Sensitivity analysis for dynamic microsimulation models

Microsimulations usually contain various transition processes such as births and deaths, relocations, and change in household characteristics. The estimation, organisation, and implementation of these processes can have a substantial impact on the simulation outcomes. We propose to evaluate a dynamic discrete-time stochastic microsimulation using the concept of variance based sensitivity analysis which is well established for indicator assessment in survey statistics. This method is suitable for measuring and comparing the impact of different input factors on the output, even when analytical methods - given the complexity in a dynamic framework - cannot be applied.