About BEAMM

BEAMM stands for BElgian Arithmetic Microsimulation Model — a tax-benefit microsimulation model. It applies the rules of the tax and social-security system to each household in a representative sample of the population, so you can see how a policy reform’s impact varies with people’s circumstances.

Here’s how it works, end to end:

  1. Creating a synthetic population
  2. Making it representative
  3. Simulating the tax-benefit system
  4. Analysis & visualisation

Expand How it works on any step to go deeper.

Creating a synthetic population

No single dataset covers everything BEAMM needs — income, expenditure, mobility, family composition. We merge several anonymous datasets into one using statistical matching, then generate a 100% fictitious but statistically realistic Brussels population from it.
How it works

Why a synthetic dataset

BEAMM needs rich data on every household — income, expenditure, mobility, family composition. No single Belgian dataset has all of it, and much is privacy-sensitive. So we build a synthetic population instead of using real records.

Merging datasets — statistical matching

Each dataset we hold covers different information but shares some common variables (age, household, location). Statistical matching uses those shared variables to fill in the gaps — imputing what each dataset is missing — so the merged result holds everything, statistically accurate for the population even though it is wrong for any one (anonymous) person.

Statistical matching fills the gapsAdministrative data holds shared demographics and income but not spending; survey data holds shared demographics and spending but not income. Matching on the shared demographics fills each gap, producing one synthetic dataset with all variables.DemographicsIncome & taxSpendingAdmin dataSurveySynthetic––matched on shared traits → gaps filled

Making it 100% fictitious

The matched data can still hold fragments of real records. To remove them, we apply generative AI — for example a Generative Adversarial Network — to produce a dataset that reproduces the real statistical distributions without reusing any real data. The result is 100% fictitious but highly realistic, validated against the real data, and is what powers this platform.

Further reading: Annoye, Beretta & Heuchenne, statistical matching using machine learning (CAPE).

Making it representative

A synthetic sample isn’t automatically representative. We reweight it so the population’s make-up — ages, households, employment — matches official figures, so results speak for the real Brussels population.
How it works

Reweighting to the population

A synthetic sample doesn’t automatically match the real population. We reweight it against the Federal Planning Bureau’s demographic and household projections — adjusting how much each household counts so the totals (age, household type, employment) line up with reality. We change the weights, not the individual records.

On the Road Ahead. Bringing the data fully up to the present — uprating incomes and prices, and letting you pick the target year — depends on data we’re still securing. See Road Ahead.

Simulating the tax-benefit system

BEAMM applies Belgium’s modelled tax-benefit rules to every household — twice: once for today’s system, once for the reform you choose — so the two can be compared. Every instrument is traced to the law and backed by automated tests.
How it works

How

For each household, BEAMM computes every modelled tax, contribution and benefit. It runs twice — the current system (pre-computed) and the reform you choose (when you press Simulate) — and compares the two.

How the engine worksEach household is run through the tax-benefit rulebook — which is traced to the law and automatically tested — twice: once for the current system and once for your reform. The difference between the two is the impact.HouseholdTax-benefit rulebooktraced to the lawautomatically testedCurrent systemYour reformtheimpact

Built to be correct

Modelling a tax or benefit correctly is exacting work, and it’s where BEAMM earns its trust. Every instrument is:

  • traced to the actual legislation, with the legal sources referenced; and
  • backed by a comprehensive suite of automated tests, so the computed amounts stay correct as the law — and the code — change.

What it covers today

The live platform lets you change five areas:

  • Personal income tax — the tax base, principal amount, reductions and credits
  • VAT
  • Excise duties
  • Car taxation — entry-into-service and circulation tax
  • Investment-income tax

More — child benefits, pensions, income support, inheritance and others — is in development; see Road Ahead.

Analysis & visualisation

BEAMM compares before and after across many dimensions at once — budget, inequality, poverty, who gains and who loses — so a reform is never judged on a single number.
How it works

The full picture, not one number

With every tax and benefit computed before and after, BEAMM turns the results into the measures that matter — disposable income, inequality, poverty, the state budget — and breaks them down by household type, income, age and more.

The interface presents these dimensions together, on purpose: a reform has many facets, and judging it on one headline number is usually misleading. You can also pair a reform with compensating measures and see their combined effect in a single run.

See it in action in the simulation.

© 2026

DISCLAIMER

PRIVACY POLICY

Beamm is a registered trademark of the Université Catholique de Louvain