BBC BLOGS - Mark Easton's UK
IN ASSOCIATION WITH
« Previous | Main | Next »

Weather and crime: what did happen in 2007?

Post categories:

Mark Easton | 16:02 UK time, Saturday, 24 January 2009

My theory on knife crime and rain has been somewhat dampened by statistics sent to me by a friend and former colleague of mine, the weather expert Philip Eden.

He has kindly dug out the England and Wales rainfall stats for the two quarters in question:

July 2007 133 millimetres (232% of the long-term average)
Aug 2007 61 mm (84%)
Sept 2007 45 mm (55%)

July 2008 102 mm (178%)
Aug 2008 120 mm (166%)
Sep 2008 105 mm (127%)

This echoes the valid point made by MarkFrank in his posts (thank you for those) but I am not going to give up on my proposition entirely. July 2007 was clearly very wet, more than twice the average, and I wonder whether the weather was just too grim for people to venture down to the pub.

That having been said, Philip reckons that even in that unusually damp month, it was only raining for about 6.5% of the time in London - compared with 3.5% normally.
If anyone can lay their hands on monthly stats which relate to alcohol consumption (sales / profits?), that would be a help.

Clearly something happened in the summer of 2007 which pushed the number of robberies at knife-point down 10-15% from what one might have expected. I doubt it was an unexpected outbreak of goodwill.

By the way, for those like SheffGillly who don't believe that weather affects crime, please see my previous post on this subject.


In any case, my broader point is that the 18% year on year increase in knife crime which has been so widely reported as evidence of a worsening crime trend, may simply reflect an anomalous situation in the summer of 2007.

Comments

or register to comment.

  • 1. At 7:07pm on 24 Jan 2009, John Ellis wrote:

    Well i know from the various loons around by us that they all sort of disappear to prison at the same sort of time maybe the plain fact is that for the period in time your talking about we happened to have a good spell and all those likely to commit knife crime were inside doing a stretch.

    We see the same with cannabis every so often there is a blip and no crops appear for a month to 6 weeks, not through busts or anything like that just people waiting on the plants to finish flowering. then suddenly there is more than people know what to do with.

    maybe we had poor full moons during the months knife crime fell due to the mystical madness that is meant to accompany the full moon being very weak.

    Complain about this comment

  • 2. At 7:31pm on 24 Jan 2009, hack-round wrote:

    Mark I think there are still a few ingredients missing to make this a meaningful piece of statistical research for criminologists

    Phases of the moon could be important to the criminal mind and the tides as well as number of available cheap flights to Spain in the month.

    We could also consider the number of prisoner being released and sentenced in different categories.

    Of course we need to factor in how the gathering and reporting of the data is changed every few months by the government ministers to render the entire thing useless

    Any way sounds like we could make it job for life perhaps a good pension too?

    Complain about this comment

  • 3. At 7:41pm on 24 Jan 2009, hack-round wrote:

    Number 1 community criminal
    No plagiarism on the moon theme you were still being moderated when I wrote my piece, now what are the statistical figures for coincidence?

    Complain about this comment

  • 4. At 11:27pm on 24 Jan 2009, John Ellis wrote:

    hehe yup :)

    Complain about this comment

  • 5. At 09:09am on 25 Jan 2009, MK_Steve wrote:

    On the ‘overall’ point about crime being so low (18%?) in summer 2007, compared to the 5 other surrounding quarters:

    It didn’t look that astonishing to me (3500 crimes approx, compared to about 4000 in most quarters), I just thought it looked like normal fluctuations. So being a sceptic I just checked it, and I find that Mark is spot on; there is almost no chance at all (statistically)that Summer 2007 was a random fluctuation of the numbers (probability is less than 0.001, [chi-square based on the 6 quarterly totals = 100.7, df=5]). Hence, this ‘trough’ in the crime figures almost certainly had a systematic cause and my initial thoughts were completely wrong.

    I still like the weather idea myself.. after all rain postponed D-Day, so it probably puts the odd criminal off too! Plus it's more fun than mundane ideas about govt policy :)

    Complain about this comment

  • 6. At 01:58am on 26 Jan 2009, JapanRam wrote:

    Peston, I think you need to take a course in how to conduct sociological research rather than wildly jump from theory to theory that agrees with your preconceived ideas of what influences crime statistics. What is the basis for your research?

    1. A list of things that you think affects knife crime.
    2. Finding a similar change in any one of these factors.
    3. Concluding that this must be the causal factor.

    Thankfully, sociological research is no longer conducted in this manner.

    Here are some basics to help you get started:

    When you do find some statistics that match - such as a reduction in cinema ticket sales (how else are people to entertain themselves?), drop in steel exports (what are we to do with the excess of steel?).

    Consider your first REAL hurdle:

    To claim that A causes B, you have to take into consideration that A and B may be the cause of something else.

    Complain about this comment

  • 7. At 09:22am on 26 Jan 2009, Spiny Norman wrote:

    "To claim that A causes B, you have to take into consideration that A and B may be the cause of something else."

    Of course, JapanRam means 'the result', not 'the cause'.

    Incidentally, don't forget that each knife crime involves a victim as well as perpetrator. If potential victims are staying indoors out of the rain, then presumably crime will decrease.

    Complain about this comment

  • 8. At 09:41am on 26 Jan 2009, AJS wrote:

    On the same days of the year when the most ice-cream is sold, the most drownings occur.

    Does eating ice cream make you more likely to die by drowning? Of course not.

    But that doesn't mean there isn't a common factor responsible for both increased ice cream consumption and increased risk of drowning.

    Complain about this comment

  • 9. At 11:22am on 26 Jan 2009, JapanRam wrote:

    result*, indeed.

    Complain about this comment

  • 10. At 12:52pm on 26 Jan 2009, Eviscera wrote:

    Maybe the lovely weather made people feel less violent ;)

    Complain about this comment

  • 11. At 3:48pm on 26 Jan 2009, MK_Steve wrote:

    "To claim that A causes B, you have to take into consideration that A and B may be the 'result' of something else."

    But hang on JapanRam...

    'A' = 'the weather'
    'B' = 'UK crime levels'

    Substitute those into the statement above... and marvel at what it says!

    On this basis the only common cause I can think of is the Earth's tilt, causing seasonal change, causing weather, but also crime as a result of dark nights. But even this is very clearly discounted by the 6 quarters data.

    Hence if I were to bet on

    A. a causal link, or
    B. A common cause

    I know where my money would be!

    Therefore I have made my due "consideration that A and B may be the 'result' of something else" and come to the conclusion that they are very obviously not!

    So a 'chance correlation' or a 'complex causal link' are still on the table.

    Also, I don't think there's any need lecture others on needing courses in research, since those in glass houses...



    Complain about this comment

  • 12. At 04:45am on 27 Jan 2009, JapanRam wrote:

    "Hence if I were to bet on

    A. a causal link, or
    B. A common cause

    I know where my money would be!

    Therefore I have made my due "consideration that A and B may be the 'result' of something else" and come to the conclusion that they are very obviously not!"MK_Steve

    Well Steve, I think I have to disagree with you.

    Obviously. we are talking about the possibility of B (UK crime statistics in question) being ‘caused’ by something other than the weather (rain, actually).

    We are not discussing whether the weather is caused by something else, nor are we discussing whether the crime stats are statistically improbable (for the sake of argument lets accept they are). So, let’s consider the weather as our causal factor. Well, let’s consider rain (why?) rather than temperature, hours of sunshine, etc. Its often cold, overcast and miserable out, without it raining.

    First, you would need year on year data that was reliable, not just comparing two years. Let’s say, comparing rainfall and crime stats over the three months for the last ten-twenty years.

    Then you should analyse the data in some clearly understandable form to check if there is any apparent correlation between rainfall and crime levels. A simple scatter graph could do this.

    These would be the first steps. Rather than to propose an explanation, actually find if there was any correlation (note, correlation does not mean causal relationship).

    However, instead of first employing a valid descriptive research design, we’ve jumped straight to a piece of explanatory research comparing only two data sets!

    The ideal data probably does not exist, as the way in which crime is reported has changed (not standardized) and there are far too many influencing variables, changing over time, that we have no control over isolating. E.g. police campaigns, legislation, social factors, etc.

    Furthermore. . .
    1. A Causal Link
    We don’t over simplify the world around us to find a causal links between two pieces of data just because they match, especially through such limited data (as presented above).
    2. Common Sense
    Thankfully, sociological research isn’t ruled by peoples subjective opinions that have little (research) validity.

    Either way, your conclusion is not justifiable!

    Hmm, maybe your world is flat?

    Also, I think I should clearly point out that I am not stating there is no causal relationship, just that the research is so poor one would not be justified in basing one off it!

    What if the only variable found to match was not:

    A = 'the weather' (you mean rain?)

    BUT

    A = 'pollen count'

    You'd be happy to conclude that we should be more suspicious of hey fever sufferers?

    Another line of reasoning that makes me wary of jumping to such hasty conclusion without appropriate study is:

    What if a higher proportion of crimes were found to be committed by minorities, you would conclude they are the CAUSE, rather than, say, suffering from social deprivation?

    Complain about this comment

  • 13. At 1:03pm on 28 Jan 2009, MK_Steve wrote:

    In response to JapanRam's comments:

    I think we're of danger of getting into peer review here! I agree with much of what you say regarding this as research; if this were this a piece of research we'd need to be more systematic and scientifically robust in some of the ways you mention. But it is a natural human heuristic to seek cause, and we're usually right... hence we've evolved thus far!

    So it isn't a crime to infer and question cause (given we're not submitting for peer review here!) and in the same way that 'correlation does not mean cause', it does not elimate the possibilty either, which is what we shouldn't forget.

    Anyway I shake your hand and conclude that we'll beg to differ on this one!!

    All the best :)

    Complain about this comment

  • 14. At 04:28am on 29 Jan 2009, JapanRam wrote:

    ...And you too, Sir.

    Thank you for the lively debate.

    Complain about this comment

  • 15. At 01:10am on 30 Jan 2009, StevieK99 wrote:

    Robert,
    I think you have missed some other major issues with the data you've been analysing. Yes, there was an 18% increase in 2008 between July and September but you neglect to discuss the 10% year on year fall for the period before (Apr-Jun).

    Could these (both the fall and the subsequent rise) be in some way related? I think so;

    A good friend of mine actually compiles these statistics as she is employed as a civilian police officer - she tells me that her boss often tells her not to input the figures this week/month for whatever reason (perhaps political pressure - or in some way related to bonuses??). It may be even something less 'sinister' than that - was there for example a sickness bug in April to June which meant the civilian police had more absence through sickness? This in turn would lead to a backlog which would mean the data for the following quarter would be bound to rise as the now fit police officers return to work and are able to catch up. It would be interesting to see what happened in the next quarter!

    I guess if these guys all fell ill due to inclement weather then your rain theory may have some indirect link in this case but we can't know without further information - perhaps you should look at the 5 or 6 police areas with the biggest % increase and just interview the chief-superintendents. You often get the best insights from asking experts, especially if it's something major that caused it...

    Finally, some people are talking about statistical significance and methods but this data is not a random sample in the classical statistical sense - as I've pointed out, all sorts of issues get in the way of this data being randomly selected variables (beaurocracy, political motives, illness, etc) so classical statistical techniques may not work - the data may be too heavily biased in some way. I've worked on many econometric models in the market research industry and believe me, if you put rubbish into a model, you get rubbish out of it (or as my previous boss used to ever so succinctly say; "sh*t in, sh*t out").
    All these sophisticated brains in the City with their lovely economic models and not one of these models had been set up to take account of mass debts and the banks having their (our) money invested in what turned out to be a pyramid scheme - albeit a very technically complex one...so not one model, not least the treasury's uber-model predicted this credit crunch and the subsequent recession...
    I guess what I'm saying is, don't take any data, even (or especially?) Govt data at face value - if there's been a big change there's usually some man-made reason effecting the data. I have great faith in the ONS - but the people at the coal-face can't be watched over every minute of the day..

    Anyway, good luck with finding out the true cause behind this...
    Steve


    Complain about this comment

  • 16. At 01:13am on 30 Jan 2009, StevieK99 wrote:

    Sorry Mark,
    I think I just called you Robert in my previous message - I had just been reading Robert Peston's blog too so it was probably due to that - I guess that proves that even inputting data as simple as a name can be incorrect!!!

    Complain about this comment

  • 17. At 01:26am on 30 Jan 2009, Ben wrote:

    Mark - whilst I loosely agree with the impact weather can have on crime - I think there's one very large piece of the equation being overlooked (and I'm sure you are aware of all the other possible influencing factors). Media coverage has a vital role to play I suspect. If talk of 'getting tough', 'stiffer penalties', 'heavier policing' is front page news then gangs in particular tend to lie low for a while. Teenagers in particular know that they are being singled out and for a few weeks may go without being 'tooled up' until the dust settles - I wager (though I haven't checked) that talk of getting tough would have been prominent then.

    Ben

    Complain about this comment

  • 18. At 2:04pm on 31 Jan 2009, MK_Steve wrote:

    StevieK99 makes an excellent point regards the data itself.

    We call it 'garbage in garbage out'. Until I read his posting I had no idea the possible extent of the issue in this case.. I stupidly assumed that crime numbers were pinned to the date of the occurence, not the date the figures were entered into the system!! That makes the data utterly unusable! All it shows that there was a statistical blip in the data entry process... perhaps the data entry is done by a duck!

    Mark, it seems you have your answer (or at least a big part of it) right there. If Stevie is accurate in what he says (and I'm sure he is), then all of our postings have been rather pointless.. if fun!

    Complain about this comment

View these comments in RSS

BBC iD

Sign in

BBC navigation

BBC © 2014 The BBC is not responsible for the content of external sites. Read more.

This page is best viewed in an up-to-date web browser with style sheets (CSS) enabled. While you will be able to view the content of this page in your current browser, you will not be able to get the full visual experience. Please consider upgrading your browser software or enabling style sheets (CSS) if you are able to do so.