Audit Scotland has published a report on the operation of the “National Fraud Initiative”, which is mainly concerned with the operation of big data in order to detect fraud by searching for anomalies. The central assumption of the report seems to be that improving the consistency of data will save money by limiting opportunities for fraud.
There is a central confusion at the heart of this approach. Following the reports of the DWP, and by extension the tests applied to local government in benefits administration, the report muddles three overlapping, but quite distinct issues: fraud, error and selectivity. Using data to review inconsistencies identifies potential sources of error. It does not necessarily identify, or relate to, fraud – which depends on deceit.
In general,
- measures to refine selectivity are liable to increase error (the more conditions there are, the more there is to go wrong. That is why the rates of error in Pension Credit are fifty times those for State Pensions.) Equally, they and they create opportunities for fraud. They do not, then, improve the efficiency of the system.
- measures to protect systems against fraud are likely, for the same reason, to increase error.
- measures to protect against error may reduce opportunities for fraud, but do not necessarily do so – it all depends on what kind of error is being addressed.
There may be opportunities to save money through big data; but anti-fraud measures are often expensive, and in a situation where many people do not receive the benefits they are entitled to, smoothing out inconsistencies might cost more.