Czech Ministry of Labour and Social Affairs – III pension system

The study of input data, the creation of analytical techniques, and the application of individual decision-making processes into a dynamic microsimulation model of the pension system are the topic of the public contract. The Czech Ministry of Labour and Social Affairs (MoLSA) has used a microsimulation model based on the Prophet actuarial software platform since 2011. The evolution of decision-making processes should result in a more accurate representation of individual behaviour, allowing for a more trustworthy evaluation of future modifications to the pension system.

The development was divided into several categories, including Concurrence of payments of old-age pension with income from the gainful activity, Commencement of retirement pension, Care for dependents, Region, Occupation, Wage, Self-Employment and Sickness and Disability. The DataSentics team handled data analysis, and Tools4F was in charge of putting the Prophet model and DCS code's new adjustments and factors into effect. 

Tactical objectives

For the selected thematic areas, analyse relationships in data and implement appropriate extensions to the model. More accurately capture the behaviour of individuals. More reliable assessment of potential changes in the pension system.

• A search of data sources on pension rights - This is a 35-page report that summarises data sources on the earned and consumed pension rights in the basic pension scheme. This report is accompanied by a detailed analysis of the data from the assessment base database provided by the CSSF.
• Analysis of available data - This is a 40-page report describing the data sources necessary for producing the model points and the derivation of the input parameters.
• Review of pension models - This 22-page report summarises the typology of pension models and the possible data sources for different pension models. It contains a description of pension models that are or have been used in the Czech Republic.
• Dynamic microsimulation model - MPSV was developed in accordance with the technical specification agreed upon between the contracting authority (MPSV) and subcontractor (Deloitte), and a functional dynamic microsimulation model was delivered. This model was developed in the Prophet actuarial software and was installed on the MoLSA computer.
• Code for the creation of model points - Deloitte provided the contracting authority with a sequence of programs in DCS and SQL to create model points based on the administrative data of the Social Security Administration and additional information on the economic status, stock status and marital status of the population.
• Manual for the use and administration of the dynamic microsimulation pension model - Deloitte provided the sponsor with the model and the code for creating the model points, together with a user manual documenting the operation, updating and methodology of the dynamic microsimulation model and a description of the model point creation code.
• Training - In January and March 2011, Deloitte organised two training sessions for selected representatives of the MoLSA. This training focused on using the Prophet actuarial software, the design of the microsimulation pension model and its outputs. The training aimed to build the technical capacity of the MoLSA to take over the model and its management after the end of the project. To ensure the transfer of know-how, in the final phase of the model development, a selected MoLSA staff member was involved in the review and testing of the model.
• Seminars and final conference - Two seminars were held at the end of the project, sponsored by the NVF and with content provided by Deloitte.

Methods

We used Azure Databricks, Apache Spark, and Python to analyse data and create the statistical models. We used FIS® Insurance Risk Suite (formerly Prophet) and NEMO model to implement the simulation model. By analyzing extensive amounts of data sourced from the Ministry, it is possible to generate models such as wage development or retirement probability that consider multiple factors.

For a more detailed description of the NEMO model, consider the following aspects:

• Dynamic microsimulation model of the pension system
• Microsimulation = at the level of individual so-called.
• Model points = millions of Czech citizens
• A model point has different properties and states
• Random events are generated monthly
• What-if analysis

Functional requirements

Ministry of Labour and Social Affairs, in collaboration with external companies. Mobilisation of the experts in the field for further discussion during the dissemination of events. 

Results and impact assessment

The series of dissemination events were prepared within the framework of the project with the primary objective of informing partners in EU countries and EU-associated countries about the conclusions and findings of the project as well as the main characteristics of the dynamic microsimulation model – NEMO. The presentation of the NEMO model took the form of technical seminars for experts in the field in each participating country. Particular attention is paid to the data used for the microsimulation model and the indicators used for the analysis. The feedback and experiences of seminar participants are expected and welcomed. Further, the eight selected thematic areas to analyse relationships in data and implement appropriate extensions to the model while capturing the behaviour of individuals. More reliable assessment of potential changes in the pension system.

Dependencies and constraints 

• It would be interesting to compare different settings of the 3rd pillar of the Czech pension system.
• As concerns the composition of the fund – made out of own contributions, employer contributions, and state support – the influence decreases over time
• Some space is likely to open up towards pension companies and possibly their projections and better overall cooperation – it would be significant to have data concerning the savings decisions of individuals etc.
• The innovation for the further project could be mirroring the legislation; it seems to be crucial to provide inhabitants with the individual information about the pension.
• There is still a lot of work on the data side; at the moment, there is only national and pension contribution data, and it is not connected (e.g. to educational level). If any pension company provided data into the system, this would need to be legislatively anchored, data individualised, with precise settings of responsibilities and data format. Cases exist in some of the countries within the Europe Union.