Recommendation 1
a work with HMRC to review the level of assurance over the integrity of National Insurance records and identify ways to reduce error in the records.
Source · National Audit Office
DWP Report on Account 2022/23
Published: 6 Jul 2023
Recommendations: 16
Type: Value for Money
NAO confirmed: 16
NAO report detail with recommendations, government responses, and any Public Accounts Committee follow-up.
Recommendations
Recommendation 2
b establish an ?early warning system? to record types of underpayment errors that happen repeatedly and take action before they develop into significant issues. As part of setting up this system DWP should:
? encourage frontline staff to refer large or unusual underpayments for analysis;
? encourage frontline staff to refer large or unusual underpayments for analysis;
Recommendation 3
c improve its experimental framework to further enhance transparency around its counter-fraud performance. DWP should aim to demonstrate that changes in its metrics are a result of its activity and not external factors, and to show how its activities are affecting the overall level of fraud and error. To achieve this DWP should:
? set targets for a greater number of individual benefits, prioritising benefits with the most significant levels of fraud and error;
? set targets for a greater number of individual benefits, prioritising benefits with the most significant levels of fraud and error;
Recommendation 4
d continue to set an annual target for annually managed expenditure (AME) savings until it sets a target for the monetary value of fraud and error and do more to justify to Parliament that the target is sufficiently stretching.
Recommendation 5
e DWP should set out a roadmap for developing its performance framework and counter-fraud activities to a point where it can demonstrate that it is operating a cost-effective control environment.
Recommendation 6
g as part of its response to PAC due in November 2023, DWP should seek to improve public confidence in its use of advanced data analytics by committing to regularly publishing summaries of:
? its assessment of bias in machine learning models; and
? its assessment of bias in machine learning models; and
Recommendation 7
i DWP should collect information to provide assurance over the quality of customer service and report on this in its 2023-24 Annual Report and Accounts.
Recommendation 8
f exercise caution in how it describes its measure of overall overpaid welfare spending to not conflate it with its existing measure of fraud and error, and replace its current methodology for the alternative measure to better focus on the impact of introducing Universal Credit on the cost of overpayments (as set out in the Universal Credit business case) by:
? focusing solely on Universal Credit and the benefits it replaces and excluding other benefits; and
? focusing solely on Universal Credit and the benefits it replaces and excluding other benefits; and
Recommendation 9
h DWP should continue to scale Targeted Case Reviews and in doing so ensure it hits the level of recruitment and productivity required to achieve the expected savings over the Spending Review period; and
Recommendation 10
DWP should: b perform more in-depth analysis to determine the root cause of a greater number of detected underpayments; and
Recommendation 11
DWP should: b develop the capability to record detected underpayments across all benefits as part of any planned updates to its IT systems.
Recommendation 12
c improve its experimental framework to further enhance transparency around its counter-fraud performance. DWP should aim to demonstrate that changes in its metrics are a result of its activity and not external factors, and to show how its activities are affecting the overall level of fraud and error. To achieve this DWP should: set its targets to cover a longer time horizon, to better align with the time its activities take to have an effect;
Recommendation 13
c DWP should: regularly publish its fraud and error forecasts; and
Recommendation 14
DWP should: c improve its experimental framework to further enhance transparency around its counter-fraud performance. DWP should aim to demonstrate that changes in its metrics are a result of its activity and not external factors, and to show how its activities are affecting the overall level of fraud and error. To achieve this DWP should: refine its savings estimate methodology to increase the level of confidence in each element so that it is robust enough to make year-on-year comparisons. This should include testing the assumptions where possible.
Recommendation 15
DWP should: f exercise caution in how it describes its measure of overall overpaid welfare spending to not conflate it with its existing measure of fraud and error, and replace its current methodology for the alternative measure to better focus on the impact of introducing Universal Credit on the cost of overpayments (as set out in the Universal Credit business case) by: focusing on the annual financial savings to the Exchequer rather than a percentage overpaid so as to not conflate this measure with the monetary value of fraud and error and to better account for differences in recovery rate and measurement period.
Recommendation 16
DWP should: g as part of its response to PAC due in November 2023, DWP should seek to improve public confidence in its use of advanced data analytics by committing to regularly publishing summaries of: any impact on customer service.