Because if you don't you might end up like Tim Blair:
Big call from Tim Lambert: "Crime and violent crime in Britain peaked in the early 90s and [have] since plummeted." His source? The British Crime Survey, an annual affair which asks some 40,000 Brits if they've been encrimed during the previous year. Is it trustworthy? Depends which side you're on:
The political discussion about crime is often a numbingly boring argument about statistics. Overall crime recorded by the police seems to have risen (so the Conservatives rely on this statistic) while crime reported by the public seems to have fallen (so Labour rely on that).
...
I'd lean towards the official police figures myself (although jerked-around crime counting methods make comparisons problematic), mainly because they're, you know, official police figures. The British Crime Survey is just a survey.
Blair might as well have said "Does 2+2=4 or does 2+2=5? Depends which side you're on." It doesn't depend on which side you are on. There is a right answer and a wrong answer. Not all crime is recorded by the police. The police recorded 5.9 million crimes, while the BCS, which included crimes that weren't recorded by the police, estimated 10.9 million crimes. Obviously the BCS number is closer to the right answer.
I suspect the reason why Blair is so dismissive of the BCS is that he doesn't understand enough statistics to tell the difference between a good survey and one that is worthless. An example of a worthless survey was published by the Bulletin to accompany Frum's article advocating the death penalty.
While Australians are evenly divided on whether the death penalty should be reintroduced here, according to an exclusive Bulletin poll, more than half (56%) of respondents believe that the government should not try to persuade the Indonesian government to reduce the sentences of the Bali Nine -- which include the death penalty for ringleaders Andrew Chan and Myuran Sukumaran.
However, if you look at the full poll results, you will discover something much more interesting: 66% of Australians are male! How come? Well the Bulletin reports that the poll was conducted by CoreData. If you check out their page, you find that they pay people to do surveys over the Internet. For a survey to be representative of the views of Australians, you must have a proper random sample, where each person is equally likely to be surveyed. The CoreData sample was biased in two different ways: first it only included people with Internet access, and second, even amongst Internet users, it just sample people interested enough in doing surveys to sign up with CoreData. As a result, the survey was worthless -- the results do not reflect what Australians think. Nor are 66% of Australians male.
But if a survey uses a properly random sample, where each person is equally likely to be surveyed and has a large enough sample size, the results are reliable. The BCS has both of these properties.
If you study statistics, you can learn the difference between good surveys (like the BCS) and worthless ones (like the Bulletin's). And if you don't know enough statistics to tell the difference, you can always find someone who does and ask them. Something Blair and the Bulletin might consider next time.
It would be interesting to look at the overlap between the police data and the BCS data, especially since they show different trends. Are, for instance, more serious crimes on the rise (and driving the police numbers up, since perhaps the more serious the crime the more likely it is to be reported), whereas a drop in less serious crimes might account for the downward trend in the BCS numbers?
If the police report more rapes or homophobic crimes, is it because there are more crimes or because people feel more confident that coming forward will do some good? Conversely, if the BCS report a change in reports of rapes and homophobic crimes, that very likely reflects a change in the number of instances.
Even having a random sample is not good enough. Polls measure *perceived* crime level, not the real one. Those are two different things.
I think that as long as we want statistics to help us in debating the utility of crime penalty, which is mostly given for murder, official statistics are OK. Why? Because practically 100% of murders are detected.
This reminds me of an alleged C19 criticism of the first applications of surveys to derive social statistics.
"These numbers are not for the whole population but only for a sample. And a random sample at that!"
I suspect the reason why Blair is so dismissive of the BCS is that he doesn't understand enough statistics to tell the difference between a good survey and one that is worthless.
I'd suggest that Blair is dismissive of the BCS because it tells him what he doesn't want to hear.
As a thought experiment, what would be his reaction to the Lancet survey if it had suggested low Iraqi casualties?
Exactly right IMHO Ken Miles.
D
Polls measure *perceived* crime level, not the real one. Those are two different things.
The BCS is actually quite sophisticated and measures both; it has a set of questions on people's perceptions of the overall level of crime in their areas and a set on whether they, personally have been a victim of a given crime (and another set on whether a family member has IIRC). Obviously for crimes like drug dealing you are only getting the perceptions since few people consider themselves to be "victims" of drug dealing but for more normal crimes you get both.
Yes, but doesn't the BCS (and I may be wrong here) measure a selection of crime types, and not all of them, whereas the recorded crime figures cover a far wide selection of crime types, ergo a straight comparison of the "one is better than the other" is somewhat questionnable?
Does the BCS include data on murder, sexual offences, drug dealing, crimes against children, and crimes against business? If it doesn't, I'd be a little dubious about it reflecting overall crime rates.
Norman Dennis and George Erdos, who are a little closer to events in the UK, appear to disagree with Lambert: "Part of the problem, Mr Dennis maintains, is that the Home Office, along with a lot of academic criminologists, insist on denying the most fundamental fact about crime in Britain: that it has been, and is, increasing."
"Have you been a victim to murder?"
Now that would be a survey!
Does the BCS include data on murder, sexual offences, drug dealing, crimes against children, and crimes against business? If it doesn't, I'd be a little dubious about it reflecting overall crime rates.
Are you telling us you didn't check your sources?
Does the BCS include data on murder,
No but one hardly needs a survey to tell whether someone is dead or not and in any case the recorded crime data for murders is heading downward.
sexual offences,
yes but they don't present the figures because these offences are so rare that there aren't enough of them in the survey to make statistically meaningful comparisons.
drug dealing,
yes, albeit in the form of "perception of the incidence of drug dealing in your area"
crimes against children,
No but the Offending, Crime and Justice survey does; there are issues of comparability but it shows the same broad trend.
and crimes against business?
thank god, at least one thing in this list that I hadn't already told you about on the other thread. The answer is no, but the recorded crime figures for non-domestic burglaries show exactly the same trend as the BCS data for domestic burglaries ie down.
If it doesn't, I'd be a little dubious about it reflecting overall crime rates.
It is plain as a pikestaff that since you can't be bothered with the four clicks it takes to download "Crime in England & Wales", which compiles both the BCS and recorded crimes data, you are going to remain "a little dubious" no matter what the numbers say so why bother eh?
Norman Dennis and George Erdos, who are a little closer to events in the UK, appear to disagree with Lambert
Unlike you, I am in the habit of checking things and so I can report back that the two sentences following the passage you quote are:
"That denial is just silly," he says. "Fifty years ago, there weren't 400 street robberies in the whole of Britain. In 2001, there wasn't a single month in the borough of Lambeth that is just one London borough in which there were fewer than 400 street robberies.
Few people would deny that there is more crime in the UK now than there was in 1956, but fewer would deny that you really are clutching at straws in a most hilarious manner by pretending that this quote supports any argument about crime in the 1990s.
"Blair might as well have said "Does 2+2=4 or does 2+2=5? Depends which side you're on."
Unfunny joke designed to take one's mind off the waffle that follows.
"It doesn't depend on which side you are on. There is a right answer and a wrong answer."
Not really.
"Not all crime is recorded by the police."
True as far as it goes but not all crime is included in the BCS. If I steal a DVD player from a shop sell it to buy drugs and then slash all the tyres at the bus depot, none of this will show up in the BCS. If Tim Lambert and Tim Blair were to visit London and be beaten up by an Aussie-Tims hating maniac this would not show up on the BCS. There are several types of crime which don't show up on the BCS- http://www.civitas.org.uk/data/crimeFigures2004-05.php
"The police recorded 5.9 million crimes, while the BCS, which included crimes that weren't recorded by the police, estimated 10.9 million crimes. Obviously the BCS number is closer to the right answer."
What a ridiculous statement, obviously the higher figure is closer to the total crime figure, but seeing as it is the trends that are being discussed, it is irrelevant. What exactly is "the right answer" seeing as the two figures measure different things? Is the BCS closer to the total number of crimes against individuals who are adult residents of the UK? Yes. Is the recorded crime figure more representitive of the total number of crimes that people thought was worth reporting to the police? Again yes. It is comparing apples and oranges, or at least satsumas and oranges.
"But if a survey uses a properly random sample, where each person is equally likely to be surveyed and has a large enough sample size, the results are reliable. The BCS has both of these properties."
Really? Seeing as the response rate for the survey is not 100%, (I think it is closer to 70% but I don't have the figure to hand), it cannot be assumed to be truly random. As James Q Wilson has pointed out it is plausible that the people least likely to respond to crime surveys may be the most likely to be victims of crime, if social isolation makes one more likely to be a victim and being a victim makes one more likely to be isolated, and those who are socially isolated are the least likely to respond to crime surveys then the figures produced are going to be flawed. This isn't a criticism of the BCS there is nothing that can really be done about it as the effect cannot really be measured, but it does prevent it from being truly random.
The BCS is a useful tool, but to elevate to the status of a comprehensive crime figure is absurd, and to pretend that it is obviously more reliable that the recorded crime figure smacks of straw clutching.
tim, even the quote you give undercuts your position: "the Home Office, along with a lot of academic criminologists" disagree with Dennis. Maybe professional criminologists know more about criminology than Dennis? The Telegraph reporters doesn't seem to have bothered talking to anyone who disagreed with Dennis to find out why. Here is an article from a reporter who did a better job:
So they are psychologists, not criminologists, were published by a right-wing think tank rather than a peer-reviewed journal and blame things on the collapse in moral values.
They conceal the decline in crime in the past ten years by comparing rates with those of 40 years ago. They don't tell you that crime trends in the US and the UK have been parallel. In both places crime increased in the 60s, 70s and 80s, peaked in the early 90s and has since declined markedly.
Oh and do you think that 66% of the Australian population is male? Maybe the Bulletin could make a correction?
I think I've dealt with this before on my own blog - yes, the BCS doesn't include murders, no, this isn't significant. Think about it - if you claim that the BCS is all wrong because it doesn't count murders and crime is really going up, there must be a truly humungous number of murders around.
Enough to turn around a trend in a sample of 10.9 million! Bodies must litter the streets! What's more, as the police reporting figures don't show a huge murder wave, the streets must be littered with unreported murder victims!
In order to say with some precision "crime is rising", one must define what number he considers to rise. Lumping all possible crimes in one category and adding them up is not wise.
Example:
Case A: five armed robberies and five rapes
Case B: ten pocket thefts and ten shopliftings
Which case has "higher crime"? Case A is only 10 crime incidents, and case B is, oh my God, 20!
So, what weights should we attribute to specific crimes?
Another comment: suppose we somehow define what we mean by "crime level" and come up with two ways to estimate it (like BCS and police statistics). Let's call them A and B, again. Suppose that A is closer to the true value of "crime level" than B. Still, the rate of change of B may be *closer* to the rate of change of the true value than the rate of change of A. Anyone who knows what is the derivative of a function will easily understand that. Thus, arguing that the trend of A will necessarily reveal the trend of real "crime level" than B, just because it comes closer to the *value* of the "crime level", is completely illogical.
Roman, if the percentage of crimes measured stays the same, then both A and B will reflect tends in the real crime rate. But if that percentage varies then they won't. B, measuring a smaller percentage has much more scope for change. Consider what happens if B goes from 25% of crimes to 50% and compare with A going from 70% to 95%.
In this we do know that a larger fraction of crime is being reported and recorded because recording practices have changed, while the BCS questions have not been changed.
Tim,
I am afraid you have a bit of an error. The critical question for a survey is not whether "each person is equally likely to be surveyed" - but whether one can specify the probability that any given person will be surveyed (in the language that I was taught, that it is a "probability sample") - and whether final results are appropriately weighted to reflect any variation in those probabilities. Many samples in the United States routinely over-sample some relatively small groups in order to get reasonable estimates of something within those groups (e.g. political attitudes among African-Americans), and those surveys are, as I understand it, perfectly legitimate.
This doesn't dispute your main point (after all - no one can specify what the probability that particular people will sign up for CoreData's program is) - but it does matter. There were a few minutes as I read your post that I started flashing back to some of the comment threads about the Lancet study, where some commenters seemed to basically be insisting that any sample that was not a Simple Random Sample (where not only every individual, but every possible sample was equally likely). Yes, an SRS is computationally easy to work with, and either an SRS or a "uniform probability sample" (I am not sure this is a standard phrase, but it seems to be a decent label for the sample you describe) is easier to explain in some ways, but much important work can be done using other sampling techniques.
Thanks for all the work you do on this blog.
"The blame for the rise in crime, according to Policing in Four Nations, is a collapse in moral values in the 1960s."
Given the time frame, especially considering the gradual fall in the last decade, one could just as easily blame it on the rise of Thatcherism. Just kidding.
Yes, the probabilities don't have to be equal if you weight the results appropriately. I was trying to keep the explanation simple.
The BCS is a useful tool [no problems here], but to elevate to the status of a comprehensive crime figure is absurd [i'm with you so far], and to pretend that it is obviously more reliable that the recorded crime figure smacks of straw clutching [um...uh...hmmm]
the issue isn't whether the BCS is perfect, it's whether it's better for identifying overall trends than the recorded crime figures. your post only demostrates why the BCS is not perfect. it does not provide an explanation for your view that trends are better supported by reference to the recorded crime figures.
hilariously, your post goes to some length to show why "social isolation" may discourage victims of crimes from participating in surveys, but you seem not to have noticed that this would also [and to an absurdly greater extent] discourage victims from reporting crime to police.
I don't know what Tim Blair is complaining about or what Tim Lambert is gloating about. There has been a fall in crime during the nineties but this vindicates cultural conservative social ideas. This, paradoxically, makes cultural conservatism less applicable for the time being. Thats why support for the death penalty is dropping.
A psychologist can be a criminologist, just as an economist or computer scientists. Basic scientific principles are portable between disciplines.
Civitas are in general, a pretty good outfit. They are not really right wing. More like first generation neo-cons ie liberals(social democrats) mugged by reality. They certainly got the critique of multiculturalism right-on.
Over the past generation a right wing think tank's social analysis is more likely to be true because, not in spite, of its right wing-ness. The Manhattan Institute, sponsor of the unspeakable Charles Murray, was one of the first right wing think tanks to concentrate on social affairs. It's critique of liberal social policy was certainly miles ahead of the curve of "peer-reviewed journals" at the time, much to the disgrace of the predominantly Wet sociological profession.
Nowadays the MI's scientific analysis and policy conclusions are now pretty much the conventional wisdom accross the political spectrum. Clinton paid very close heed to its type of thinking and Guiliani swears by it.
More generally, crime rates escalated from the sixties through nineties largely because, as Civitas says, there was a "collapse in moral values" as expressed in behaviour. The minority-empowering cultural revolution caused a massive disruption in social structure. Democratic states were not immune and endured a frenzy of libertarian cultural and financial deregulation.
Instead of crime being a white, male, under-class adult thing it suddenly became all the rage for every minority group - teens, women, immigrants, indigenes, entrepreneurs etc - who were by and large able to get away with it scot-free. We had equal opportunity in crime and therefore more of it.
Then, through the eighties and nineties, voters and politicians started to take notice of what "right-wing think tanks" were saying by tightening welfare, more selective immigration, tougher sentences, more police, zero-tolerance etc. Also, many criminals and their kin killed, maimed themselves or got older and wiser after a period of inside. Result: less crime from the nineties through naughties.
In short, the fall in crime proves that Anglosphere cultural conservatives won the Culture Wars: first in theory and now, with the tough-on-crime related collapse in crime, in practice.
Jack,
I would have thought that crime is lower mainly due to lower unemployment, wealthier societies, and an ageing population (ie less young poor people). I'd also give better policing it's share of the credit but not because of 'zero tolerance' policies which didn't exist outside of a few places in US but because of the increase in the use of technology from DNA to computer databases as well as better training and support services. In Australia the police busting of the main heroin importation ring had a dramatic effect on property and drug related crime. It wasn't tougher sentencing or selective immigration that lead to the heroin drought but good police work.
SimonC March 9, 2006 01:48 AM
Not necessarily. The post-modernising world during the sixties saw sensational improvements in prosperity, particularly for the newly-empowered minorities. They were also a time of rapid increase in crime rates, particularly amongst the newly-empowered minorities.
The nineties through naughties have certainly been prosperous. This has probably helped reduce crime a bit, especially when combined with tighter, mutual obligation welfare and pro-active paternalistic social work.
SimonC says
Yes. Older people are less prone to commit crime. Improvements in health and medical technology have caused more people to live longer and have thus brought down the overall crime rate. This is an apolitical trend which no ideological side can claim credit for.
SimonC says
Beside the point. Crime in general, including drug crime, has gone down due to tougher sentencing.
Unselective immigration increases the crime rate over time, through generational "cultural degeneration". First generation immigrants, whether illegal or legal, tend to be old-fashioned and keep their heads down. A Harvard study showed that second and third generation children of poorly-selected, or poorly-settled, immigrants tend to be more way-ward as parental controls are loosened. Particularly in the context of a more general "moral collapse".
So more selective immigration will tend to bring down crime over time.
Target hardening has also reduced crime. We have much more private security, fences, locks, cameras etc than we did a generation ago. This all costs money.
The rising number of heroin traffickers who have been jailed under tougher sentencing laws means less traffickers on the street. The heroin market was knocked by more authoritarian regulation - Salvation Army Major Brian Watters ran the anti-drugs taskforce. As Miranda Devine says, it pays to be tough on crime:
The stepped up War on Drugs is one of the great unsung achivements of the Howard government. His policies here have saved countless lives and he has gotten little credit from "cultural workers" who have been clueless on cultural policy. Cultural conservatives should be bragging about this. I think that Tim Blair is a libertarian, perhaps he would care to enlighten us.
SimonC says
True, but this proves my point. Cultural conservatives love good police.
SimonC my lengthy reply to your comment seems to have been swallowed without trace by Deltoid's spam filters. Perhaps Tim Lambert might find it lying in the bowels of his server.
So, TimL, the best you've got is an appeal-to-authority argument and another mention of a survey which is nothing to do with the matter in dispute.
If you're not going to try, why bother playing?
errr, tim, it was *you* who appealed to authority. Tim L's point was that your appeal to authority was in fact an appeal to a non-authority, and all the real authorities disagreed with him.
Have you any idea how silly that looks?
In addition to dsquared's points, I should not that I also explained why Dennis and Erdos were wrong. tim somehow missed that bit.
And the other survey is quite relevant, since my post is about reasons to study statistics. CoreData pay $2.50 per survey you do, so with 2000 responses it must have cost the Bulletin at least $5,000 for a worthless survey. If you know more statistics, you could have saved the money.
" your post only demostrates why the BCS is not perfect. it does not provide an explanation for your view that trends are better supported by reference to the recorded crime figures."
My view? I did not express a view on which way crime in the UK is going (it is irrelevant, but my own belief is that overall crime is falling but certain categories of crime are rising, and the government is more responsive to those which show up on the BCS). Simply that the BCS has as many faults as recorded crime figures and therefore it clearly wrong to claim that it is obviously more accurate. Where the two figures conflict it is as Tim Blair said a question of which side you are on as to which figures you rely on.
"hilariously, your post goes to some length to show why "social isolation" may discourage victims of crimes from participating in surveys"
If you find one sentence to be lengthy it may explain your inability to understand what I actually wrote.
My view? I did not express a view on which way crime in the UK is going
pardon my mischaracterisation. i had assumed you were saying the reported crime statistics were more reliable than the BCS figures for trend purposes, when in actuality you were saying they are as reliable as each other.
anyway, the point is that the only reason you offer in support of this opinion -- "social isolation", which you say leads to an unwillingness to participate in surveys -- has a far greater [detrimental] effect on the reliability of recorded crime figures. think it through: is "social isolation" more likely to lead to (a) an unwillingness to approach legal authorities, or (b) an unwillingness to be approached by surveyers? it's not a hard question.
anyway, it's not simply a matter of flipping a coin to decide which figures you should go with. the BCS figures are more reliable. that tim blair thinks you should go with the figures which better support your pre-conceived view is a comment on his intellectual seriousness, and nothing else.
also:
As James Q Wilson has pointed out it is plausible that the people least likely to respond to crime surveys may be the most likely to be victims of crime, if social isolation makes one more likely to be a victim and being a victim makes one more likely to be isolated, and those who are socially isolated are the least likely to respond to crime surveys then the figures produced are going to be flawed. This isn't a criticism of the BCS there is nothing that can really be done about it as the effect cannot really be measured, but it does prevent it from being truly random.
i count two sentences.
As if Blair is interested in any real discussion of the issues. His latest offering is http://timblair.net/ee/index.php/weblog/comments/guns_do_good/
in which the evidence consists of two anecdotes : the guy is a laugh riot.
Maybe Michael Mann should learn statistics as well!
Regards
Peter Bickle
Tm Blr strtd thrd bt cnflctng K rprts n crm, shwng tht n (crm s dclnng) ws ndrsd b Blrt (WMD) Lbrts, nd th thr (crm s ncrsng) b Cmrn's Trs. Tm Lmbrt's hjckng f ths tpc shws () h s plgrst, () dctfl, b mplyng tht Blr ws Tr mrl b cmmntng n ths (h m wll b bt ths ln s nsffcnt vdnc), nd () tht h s lckng rgnlt n trms f strtng thrds. s thr n hnr mngst blg hsts?
Instead of crime being a white, male, under-class adult thing it suddenly became all the rage for every minority group - teens, women, immigrants, indigenes, entrepreneurs etc - who were by and large able to get away with it scot-free. We had equal opportunity in crime and therefore more of it.
*wiping away tears of laughter*
I was wondering why all me mates were suddenly out burgling.
As Helen has reopned this thread perhaps I could explain why my last was put into esperanto or some such by Dr Lambert - it seems he did not like my sugestion that he had in effect plagiarised Tim Blair's blog, or at least taken over TB's posting on UK crime statistics to start his own.
"Instead of crime being a white, male, under-class adult thing it suddenly became all the rage for every minority group - teens, women, immigrants, indigenes, entrepreneurs etc - who were by and large able to get away with it scot-free."
Damn entrepreneurs!!!!
why should people in accountancy understand statistics
why should people in accountancy understand statistics
the answer i got from you is not clare, please can you make it more clareer for me. thanks