The use of data analytics in the fight against welfare benefits fraud can be very beneficial for government. In Denmark, a specialised unit, the Data Mining Unit, was established in 2015 to use data coming from the Basic Data Registries and other sources to detect fraud. The unit was profitable rather quickly in generating an equivalent of EUR 10.9 million worth of paybacks and prevention of wrongful payments in 2016. The Return on Investment has become even higher over the years with the revenues amounting to an equivalent of EUR 61.9 million in 2019, while keeping the yearly investment steady at approximately EUR 3.4 million. Moreover, by catching and stopping more wrong benefits payments at an early stage, the burden for the competent authorities of administering the benefit in the first place and later retrieving wrongly paid benefits is alleviated.
But what about the Danish citizens? Do they experience any advantages from this programme besides the indirect effects of saving tax payers’ money?
Perhaps drawbacks of systematic surveillance, profiling and being treated as a potential fraud suspect come to mind first, rather than what citizens have to gain. Critics claim that the social and political debate on data ethics in Denmark - not just in the area of fraud detection - is insufficiently present. Public values such as transparency and proportionality would be in dire need of being discussed with regard to the ambitious digitisation efforts demonstrated by public authorities.
However, over the years the Danish government has increasingly given thought to the position of citizens, their rights and how they can benefit from the use of data analytics in relation to welfare benefits.
Helping citizens amidst administrative complexity
The point of departure for the Danish Government is that it is easy for citizens to fall victim to the complexity of the welfare benefits system. On the one hand, they can make mistakes themselves, as it is difficult to understand what benefits they are and aren’t entitled to and what impact changes in personal circumstances may have. This can easily lead to errors in benefit applications. On the other hand, the complexity of the system may lead to welfare authorities making mistakes and paying out benefits they shouldn’t.
This presumption of error doesn’t mean that the Danish government is blind to purposeful welfare fraud on the part of citizens. Several of the algorithms used by the Data Mining Unit have been designed to track down non-accidental irregularities in benefits payments. However, the procedures followed by the caseworkers investigating the potential fraud cases have been designed with the weak rather than the fraudulent citizen in mind. Citizens can count on several legal safeguards, such as receiving information on the investigation, being asked consent for further data collection from private entities, getting the opportunity to clarify the situation and having the right to appeal. The investigating authority has the burden of proof and needs the citizen’s cooperation on several aspects.
A certain protection to citizens may already be incorporated in the control procedure, but the Danish government has realised it can do more.
Early intervention as a benefit to citizens
Having to pay back months or years of wrongly received benefits payments constitutes a real burden for citizens, especially for those who already find themselves in a difficult situation. ‘The earlier the error is detected, the better it is for the citizen and government’, is the leading thought behind the Data Mining Unit’s new strategy.
The unit is therefore expanding its activities in the detection of potential fraud and error in the earliest stage possible. Instead of having to ask a citizen to pay back a benefit (s)he incorrectly received, the Danish government wants to discover irregularities in the application phase, before any payments are made. For this purpose, the unit is running some tests to develop a web service for new applicants. It also trains its machine learning algorithms in detecting irregularities with regard to recent changes in personal circumstances of benefit recipients (e.g. change of address). This allows caseworkers to correct erroneous changes in benefit payments in an early stage (usually one month), thereby preventing a potentially fraudulent situation.
Besides preventing a potentially difficult situation for citizens, early intervention may also be a good idea from a human rights perspective. Interconnecting and analysing citizen data are now focused on the payment of correct benefits rather than investigating overpaid citizens.
Serving and protecting citizens in Europe
Do you think the actions taken by the Danish government suffice to serve citizens while protecting their rights? How are citizens’ rights protected in your country related to the use of data analytics in the welfare domain? Could the Danish approach work in your context?
Could European countries take it one step further and use data analytics not only to prevent wrong payments, but also to facilitate the right ones? As demonstrated in India and Mexico, data can be used to detect unaware eligible citizens, help them to exert their rights and get the benefits they’re entitled to. This would mean leveraging data analytics even before the application phase.
Please share your thoughts as a comment below this post. If you want to know more on the development and results of the Data Mining Unit, you can consult the complete case study ‘Welfare Fraud Analytics in Denmark’.