The latest crime statistics for England and Wales have been released. Sharon Howard has a good round-up of reactions to the statistics:

What’s happening here is that the British Crime Survey is suddenly being discounted by Tory politicians because it’s showing falling crime levels (and, indeed, has been since the mid-1990s), whereas the police statistics record increases in violent crimes (but falls in most other categories). They’ve latched onto the one category and set of stats that are of use to them. David Davis, the shadow home secretary, will no doubt have criminologists everywhere in stitches—or in shock—with this assertion: “The most reliable measure of crime is that which is reported to the police”.

As Howard goes on to explain,

when it comes to trends there should be little doubt that the BCS will be more reliable than the ‘official’ police figures. Unless it’s inconvenient for you as an opposition politician in the early stages of the run-up to a general election.

Or you’re trying to “prove” that gun control in England caused an increase in violent crime, like John Lott here and Joyce Lee Malcolm here.

This article has some more rationalizations from Joyce Lee Malcolm as to why the inconvenient BCS figures can’t be trusted:

The government now seems to be insisting that the police actually record a higher level of the crimes reported to them, but don’t seem willing to explain what proportion or how it is being done. As a result, as crime rates go up dramatically in police statistics, especially for violent crime, the government keeps saying they are not really going up, it is just that the police are using a different method of recording crime. But for some crime, such as murder, the police could not easily have been under-reporting in the past, although they do track murders to pull them from the totals if the final judgment is anything less than a court finding of murder. At this point the government has used the “new method of calculating crime” excuse so repeatedly and without explanation that I am inclined not to trust their assurances that crime is going down or remaining steady. Murder, for instance, is at the highest level since statistics were kept.

If you look in the crime report you will find explanations of the changes in the police statistics. Explanations that Malcolm pretends do not exist.

I think the international crime victimization study released in 2002 is more reliable and offers a comparison of how England and Wales are doing compared to other industrial countries. Sadly, England has many times the violent crime of most European countries. But their methods of fighting crime by disarming and prosecuting victims is so counter-productive that the results do not surprise me. Unfortunately it is in the government’s interest to demonstrate that its crime-fighting initiative is successful, which makes its assertions doubtful.

My comments on each of Malcolm’s four sentences:

  1. Malcolm does not give any reason at all why the ICVS study should be more reliable. The BCS has a sample size twenty times as large as the ICVS and no reputable statistician would consider the ICVS to be more reliable.
  2. The ICVS does show that England has more violent crime than most European countries, but Malcolm has greatly exaggerated the difference. The study shows that English rate of contact crime (3.6%) is not “many times” that of most European countries (average 2.2%).
  3. As explained here, Malcolm’s claim that England prosecutes victims is concocted by selective quoting from cases and doctoring quotes from law texts.
  4. Malcolm is a noted pro-gun advocate, so it is in her interest to demonstrate that English gun control has increased crime, which makes her assertions doubtful.

Comments

  1. #1 Shaun Bourke
    August 4, 2004

    Hilarious……

    “The survey is carried out in England and Wales. It started in 1982, and the 2000 BCS is the eighth survey. The 2000 BCS is being carried out by Social Survey Division and the National Centre for Social Research.

    One person (aged 16 or over) is selected, at random, for interview at each selected address. The interview is administered using CAPI. There are two self-completion sections at the end of the questionnaire and the respondent is encouraged to do these on the laptop themselves.”

    …..this is nothing but an opinion poll.

    And quite typical from the Left when they need to counter factual arguments.

    Tim…. and you run a computer school at the UNSW ?? boy have their standards slipped in a few decades.

  2. #2 John Ray
    August 4, 2004

    Tim
    As you know, I take only a cursory interest in these things but you seem to have fallen into a grave methodological error in the above. If gun crime is concentrated heavily in just a few areas — which it is — and we both know where — a nationwide survey will not at all adequately represent it. And nationwide surveys are not good at representing anything of low overall incidence anyway. Quite aside from purely statistical error the other sources of error (non-co-operation with the survey etc) can easily be much greater than the effect being examined.
    I stick with reported homicide followed by reported gun crime as the most believable statistics
    JR

  3. #3 John Ray
    August 4, 2004

    Tim
    As you know, I take only a cursory interest in these things but you seem to have fallen into a grave methodological error in the above. If gun crime is concentrated heavily in just a few areas — which it is — and we both know where — a nationwide survey will not at all adequately represent it. And nationwide surveys are not good at representing anything of low overall incidence anyway. Quite aside from purely statistical error the other sources of error (non-co-operation with the survey etc) can easily be much greater than the effect being examined.
    I stick with reported homicide followed by reported gun crime as the most believable statistics
    JR

  4. #4 John Ray
    August 4, 2004

    Tim
    As you know, I take only a cursory interest in these things but you seem to have fallen into a grave methodological error in the above. If gun crime is concentrated heavily in just a few areas — which it is — and we both know where — a nationwide survey will not at all adequately represent it. And nationwide surveys are not good at representing anything of low overall incidence anyway. Quite aside from purely statistical error the other sources of error (non-co-operation with the survey etc) can easily be much greater than the effect being examined.
    I stick with reported homicide followed by reported gun crime as the most believable statistics
    JR

  5. #5 John Ray
    August 4, 2004

    Tim
    As you know, I take only a cursory interest in these things but you seem to have fallen into a grave methodological error in the above. If gun crime is concentrated heavily in just a few areas — which it is — and we both know where — a nationwide survey will not at all adequately represent it. And nationwide surveys are not good at representing anything of low overall incidence anyway. Quite aside from purely statistical error the other sources of error (non-co-operation with the survey etc) can easily be much greater than the effect being examined.
    I stick with reported homicide followed by reported gun crime as the most believable statistics
    JR

  6. #6 sp
    August 4, 2004

    but how could you prove that gun crime was concentrated in some areas if there wasn’t a nationwide survey? It would be the best way to minimise bias

  7. #7 Tim Lambert
    August 4, 2004

    As my title states, the BCS shows that there has been a decline in violent crime. This is not the same as gun crime. Lott and Malcolm have claimed that gun control in England has caused an increase in violent crime. The BCS shows that this claim is false.

  8. #8 Spot
    August 5, 2004

    Since the BCS Shows that only about 40% of crimes are reported, and there were 12+ million crimes, why would you a.) consider the police reports the best source, and b.) think that this is something of low overall incidence? In what universe is 12+ million incidents a low number? Plus, the police stats over the last two years were affected by a change in recording methods, making them even less useful for the types of comparisons that Tim is making. Crime in England is concentrated in a few places (like Yorkshire, Manchester, London, and Humberside), but the large drop in crime in England over the last 7 years cannot be accounted for by saying that crime in these cities (which represent the bulk of crime in England, and therefore have the most influence on the overall statistics) has gone up, while it has gone down in the rest of the country. Furthermore, the BCS data shows that that is not what has happened.

    Does John Ray then think that we should go with the most obviously inaccurate statistics, because, as the ones that confirm his world view, they are the most believable?