Sex offenders are being wrongly released because the system used to assess their risk to the public is flawed, a new report has claimed.

Researchers warn that the Risk Matrix 2000 system, used by the police, social workers and the Scottish Prison Service, to determine the likelihood of sex attackers reoffending, is unreliable and should be scrapped.

Risk Matrix predictions are presented to judges and sheriffs when offenders are sentenced and are part of the criteria on whether they should be considered for early release. They are used by the police to determine which offenders should be monitored most closely on release.

There are more than 3,500 people on the sex offenders register in Scotland, the majority of whom are assessed using Risk Matrix. They are given scores based on factors including the type of offence committed and who their victim was. They are categorised as at a low, medium, high, or very high risk of reoffending.

Psychologists at Glasgow Caledonian University, whose findings have been published in the British Journal of Psychiatry, say that the system’s chances of predicting the behaviour of individuals is as low as 3%.

“It’s very popular and has been pushed by the government because it’s cheap, but it doesn’t actually work,” said Professor David Cooke, a clinical psychologist who was part of the research team. “Often offenders will be deemed to be low risk when they are actually high risk, and the opposite is true as well. The public is suffering because people who are high risk will be slipping through the net.”

Stuart Leggate, who murdered eight-year-old Mark Cummings in 2004 in Glasgow, was considered “medium risk” at the time of the killing. Leggate had been released early from a prison sentence for sex attacks on children.

The Risk Management Authority (RMA), a government group on risk assessment, published a study into Risk Matrix 2000 this year, which found it was a viable screening tool with a moderate level of accuracy.

Roisin Hall, the authority’s chief executive, said the system was used as a “starting point” for assessing offenders and many other factors were taken into consideration when determining risk status.

“We can never predict human behaviour and there’s always the danger that people will be wrongly classified.

“What the RMA is saying about the pros and cons of Risk Matrix is be sensible, be proportionate about it, and take other factors into consideration,” she said.

Times Recommends

• Revealed: how Eurocrat leaked secrets over dinners
• Politicians use the net to brainstorm
• British spy in longbow plot to kill Himmler