Sunday, May 02, 2004

Mea Culpa

I owe Tim Lambert a small apology. In a previous piece I made a pretty stupid statistical error which he caught, and I have, until now, failed to correct it. I will do so now.

In I Pound My Head Against the Wall Because it Feels So Good When I Stop I stated:
To me that isn't as important as the fact that England, according to the British crime survey, suffered 276,000 robberies in 2000, and the U.S. about 408,000. With six times England's population, that makes the English rate four times the American rate.
Tim followed the provided links and responded:
Oh, and you blew the comparison of robbery rates. You have compared the survey measured robbery rate in England with the police reported robbery rate in the US. The police reported number in England is 78,000 (it's right next to the 276,000 figure you reported) that's roughly the same rate as you get with 408,000 robberies in the US once you adjust for population.
Tim was correct, I did mix crime survey and police reported levels of crime, and that was an error. My apologies. It was not intentional. However, it was my intent to use survey results for both, rather than police reported crime numbers, because there is some significant doubt as to the accuracy of the actual levels of crime as reported by police agencies in England.

To illustrate this doubt, let me preface by providing this Telegraph story from 2003:
Britain the most violent country in western Europe

By John Steele, Crime Correspondent (Filed: 25/10/2003)

Britain has the worst record in western Europe for killings, violence and burglary and its citizens face one of the highest risks in the industrialised world of becoming victims of crime, a study has shown.

Offences of violence in the UK have been running at three times the level of the next worst country in western Europe, and burglaries at nearly twice the rate.

Britain has the highest level of homicides in western Europe and the totals for robberies and thefts of motor vehicles have also been close to the highest in the European Union, outstripped only by France, the Home Office figures show.

Only Germany, which has 20 million more people, recorded more crimes overall in 2001, the most up-to-date figure in the research - International Comparisons of Criminal Justice Statistics 2001, with data collected by the Home Office and the Council of Europe.

But the "victimisation risk" - showing the risk of suffering a crime - in England and Wales is higher for overall crime than anywhere else in Europe, and higher than in America. The same is true of falling victim to "contact" - violent - crime.

England and Wales also had markedly fewer police officers per head of population than France, Germany and Italy, according to the study.

The Home Office points out that police have achieved some reductions in violence and robbery in 2003.

The study is also accompanied by warnings about the difficulties in making comparisons because of differing definitions and methods of recording crime. But the sheer scale of offending in the UK in recent years is apparent from the figures.

Britain had 1,050 homicides in 2001, three ahead of France, the next worst in western Europe.

In 2001, UK police recorded nearly 870,000 violent crimes, a figure hugely above the next highest total - 279,000 in France. Germany recorded 188,000 violent offences.

There were around 470,000 domestic burglary offences in England, Wales, Scotland and Northern Ireland. Spain recorded 247,000 offences, France 210,000 and Germany 133,000.

The figures for robbery, which surged in Britain around the turn of the Millennium, showed about 127,000 offences in 2001.

This was surpassed only by France, with a total of 134,000. Both countries were ahead of Spain (104,000) and substantially ahead of Germany (57,000) and Italy (66,000).

Overall, in 2001 nearly 6.1 million crimes were recorded in the UK. Only Germany had a higher total (6.3 million).

Hazel Blears, the Home Office minister for crime reduction and policing, said: "This report shows the picture in 2001.

"Since then we have cut crime further and dramatically increased the number of police on our streets."
Those are pretty serious numbers, don't you think?

Now in this slightly earlier piece there seems to be some question as to the accuracy of the data:
Rising crime, falling accuracy

By Philip Johnston Filed: 05/04/2003)

What has happened to crime statistics? Once they were the gold standard of the criminal justice system against which could be measured the success of the police against the villains.

We relied upon recorded crimes - those reported to the police - as a guide.

But, increasingly, the Government has come to rely upon the British Crime Survey. This used to be conducted every two years (it is now annual) among a pool of about 20,000 people who give their personal experience of crime. It has a major flaw in that it excludes under-16s.

Ministers began to notice that the BCS told a different story to the recorded crime figures: it was registering a decline. So, the survey became the new guide for the Government, talked up by ministers as the only true measurement of crime.

Furthermore, the Home Office was unhappy with the way the police recorded their statistics and so it introduced a new National Crime Recording Standard - a sort of statistical quality control.

This, then, is where we stood yesterday when the latest quarterly crime figures were produced. "Crime is down," said Bob Ainsworth, the Home Office minister. "These figures show government measures to reduce crime are working."

Well, do they? Let us take the claim that domestic burglary fell by 11 per cent from just over one million to 948,000 in 2002. This is not a real figure but an estimate calculated using interim population figures supplied by the Office for National Statistics. So, too, is the 17 per cent "drop" in vehicle thefts. Why is the Government relying on a survey to establish the theft of a car or a house break-in? Who does not report a stolen car or a burgled house?

When we look at the crimes recorded by the police a different picture emerges. Over the three months to December, domestic burglary fell by less than two per cent and vehicle theft by just three per cent, both of which are "statistically insignificant".

Total recorded crime rose by more than four per cent over the quarter and by eight per cent over the year as a whole. The Government finesses this by "adjusting" the figures to account for the new recording standard. And, lo and behold, they then go down. Instead of the four per cent increase in the three months to December, we discover that it has, in fact, miraculously fallen by seven per cent.

However, this adjusted figure is also an estimate. Needless to say, the Home Office highlights the two estimated measures of crime - the BCS and the new recording standard, which show a decline - and ignore the recorded crime figures that show an increase.

Or take violent crime, which the Home Office said "appears to have levelled off". The recorded crime figures show a 28 per cent rise in the final quarter of 2002. Yet after "adjustment", this declines almost to zero on the grounds that "most offences are relatively minor assaults". Adjustments are always made to make the figures look more positive.

This statistical jiggery-pokery is making it almost impossible for observers to know what is going on. The Home Office stopped publishing monthly asylum figures because they produced bad publicity on a regular basis. Recently the Home Office issued figures claiming that the reconviction rate among young offenders was falling. Closer scrutiny showed this just was not true. An official complaint has been lodged with the Statistical Commission about the way race figures have been used.

In the short term, the Home Office's inventive use of statistics may get favourable headlines. In the long run, it risks damaging its reputation for straight-dealing, perhaps irreparably.
It's tough to know what to believe when the guidelines keep changing. And then there's the declining trust in the police to do much for you when you've been robbed. The British Government uses the British Crime Survey numbers because - even though the numbers are massively higher than the police reported numbers, the BCS numbers are coming down while the police recorded numbers are going up. Seeing as the BCS numbers - although they exclude victims under the age of sixteen - are supposed to represent reported and unreported crime, those are the ones I intended to use. However, to be consistent, I needed to use U.S. National Crime Victimization Survey numbers in comparison, not the police recorded number.

According to this Dept. of Justice Report in 2000 there were 732,000 attempted and completed robberies in the U.S. in the year 2000. That's 732,000 estimated under the National Crime Victimization Survey, as opposed to the 408,000 recorded robberies, a ratio of 1.79:1. And as opposed to the 276,000 estimated robberies according to the British Crime Survey compared to the 78,000 recorded robberies as reported by British police forces, a ratio of 3.54:1.

So, with one-sixth the population of the U.S. England and Wales managed to have a robbery rate not four times higher, but only 2.26 times higher than ours.

In the year 2000.

Way to go England!

Oh, and our robbery rate has continued to decline precipitously. According to this report NCVS estimates show robbery fell to 630,690 in 2001, and to 512,490 in 2002. Robbery has decreased in England and Wales over the same period, though.

Maybe.

According the British Crime Survey,
In 2002/03, the number of robbery offences in England & Wales for people aged 16 and over was 300,000.

This compares with 97,000 robberies of personal property recorded by the police in the same period.

The BCS does not measure robbery offences among victims under 16 years.
However, a study of 2,000 police files found that:
22% of recorded robbery victims were between 11 and 15 years old
23% were between 16 and 20
5% were over 60
Apparently a LOT of Brits no longer bother to report robberies. I wonder how many are missed by the BCS? At any rate, a comparison of 512,490 robberies in the U.S. and 300,000 in England & Wales means the per capita robbery ratio has increased to just over 3.5:1.

Now, if you want to talk recorded crime, take a look at this Home Office paper from January 2003:
Recorded offences of robbery have risen sharply in recent years despite the fact that recorded crime overall has fallen over the same period. Between April 2001 and March 2002 robbery offences recorded by the police increased by 28 per cent. This followed a 13 per cent increase the previous year, and a 26 per cent increase before that.

--

The British Crime Survey routinely collects information on 'muggings', which includes personal robberies and snatch thefts. The latest BCS estimates that there were 441,000 muggings including 362,000 robberies.

--

Offences recorded as robbery (personal and business) by the police in England and Wales have more than doubled over the last ten years. Some of the largest increases, in terms of volume, have been in recent years.
I hope to shout! Check out this graph:

Now, this next part is really interesting:
Personal robbery accounts for the bulk of recorded robbery in England and Wales. Between April 2001 and March 2002, personal robbery accounted for 89 per cent of all robbery, and almost all of the increase. Personal robbery continues to increase at a faster rate than business robbery. Business robbery increased by 6 per cent in 2001/02 compared to the previous year, while personal robbery increased by 31 per cent.
Now, why might that be?

And if you really want to compare international recorded crime instead of estimated, there's this graph:

Anyway, I apologize for the error, Tim, and I'm glad you caught it. It's important to get these things right.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.