The ethics and implications of a hyper-means-tested digital benefit

In addition to the greater administrative burden created by a hyper-means-tested welfare system, there are also implications for claimant’s privacy and for their trust in government.

Collecting additional data to personalise, target or means test a service comes at a cost to a users time and to their privacy. One of the drivers towards less means-testing and more universal benefits in the middle of the 20th century was the question of privacy. In the 1930s there were public protests and riots against the means-testing of unemployment benefits. The ‘household means test’ saw government inspectors visiting people’s homes to assess how poor people were based on the state of their bedding or utensils in their kitchen.1 2 Universal Credit has taken the working-age benefit system back to something closer to the household means test. For example, to operate the two-child rule and the associated exceptions, DWP has presumably created a de facto database of rape victims.3 By reserving the option of splitting payments to each member of a couple, it has presumably created a de facto database of victims of financial or domestic abuse. In addition, every Universal Credit claim contains a detailed picture of an individual’s work and home life, including details of their children, their work-life and their health.

There is a strong case to be made to review the data the Universal Credit policy requires to personalise and means-test, and to look for opportunities to rationalise it. This would mean less control for policy makers, but there should be a fair assessment of the utility in meeting a (verifiable) policy outcome, verses the additional administrative burden for a claimant and the impact on their privacy.

Trust and fairness

The impact on privacy is one aspect of a larger question: how does Universal Credit make people feel? Does it feel ‘fair’? That might seem like a secondary concern when viewing Universal Credit from the point of view of the business case, or the political origins, but it is not. That’s because, as Donald Moynihan, Pamela Herd and Hope Harvey note in their paper on ‘administrative burden’4:

Individuals care as much or more about the process of their interactions with the state as they do about the outcome. This implies that procedures perceived as consistent, fair, and equitable are fundamentally important to citizens

Right back to its origins, Universal Credit has been an exercise attempting to get people to change their behaviour. But changing behaviour using a system that does not feel fair, is surely an uphill battle?

Although the to-do list and the claimant commitment explain what a claimant needs to do, there is much in Universal Credit that is unknown or unknowable for claimants. In addition to the lack of a definitive list of changes of circumstances, it is unclear exactly which activities might result in a sanction.5 The result of the calculation is unknown until the end of the month. The conditionality thresholds can also change. The routes for appealing are opaque and the lack of a digital option for appealing DWP decisions runs counter to the policy intent around the use of digital.

If DWP is going to embrace digital fully, that will sometimes mean designing digital services that advantage their end-users over the department. It should, for example, add a unified digital journey for appealing decisions, along with ways for claimants to track their progress through the mandatory reconsideration and tribunal stages. It may be that the internal incentives of the department make this hard to do, so it could also consider outsourcing the design of the appeals process to another digital unit in government, or to an outside organisation.

Explaining to claimants how data is used

Another way to build trust in the system is to do a better job of explaining how data is used. While it is positive that the journal provides a history of what has happened to a claimants account, the view it provides is incomplete. Data about delegated access, automated decisions, and when an account has been reviewed by a work coach appear to be missing.

Taking the transparency portals of the Estonian government and India’s Aadhaar system as inspiration, the journal should be adapted to include further information about who has accessed data about a claimant, when automated decisions have been made and when there have been manual overrides (for example in the monthly calculation). If user research shows that it is not possible to include this information in the context of the journal without confusing claimants, DWP should explore other designs.

There are obviously limits to what an individual claimant can do to understand how data is being used. The needs of third-party support organisations should also be considered, so that DWP understands what information they need to support claimants, for example, in appealing an automated decision.

There is also a role for independent auditing, especially in areas like fraud detection, to ensure that use of data, appeals processes and automated decisions operate in line with the rights of claimants (for example, ensuring they are not unfairly targeting claimants with protected characteristics). Central government could create a new organisation to do this, or existing organisations like the NAO could take on this function.

Are activities effective?

The to-do list raises a particular ethical question. Together with the claimant commitment, it represents a mechanism for government to make members of the public perform very specific tasks in exchange for a payment to cover their most basic needs. The aim of these tasks is supposed to be to help claimants earn more money, but this raises the question: does DWP know which activities work? If it is unable to answer that question, it lays itself open to the criticism that Universal Credit is like a government version of Task Rabbit6, something for generating “make-work” tasks to keep claimants busy. Similarly, one of the original aims of Universal Credit was to increase the value of the face-to-face time spent with claimants. Again, and especially given the closure of jobcentres, is this providing value for claimants? Is there enough time available with DWP staff to support claimants in circumstances where a digital system cannot?

These questions are core to understanding if Universal Credit is meeting its aims and if it is doing so ethically. They are two of a broader set of questions around the transparency required to understand if Universal Credit is working, more of which are considered in the next section.

  1. Stephanie Ward, “Unemployment and the State in Britain: The means test and protest in 1930s south Wales and north-east England”, Manchester University Press, p94 2013 

  2. Justine Leblanc / IF, “A Look Back at Privacy” p5, 

  3. To register for this exception, claimants must complete a form and hand it to a work coach or case worker. This presumably means DWP maintains a de facto database of rape victims to calculate claims correctly). See: HM Government, “Support for a child conceived without your consent form”,, retrieved 11th October 2019 

  4. Moynihan, D., Herd, P., & Harvey, H., “Administrative burden: Learning, psychological, and compliance costs in citizen-state interactions”, Journal of Public Administration Research and Theory, 25(1), 27th February 2014 

  5. It is possible to be sanctioned for something that predates a claim 

  6. “TaskRabbit”, Wikipedia,