This paper introduces a model-based framework for generating close-to-reality synthetic population data for household surveys, addressing the need for realistic micro-level data that can be shared without exposing real respondents. The approach combines calibrated resampling of the survey’s structural variables with model-based simulation of the remaining characteristics, so the synthetic population reproduces the joint distributions and dependencies of the source data. It is demonstrated on the European Union Statistics on Income and Living Conditions (EU-SILC), a setting with complex household structure and a multi-stage sampling design. The methodology became the statistical foundation later implemented in the simPop R package and remains a reference point for synthetic-population generation in official statistics.