Reuters, ITV, the Mail and the Express all ran with variations on the theme "Violent crime soars 12%". The Times headline referred to "One million violent crimes". The Mirror led with "Lawless UK". And the Sun, under the heading "Drunk and disorderly", reported how "Boozed-up yobs were blamed yesterday for a whopping 12 per cent surge in violent crime."

That was one version of the different crime figures published last week. The other was represented in the Guardian, under the headline "Longest period of falling crime for 106 years"; and, more stridently, in the Independent, which proclaimed: "Crime: the truth. New figures reveal that crime has fallen 39 per cent over the past nine years - the biggest sustained fall since the 19th century."

This conflict of evidence has become a recurring theme in the crime debate. On the one hand, the Sun can declare - correctly - that: "The number of violent offences last year topped a million for the first time . . . Assaults, woundings and threats to kill have TREBLED under Labour, according to the shock figures. Meanwhile, police detection rates have FALLEN in every region of England and Wales."

On the other hand - and equally correctly - the Independent can demonstrate how the decline in crime has been "quite staggering: car thefts, burglaries, domestic violence and assaults on people who are known to each other have all dropped by about half . . . In every category of crime - including violent crime - there has been a decrease. The risk of becoming a victim of crime has fallen from 40 per cent in 1995 to 26 per cent."

So whose interpretation of these apparently conflicting statistics does one believe? And what does it mean in terms of public policy on crime and punishment at a time when the government is sending out seemingly mixed messages announcing that crime is under control and falling, and at the same time that it is necessary to ditch the so-called "1960s liberal consensus" and get tough on law and order?

First, the figures. The 12% hike in violent crime that made most of the headlines last week comes from police figures for recorded offences, which showed a total of 1,109,017 violent offences in the UK in 2003-04, up from 991,603 the previous year. Altogether, on this reckoning, the number of recorded crimes rose 1% to 5,934,580.

The Guardian, Independent and the government, however, preferred to rely on the British Crime Survey, in which 40,000 people aged 16 and over are interviewed about their personal experiences of crime. This is considered to provide a more accurate picture because it includes unreported crimes. About half of all crimes are not reported to the police, so the BCS figures show an estimated 11,700,000 crimes committed in the year ending April 2004. This may be almost twice as high as the number of crimes recorded by the police, but it is a massive reduction from the peak of 19,300,000 crimes reported in the 1995 British Crime Survey.

"Not since 1981, when the total was a little over 11 million, have the figures been so low," the Independent declared triumphantly. "The last time Britain enjoyed such a sustained fall in crime was in 1898 - a decade in which, ironically, Jack the Ripper was terrorising Whitechapel, east London."

So is it just an "urban myth" that fear of crime and anti-social behaviour is rising, as Professor Paul Wiles, the director of research, development and statistics at the Home Office, believes? And have David Blunkett and Tony Blair miscalculated in their emphasis on a crackdown on social disorder and crime?

The answer is that neither the precise figures nor the detailed trends matter as much to most people as what is happening in their immediate neighbourhoods. And whichever way you look at, more than a million violent crimes in a single year means that a very substantial proportion of the population will have had direct experience as a victim, or as a friend or relative of a victim, of those crimes.

Look at people's experience of violent crime over a longer period (our perceptions are shaped not just by what happened in the year 2003-04 but by what occurred in the five, ten or 15 years before that too), and it becomes clear that the fear of crime is not urban myth but reality. The statistical risk of becoming a victim of crime may have fallen from 40 per cent in 1995 to 26 per cent in the year just gone, but spread that figure over four years and it becomes a 104% probability, or 260% over ten years. Spread it over a lifetime and the odds are that an individual will have been the victim of more than 18 different crimes. And since crime is not distributed evenly, some people are going to have been on the receiving end of a great deal more than that.

It's not necessary to rail against some supposed "1960s liberal consensus" on crime to recognise that safeguarding the right of citizens to live free from fear is one of the first responsibilities of the state. Those 1,109.017 violent offences recorded in the UK last year include 955,752 offences of violence against the person, a 14% increase. Threats to kill were up 23% to 22,232; serious wounding up 8% to 19,358; racially-aggravated wounding up 11% to 4,840; and harassment up 26% to 152,269. Sex offences increased by 7% to 52,070 - including an 8% increase in rapes of women to 12,354.

These are all statistics that merit urgent action, whichever way you choose to interpret them.