Advanced analytics helps policymakers determine how new tax measures would affect citizens

Belgium's Federal Public Service Finance uses SAS^® Viya^® to better understand the effects of tax reform 

There’s nothing as certain to raise public outcry as unfair taxes. But it can be complicated to predict the impact of even a small change in tax laws on a vast and varied population. This is especially true for Belgium’s personal income tax because the calculations are so complex.

To complete a personal income tax filing, the average Belgian fills in 10 to 20 fields of a form and later receives a two- or three-page document that calculates their final tax bill. From a technical point of view, a tax declaration consists of 3,000 distinct declaration codes, and the calculation involves 65,000  lines of business logic. Multiply that by 7 million taxpayer returns.

Adriaan Luyten, Head of the Tax Policy Unit in the research department of Federal Public Service (FPS) Finance in Brussels, says that getting a better handle on the potential effects of future tax reforms is a valuable contribution to public policy making. He and his team have been assigned to profile the “winners” and “losers” before regulations are implemented.

Benefits of digital twin technology

“The calculation of personal tax income is something that, operationally, has been possible for a long time,” Luyten says. Traditionally, the analysis was based on a population sample because analyzing such a large data set is difficult and time-consuming. But sampling means the answers are biased. “So you have a partial explanation of reality at the moment,” he says.

To get a true representation of the impact of future tax reform, the data scientists at FPS Finance would need an analytics tool precise enough to zoom in on small segments of the population and powerful enough to analyze all 7 million returns quickly.

The agency’s new tool, Aurora, offers that flexibility. Aurora is a digital twin of the calculator that processes Belgium’s income taxes. It includes a copy of Belgium’s personal income tax data, the complex calculations and the multitude of business rules. Regardless of the scope, Aurora, based on SAS Viya, delivers fast, accurate answers.

Before Aurora, policymakers relied on the output of a simplified computational model based on a sample of taxpayer data. But when data scientists need to evaluate the effects of a tax change on a microsection of the population, a sample won’t accurately represent the portion of the population they are hoping to study. They need to look at the entire population.

“We use Aurora whenever we need very accurate results or an analysis of a tax regime that is very rare,” says Luyten. “The better our estimations are, the better our policymakers are informed and the better the results will be.”

Dynamic visual reports

Getting the information in the hands of data scientists is important, but it’s also crucial that policymakers can understand the details. “I also like the presentation capabilities in SAS Viya,” Luyten says. “SAS Viya lets my team create dynamic visual reports to get the right information to the right audiences. Even during a meeting, you can decide to zoom in on a particular group of taxpayers for more detail.”

The functionality and presentation capabilities are key to understanding such complex outcomes.

“You can have a lot of results from the data, but you need to be able to present it visually,” he says. “You also need to be able to drill down to answer questions about the data such as the region, age or type of customer who’s impacted. By combining the correct graphs – visually appealing graphs – on modern digital devices like tablets and smartphones, the data is extremely valuable to our information consumers.”

Better decisions for the country and its citizens

In a field where accuracy makes all the difference, data scientists at FPS Finance use analytics to help Belgium’s government make precise predictions. The result? Better decisions that benefit the country – and its residents.

FPS Finance can also retroactively check whether Aurora’s assessments were correct. How accurate are they? According to Luyten, “Out of a total of about 67 billion euros, there was a difference of 6.70. But even that depends on rounding margins.”

This story was written based on the AI Heroes page and its corresponding FPS Finance customer videos.