(From: Didier Blanchet, Sophie Buffeteau, Emmanuelle Crenner and Sylvie Le Minez, 2009 ; IMA conference paper)

The Destinie model has been developed and maintained at the French national statistical institute (INSEE) since the mid-1990s. A new version is progressively becoming operational and this paper presents the main options that have been retained for this new model. The main goals of this new version have been the improvement of consistency with official demographic projections, the incorporation of more complex labour market trajectories, and increasing portability and facility to use. The new model makes a complete separation between two components. The first component is a generator of demographic and employment biographies, whose results are then stored in two intermediate output files. The second component is a library of subroutines allowing an easy programming of pension projections based on these two output files resulting from the first stage.

Related Information

Main strengths/weaknesses of the initial model Improvements or corrections brought by Destinie 2
Strength: closed population approach, allowing a simple simulation of family and kinship ties. Unchanged
Strength: separate simulation of the most important pension schemes (CNAV, ARRCO, AGIRC, Civil Servants) Maintained, with the additional possibility ofsimulating heterogeneous careers (mobility betweensectors and part/full-time.)
Strength/weakness: a behavioral module for simulating retirement behavior (Stock and Wise model). Better than nothing, but many doubts have been expressed about the empirical relevance of this module. The existing module has been maintained, but is onlyone among several possibilities for simulatingretirement behavior, with improved facilities forchanging preference parameters.
Weaknesses: heaviness, execution time, entry costs, lack of portability/versatility. The model has been split in two subcomponents:

1. Demographic and labor market events are simulatedfirst, with results stored in intermediate files.

2. Pension projections (or other applications) are madewith small ad hoc programs, based on relativelysynthetic and easy to use libraries that use these files asinputs.

Weakness: Insufficient consistency with macro projections Automatic adjustments of transitions probabilities tosatisfy some predefined macro-targets
Weakness: short run behavior Initial conditions are partially imputed to avoidinconsistencies that can exist between initialobservations and projected values. This also facilitatesthe simulation of counterfactual initial conditions (e.g."what would have been the starting point in 2003 if noreform had taken place in 1993?")
Scroll to top