The Destinie 2 microsimulation model
The Destinie 2 model (modèle Démographique Économique et Social de Trajectoires INdividuelles sImulÉs, version 2) is a dynamic microsimulation model developed and maintained by INSEE, whose main objective is to provide long-term projections of the pension system. It is the successor to Destinie 1, the first microsimulation model developed in this field in France, since 2010.
Usage
This model is used in long-term projection exercises conducted by the Conseil d’Orientation des Retraites (COR) and the European Commission, as well as in research studies. The construction of a demographic sample, coupled with the establishment of family ties and career paths, gives the model a more general perspective that extends beyond pensions in the strict sense. It has thus been used for studies on dependency, health expenditure and wealth accumulation. Its ‘household’ dimension distinguishes it from other microsimulation models. Furthermore, the modest size of the sample allows for rapid simulation and thus the creation of numerous variants, at the cost of lower accuracy of results.
Methodology
The model is based on a pseudo-population representative of the French population as of 1 January 2018, constructed from the 2017-2018 Household and Finance and Consumption Survey (HFCS), which questioned approximately 13,000 households about their family ties and professional careers. This pseudo-population is obtained through procedures for re-imputing missing values or links, as well as through post-calibration duplication. It comprises 30,000 households.
The simulation is carried out at the individual level. The model renews the initial population by simulating births, deaths and migration flows. It then projects the professional careers, earned income and retirement of each individual.
Employed persons are divided into three main groups (private sector employees, civil servants and self-employed persons), as are pensioners (in as many basic schemes). This model simplifies a number of legislative elements (for example, the self-employed are simulated as all being subject to the same rules for calculating entitlements). Destinie 2 makes it possible to reconstruct an individual's entire career path (activity status and income) and simulates the settlement of their pension according to the rules in force for each scheme.
Three main modules
The model comprises three separate modules, written in C++ and interfaced in R: the demographic biography generator module, the career trajectory generator module, and the ‘retirement’ module.
The demographic trajectories are based on INSEE demographic projections published in 2021, covering the whole of France over the period 2021-2070. The number of deaths, births and migration flows are thus calibrated each year of the projection.
Individuals' career paths are observed until 2017, the base year. From 2018 onwards, their careers (employment status and income) are projected in accordance with calibration constraints based on macroeconomic assumptions. These assumptions relate to labour productivity gains and the unemployment rate, both taken from COR publications, and the activity rate, based on INSEE labour force projections.
The ‘retirement’ module simulates retirement and calculates the amount of associated entitlements. These entitlements include direct entitlements and survivor's pensions from basic schemes (the general social security scheme, the Service des Retraites de l’État (SRE), the Caisse nationale de retraites des agents des collectivités locales (CNRACL), Sécurité Sociale pour les Indépendants (SSI)) and certain supplementary schemes (Association générale des institutions de retraite complémentaire des cadres-Association pour le régime de retraite complémentaire des salariés (Agirc-Arrco)). In addition to survivor's benefits, the ‘households’ dimension allows to simulate the allocation of the Allocation de Solidarité aux Personnes Âgées (Aspa).
Several simulation options
The model allows for customised simulation via a wide range of options. Some options allow the robustness of the results obtained to be assessed, while others have been added for specific studies. The options include:
- the INSEE demographic scenario used in projections relating to mortality, fertility and net migration;
- the legislation to be applied: either the latest legislation incorporated or previous legislation;
- the type of mortality: whether or not to take into account differentiated mortality according to level of education.
The other options available in the model are documented in the online repository (see ‘Learn more’).
Recent publications
- Quelles règles d’indexation des retraites pour diminuer la sensibilité du système à la croissance ? – M. Chabaud, T. Olivia, J. Rubin et D. Blanchet, Insee Analyses no 109, 2025
- Quels effets budgétaires et redistributifs des règles d’indexation des retraites ? – D. Blanchet, M. Chabaud, T. Olivia et J. Rubin, Document de travail no G2025/08, 2025
- Une actualisation des projections de population active tenant compte de la réforme des retraites de 2023 – M. Fabre, T. Olivia et J. Rubin, Insee Références « Emploi, chômage, revenus du travail » Édition 2023, 2023
- L’accumulation du patrimoine au cours du cycle de vie : une approche par microsimulation comportementale – L. Galiana, et L. Wilner, Document de travail no G2023/04, 2023
- Projections de population active : le nombre d’actifs diminuerait à partir de 2040 - N. Bechichi, M. Fabre et T. Olivia, Insee Références « Emploi, chômage, revenus du travail » Édition 2022, 2022
- Projections de la population active à l’horizon 2080 – N. Bechichi, M. Fabre et T. Olivia, Document de travail no G2022/06, 2022
- L'introduction d'un gradient social dans la mortalité au sein du modèle Destinie 2 − A. Cazenave-Lacroutz, F. Godet et V. Lin, Document de travail no G2018/12, 2018
- L'espérance de vie en retraite sans incapacité sévère des générations nées entre 1960 et 1990 : une projection à partir du modèle Destinie - A. Cazenave-lacroutz et F. Godet, Document de travail no G2017/03, 2017.
Learn more
The source code, the COR parameters required for the simulation, and detailed documentation of the model are available in an Ouvrir dans un nouvel ongletonline repository.
