the extensive development of microsimulation modeling, the professional literature focuses primarily on the development of simulation algorithms, while the issue of the suitability of various sources of input microdata remains relatively under-explored. The aim of this study is to analyze the importance of input microdata in population microsimulation modeling and to experimentally assess the impact of three official statistical sources—the Labor Force Survey (LFS), the European Union Statistics on Income and Living Conditions (EU-SILC), and the population census—on the results of an identical population microsimulation model. The research is based on a controlled experimental design in which all components of the model—including simulation algorithms, parameters, demographic processes, the time horizon, and validation procedures—remain unchanged. The only experimentally varied variable is the source of the input microdata, or the parameters derived from these data. This approach allows us to isolate the impact of the input data set from other factors influencing the results of the microsimulation. The results show that the choice of input microdata has only a limited impact on aggregated population indicators, such as the total population or the number of births, with all analyzed databases yielding very similar long-term trends in population development. Conversely, in projections of the economically active population and its age and educational structure, differences between individual input databases gradually accumulate and lead to divergent results. The experiment confirmed that the sensitivity of the population microsimulation model to the choice of input microdata depends on the type of output indicator analyzed. The main contribution of the study is the proposal of an experimental methodological framework that allows for the systematic evaluation of the impact of different microdata sources while maintaining an identical microsimulation model structure. The results highlight the need to pay closer attention to the selection and validation of input microdata, particularly when modeling the structural characteristics of the population, and provide methodological recommendations for the development of population microsimulation models using official statistical data.