Complicated rules govern how police record incidents
The chief inspector of constabulary says some police are failing to class all acts of violence as crimes.
The watchdog reviewed decisions by officers to class some reports of violence and assault as "no crime".
It found that more than a third of the cases from the sample of 479 incidents had been wrongly categorised as not warranting further investigation.
Denis O'Connor, the chief inspector, said the findings meant some victims had not got the service they deserved.
'Cause for concern'
He said Her Majesty's Inspectorate of Constabulary (HMIC) would be launching a major review next year into how the forces are recording serious violence to establish whether the findings from the small study have serious national implications.
Policing and crime minister David Hanson said the report showed the majority of forces were performing well when classifying violent crime, but there were some issues that gave "cause for concern".
"Some cases where a decision was made not to record a crime are clearly unacceptable. We must better understand what is behind these errors to ensure victims get the right support."
The HMIC's findings came after an investigation into why some police forces were incorrectly classifying some acts of violence.
Mr O'Connor said the study, launched after flaws in figures emerged last year, had concluded that nine out of 10 decisions on how to classify violent incidents had been correct.
VIOLENT CRIMES RECORDED AS 'NO CRIME'
5.3% of all incidents annually classed as 'no crime'
479 selected for HMIC study
308 'no crime' decisions correct
26 were actually 'most serious violence'
145 should have been 'assault with less serious injury'
Sample too small to draw firm national conclusions
But the study then uncovered new mistakes in how officers had dismissed some reports of violence as being not worthy of being recorded at all.
The team sampled 479 decisions to categorise a report of violence as "no crime". In almost 60% of those cases, the decision to dismiss the report had been correct because the report from the public had effectively been wrong.
Examples would include someone claiming to have been hit when they were making it up to cover their own embarrassment.
But the HMIC found that just over 171 decisions - about four out of 10 of those looked at - were flawed because there had been a violent incident.
Mr O'Connor said: "It's a very high error rate on a small sample. For us as the regulator, it's a matter of concern."
The chief inspector warned that it would be "statistically irresponsible" for the watchdog to extrapolate national figures from the small sample and that next year's review would try to reach firm conclusions.
But based on the HMIC's own figures, a repeat of the flaw across all forces could mean that police in England and Wales annually failing to record up to 6,000 acts of violence resulting in injury.
Staffordshire assistant chief constable Douglas Paxton, lead on crime statistics at the Association of Chief Police Officers (Acpo), said: "Sorting incidents of serious violence between detailed categories of crime can be complex, but as part of protecting the public it is vital that we have the best quality data available.
THE TROUBLE WITH JOURNALISTS ...
Managing bad news is the job of the Home Office press team. And some of their tactics on the crime story made their way to the BBC newsroom. Less than an hour before the HMIC figures were released, one press officer messaged colleagues urgently, seeking a suitable rebuttal to fend off pesky hacks. "If journalists are hell-bent on extrapolating victims 'failed' by a no crime decision, is it fair to say they would use 1.5% of total police recorded crime Violence against person with injury figures for 08/09? ASAP so I can clear before minister goes to Millbank [BBC studios]." One of the recipients, Claire Gipson, Head of Strategy and Delivery in the Violent Crime Unit, replies. "We can be a bit more robust in challenging the No-crime = No Victim Support conclusion"
"This review indicates that in the vast majority of cases recording is accurate, but that we need to look closely and carefully at some areas of variation and inconsistency. We will look at the lessons."
Mr O'Connor called on ministers to simplify the complex and 150-years-old classifications of violence. He said the various categories made it difficult for the police to categorise incidents - but also harder for the public to understand the true extent of crime.
Mr O'Connor added: "There are some fairly well-rehearsed perverse incentives around targets and 'no crimes' is one of those potentially - but we cannot say any more than that."
But Shadow Home Secretary Chris Grayling said: "All of this just further undermines confidence in the crime figures and in the criminal justice system.
"People are never going to believe in our police and criminal justice system unless they can be absolutely certain that they are being given a true picture of what's really happening."
The review of violence comes as quarterly figures for April to June 2009 showed that overall crime in England and Wales fell 4% compared to the same period of 2008.
But the detail also reveals a 3% rise in home burglaries and a 1% rise in robberies. Violent crime fell by 1%. The latest figures from the separate British Crime Survey say that the risk of being a victim remains at a historic low.