Legacy, unautomated IT was root cause of human error in pensions underpayment

Disjointed legacy IT has been identified as the root cause of human error that resulted in up to 134,000 state pensions being underpaid by the Department for Work and Pensions (DWP).

In its report Investigation into the underpayment of state pension, the National Audit Office (NAO) said the DWP will need to pay the affected pensioners a total of £1.053bn, representing an average of £8,900 per pensioner affected.

In the report, the NAO said the underpayments were official errors, due to repeated human error over many years. The state pension rules are complex and are only fully understood by a small group of specialists. According to the NAO, the complexity and lack of expertise of the state pension’s rules, combined with outdated and unautomated systems, meant the administration needed a high degree of manual review and understanding by case workers.

In its report, the NAO noted that this lack of automation and the need to use multiple legacy systems had led to human error. “When working on a claim, staff must access information from different IT systems and manually copy across information from one system to another to calculate the award,” it said. “Different parts of the award are often assessed by different case workers in different teams and each case worker needs to use information from at least three of four separate IT systems to calculate a claim.”

These legacy systems are: the National Insurance Record System owned by HM Revenue & Customs (HMRC), which holds the history of national insurance contributions (NICs); the DWP’s Pension Service Computer System, originally developed in 1988, which holds the award and calculates payments; and the 2006 Customer Account Management (CAM) system, which holds the claimant’s case record history and personal details.

Another system, called CAMlite, also launched in 2006, is used for customers who live or have lived outside the UK. But in the report, the NAO said case workers with access to CAM do not have access to CAMlite, and vice versa. This meant the international and UK teams had to work together to understand a pensioner’s full claim history, it said.

A root cause analysis conducted by the DWP in June 2021 found a number of instances where caseworkers failed to take manual intervention. For instance, caseworkers sometimes failed to manually set IT system prompts to review the case at a later date or sometimes failed to action manually set prompts or notifications, such as of a person’s death.

On top of problems with individual users of the various IT systems at the DWP, the NAO report highlighted failings within the department to recognise the importance of processing prompts once set. The NAO said this was due to the department prioritising the processing of new claims over the review of existing claims.

The DWP’s root cause analysis also found that once set, prompts were actioned incorrectly because of instructions that front-line staff found difficult to use and a lack of training on complex cases. The NAO report said: “It also found that its centralisation of pension services to a few national specialist pension centres led to different teams looking at different elements of each case and no single person taking responsibility for the overall award.”

According to the NAO, this gap in quality assurance was partly due to the DWP’s historic approach to fraud and error, which focuses on measuring, identifying and tackling the largest sources of fraud and error across benefit expenditure. “While this is in line with our understanding of good practice, by itself it is susceptible to systemic causes of error that is small in terms of annual monetary value, but large in impact on individuals,” the NAO report said.

For instance, it reported that the DWP does not record how many people contact the department to complain about underpayments or how many arrears payments it makes. There is also no assessment of whether individual detected underpayments or overpayments have a systemic cause that might indicate a wider error, said the NAO.