Fight against Tax Fraud in Belgium

Author

Toni Verbeiren

Published

January 25, 2012

The (brand new) Belgian government has decided to raise the stakes in the fight on tax fraud. That’s a good thing, especially in times of tight government budgets.

In the news last week, it is said that the group of people actively involved in detecting tax fraud will be enlarged by a quarter (a little less than 300 new hires). That means that in total, for Belgium alone, more than a thousand man and women are looking for ways to spot fraud. The involved government agencies will be looking for Masters in fields like ‘accountanting, law but also informatics’.

This sounds like the good old ‘brute force’ method: throw a bunch of people on a problem and when the problem gets bigger, throw more people at it.

Basically, there’s one major field of interest missing from the list profiles in the press release: statistics. People trained in statistical analysis may have heard of Benford’s Law or related concepts. They may take a completely different approach to dealing with large data sets. Statisticians may propose to use relatively simple heuristics first, in order to select the subset of targets with the highest probability of fraud.

An example of what the process might look like:

  1. From a very large dataset, find a way to select a small subset that a) has the highest probability of fraud, b) has the highest probability of being able to prove fraud and c) actually benefits the government by selecting only the ‘big fish’.

  2. Get the informatics specialists to implement this heuristic algorithm as fast as possible, with ways to tune the individual parameters for deeper analysis.

  3. From the resulting small selection, invite accountants and lawyers to focus on those.

Have I forgotten something? I’m tempted to try and contact the federal government agency responsible for tax fraud and propose to start with 1.

Update: Thank you Paul for correcting some errors.