Project Aurora predicts the impact of new tax rules: ‘We immediately see who wins or loses’

By Pieterjan Van Leemputten, editor at Data News

Death and taxes are the only two certainties in life. For the latter, the FPS Finance has an analytics tool to calculate every measure to the cent. This allows them to offer fast and correct advice and answer parliamentary questions.

When tax rule X is implemented, raised or lowered, what will be the impact on population group Y? Or on all seven million personal income tax returns the FPS Finance receives every year? The public service regularly needs to answer these questions for the government, parliament, local authorities or international institutions. But the answer mainly consists of an estimate after careful and complex calculations. “This is based on samples and simplified models, but it also means there may be a bias. This is the reason why some estimates in the past have turned out to be less accurate”, says Dierk Op ‘t Eynde, senior data scientist at the FPS Finances. “Even our greatest experts had difficulties to estimate who was most affected by a certain measure.”

“Today we link a simulation with 150 social characteristics, occupational groups, illness, whether someone works or has children, and we immediately see the ‘winners’ and ‘losers’ of a particular measure.”

Project Aurora changed that. The FPS Finance has implemented an analytics platform from SAS that immediately recalculates new measures or adjustments to the personal income tax for all returns and gives a detailed account of citizens who benefit from it or are disadvantaged, or how much it may or may not yield. “The average Belgian fills in ten to twenty amounts and later receives two or three pages with a calculation and the final amount, but there are about three thousand declaration codes. The combination of these means we have almost 65.000 business rules, calculation rules, that lead to 2.700 different intermediate results. Therefore, the personal income tax for the entire population is a complex issue.”

Project Aurora has now processed those thousands of rules. If the federal government considers a new measure, or when a member of parliament asks about the impact of something on a population group, it takes two hours at most to perform a calculation for all Belgians. “The big advantage is that we now get an overall picture. In the past, we would calculate and only see if the final total is different, or whether the target group of a measure is benefitting from it. Project Aurora allows us to see the impact on all taxpayers.”

Aurora is, to be clear, no replacement or alternative for Tax-On-Web. Whereas the latter is primarily the operational part of the personal income tax, intended for the effective introduction and calculation, Aurora is rather a workable part to precisely calculate the impact of new rules. For the latter, the FPS has 1/4^th manpower, or the equivalent of one person for three months, to keep the entire system up to date with the latest tax rules. For comparison, Tax-On-Web needs 2 to 4 fulltime employees.

By using both systems together, the FPS Finances can also retroactively check whether the assessment notes are correct. How correct? “Out of a total of about 67 billion euros (67.000.000.000,00), there was a difference of not even seven euros (6,70 euros) last year. But even that depends on rounding margins.”

Digital twin

Project Aurora has been operational for around a year and a half and is a good example of a digital twin. “A policy measure is not just tested and rolled out. It is often worked on gradually to determine whether something counts for roughly 25 or 30 percent. It was often a chain of notes to process and then calculate. Today we link a simulation with 150 social characteristics, occupational groups, illness, whether someone works or has children, and we immediately see the ‘winners’ and ‘losers’ of a measure”, Op ‘t Eynde explains.

The information can also be very detailed. A self-service BI analytics tool has recently been added to enable policy makers to click through and go into the detail of numbers. This immediately gives a much broader view on the effect of a measure. Of course, project Aurora is privacy sensitive. “The data owner is the Finance minister and the chairman of the FPS Finance. Direct access to the tool is restricted to the administration. All calculations happen on an individual level, no names are mentioned. But in combination with several parameters, you could easily find out who it is.”

“With the right tools, we now get reliable data based on well-founded algorithms. Users don’t need to understand the finesse behind it, but can now make more correct decisions”, Op ‘t Eynde concludes.

Original text in Dutch & French written by Pieterjan Van Leemputten, published on datanews.be, 15 November 2019