simPop

simPop is an R package that turns a household survey into a synthetic population: a complete, invented set of people and households that keeps the statistical shape of the survey and of known census totals.^2,3 Researchers use this kind of file when they want to run an agent-based or micro-level model of a country or region but cannot release, or do not have, a record for every resident.³

The main authors are Matthias Templ, Alexander Kowarik, Bernhard Meindl and Andreas Alfons; the code is kept on GitHub under the statistikat account.^1,2 The International Household Survey Network (IHSN) and the World Bank's Development Data Group backed its development to encourage the use of existing survey and census microdata for agent-based simulation, funded through a DFID trust fund administered by the World Bank.^2,3 The IHSN describes synthetic data as largely solving the risk of disclosing real respondents, and as a way to fuse data from several sources.³

The package combines model-based methods, calibration and combinatorial optimisation.^2,3 You install it from CRAN and work in R.² The method is described in a 2017 paper in the Journal of Statistical Software by Templ, Meindl, Kowarik and Olivier Dupriez.⁴

It is licensed GPL-2 or GPL-3.² The CRAN release is 2.1.3 from January 2024; the GitHub repository carries a slightly newer 2.1.4 and its last commit was in January 2026, so it is maintained but quiet.^1,2 The IHSN project page marks the World Bank project itself as closed, with follow-up activities.³