The Road Ahead: Beamm evolving

Beamm is a permanently evolving project across all steps of the pipeline, from synthetic data generation algorithms to web-interface design. It will remain under continuous development throughout the coming years and decades. The tax-benefit reform simulation functionality in beamm.brussels is the tip of the iceberg. At the heart of BEAMM is a robust and scalable codebase that supports data-intensive economic research projects. This page gives an overview of planned BEAMM development, with emphasis on what users of beamm.brussels can expect to change over the near and mid-term future. It outlines the project’s origins, its present state, and its anticipated direction in the near and medium term.

Origins: where we come from

Taxes, social contributions and social benefits make up about half of our GDP. How we organize them, as well as how much we raise and spend, has an enormous societal impact, both because of their sheer size and because of the indirect effects that they have on our economy by altering the behaviour of firms and households (labour market activity, entrepreneurship, investment decisions etc.). Moreover, our fiscal and social security systems are the primary instruments for redistributing incomes and wealth. All this makes a sound democratic debate and careful analysis of fiscal and social policy essential. In practice, however, assessing fiscal and social policies is complicated from a legal, institutional, economic, statistical and practical viewpoint (e.g., data access).

In 2019, we started planning a project to build an online, open-access platform that would allow policy makers, civil society, media and citizens to simulate, analyze and assess fiscal and social policy reforms with relative ease, and that would function as a bridge between our academic research and policy making. In 2020, the development of the first platform for the Brussels Region started, co-funded by the Prospective Research for Brussels program of Innoviris. In early 2024, 4 years of development culminated in the launch of our Beamm.brussels platform.

Present: the complexity of collective research

Since the launch of the Beamm.brussels platform, our focus has temporarily shifted to consolidation. Over the past few years, our team has grown from 5 researchers early 2020 to about 30 in 2025, and ERDF funding via the Brussels Region allowed us to further professionalize the project. Likewise, the complexity of the project has grown exponentially. This complexity is in part due to the nature of the model, but more so due to the project being a multi-disciplinary collective research project with a dual purpose: academic research and platform development for societal impact.
Beamm progresses because all CAPE researchers develop their research projects (academic and applied) in an integrated fashion:

• statisticians and mathematicians build AI tools for statistical matching, synthetic data generation (GAN and VAE), forecasting, explainable AI etc.,
• economists build code to simulate public policies, predict behavioural reactions, analyse outcomes, optimise public policies etc.,
• engineers and software developers improve common infrastructure and code base,
• legal scholars are working on the legal and institutional precision of the platform.

CAPE adopts shared coding styles and design principles that allows for research output to be integrated into a growing central project. As such, Beamm as a collective research project means that all researchers invest in the common good, and in exchange benefit from and depend on the contributions of all other team members. This interdependency is both an enormous strength and a source of complexity, especially in an academic context where research staff rotates rapidly.

Present: a continuous development environment for academic research

All parts of the Beamm software itself are in active development. Further, the project itself is embedded within a research environment that is constantly changing and evolving, both in individual projects, scope, and personnel.

For these reasons, the year 2024-2025 was used for consolidation, building an inclusive research context for continuous development in an academic context that will be the foundation for the further development of both Beamm the collective project, and CAPE the research group. To this end, we developed (and are developing) the following elements:

• An internal wiki site extensively documenting the different parts of the project (data, organisation, reading suggestions, procedures…), and a set of tutorials for code development and research in the context of Beamm,
• Adopting best practice software development routines and workflows to improve communication and coordination across the different project elements and its contributors.
• Redesigning our codebases for robustness, flexibility, and scalability.
• Establishing data pipeline versioning and automating for improved reproducibility and efficiency. This allows improved automation and portability across our pipeline elements (e.g. data selection - cleaning - renaming - matching - generation of synthetic data - forecasting).

The road ahead

This consolidation enables us to move ahead with the further development of Beamm. Over the next year, we will be working on the following developments, which will be pushed into the online interface frequently:

• Improving data processes. Statistically combining different raw datasets, replicating the rich interactions between different individuals in the synthetic data, and nowcasting and forecasting data with neural networks.
• Validation and reporting There is randomness (and therefore uncertainty) inherent to all synthetic data and economic modelling processes. Use of the online interface for policy reform assessment requires better information on how this inherent uncertainty affects the simulation results - and any subsequent policy reform comparisons. A priority development strand is validation and reporting of reliability and uncertainty.
• Re-activation, validation and enhancement of taxes and benefits. Some taxes or social benefits are still under re-construction, e.g., because of partially missing data. Other instruments need further validation to improve their precision. These taxes and benefits will gradually re-appear in the interface as we make progress. For the coming years, this exercise includes the following instruments:
  • Inheritance and Gift Tax
  • Real Property Tax
  • Social Security Contributions
  • Pensions
  • Income Support
  • Maternity leave + child benefits
• Tax-benefits code updates and the time dimension. The initial Beamm model was designed to simulate the legislation of one specific year (2020 originally) and enable easy update of the code over time. We are re-orienting this towards a single code-base that can deal with many years. This will in the medium term make Beamm a platform that can simulate the tax-benefit systems of recent past, present and near future. Of course, updating the tax-benefit code is part of this effort.
• Interface functionalities. Our researchers are constantly improving methods of visualising economic results to achieve more insight and intuition. We have plans to greatly improve the quality, quantity, and interactivity of the available visualisations on beamm.brussels.
• Documentation of the tax-benefit system. Under the tax benefit system item in the Beamm interface, you find an extensive summary of how different tax-benefit instruments work. Over the next two years, we will upgrade this documentation in 2 ways:
  • First, we will update, validate and extend this summary to cover multiple years of legislation.
  • Second, we will mobilize our data to integrate extensive visualisations and data tables into this documentation to quantify the workings of these different taxes and benefits, and will automate this process. This will make this documentation both a summary of fiscal/social legislation and a quantitative documentation of the workings of our tax-benefit system.
• Behavioural models. Different models capturing how households react to changes in tax-benefits policies are in different stages of development, and will gradually be integrated in the platform. This includes a labour market model, a consumption model, a model of investments in durables, an intra-household model etc.
• Video channel. Towards the end of 2025, the Beamm Youtube channel will be launched. It will feature short videos explaining elements of the tax-benefit system, elements of Beamm or research results.
• Co-creation. We look forward to further developing the Beamm platform with different partners in a process of co-creation, in order to better tailor Beamm to user needs. In a first round, this concerns different actors in the Brussels Region. In the medium run, we also look forward to exploring Beamm for participatory democracy, citizen science, secondary education and other projects.