Oh dear!

I knew it would go wrong the moment I heard it. “We have taken 513 police officers off the streets.”

It was at the Lancashire Police Authority Meeting last week, where the Acting Chief Constable was running through the latest crime figures, and having given the fairly good 12-month comparison had decided to release a comparison of the one single month that had finished in this year, which was less good, leading to the speculation as to why that might be.

I was sitting in what I will call the spectators’ section, and my senses may have been heightened by the presence of the local reporter who had taken a seat just in front of me, as his pen jumped to life, matching the actions of several round the table, which included two-thirds of the local Labour PCC shortlist.

The Jubilee weekend would keep the story away from the front cover till today but, as you can see, not forever. And then a tweet or two and the help of the Press Association, and before you know it, it is being quoted by the Shadow Policing Minister, and transformed in no less an organ than the Telegraph to say “Burglary and violent crime is rising in some parts of the country because of the swingeing cuts in police numbers, a police chief has warned.

Surprisingly, what didn’t make the story was his anecdote about people shoplifting meat, bread and cheese and how this was “a commentary on social pressures”, so that one member of the authority expressed a concern that they wouldn’t like to see the criminalisation of people committing crime out of necessity. Perhaps reporting all that would have made the meeting sound less like a Police Authority meeting and more like a Party Political Broadcast on behalf of the bleeding-heart-liberal party?

My mind went in several directions, to half remembered scenes from films depicting harsh punishment of starving French peasants, to my undergraduate self reading All England reports stamped ‘from the library of Mr Justice Diplock’ as I tried to understand what was required to fulfil the demands of the legal defence of ‘necessity” or ‘duress of circumstance’. But mostly what filled my mind was the thought “Oh dear!” as, whether intentionally or otherwise, the Acting Chief had become the pin-up of every wannabe Labour Police and Crime Commissioner in the country, including tacit endorsement of the campaign slogan of one sitting opposite him.

And they say policing is not political.

You may imagine that last week a whole host of requests for clarification found their way from yours truly to the Constabulary, and when that information is received, I will be sure to give you the highlights. Suffice it to say for now that the varying levels of crime, whether that variance is up or down, is neither neatly or solely related to the number of cops there are, and there are some categories of crime that should increase if the police are doing a good job (see the third bullet here).

There was a time when I would have been concerned about the impact that this would have had on fear of crime locally. This was before I realised that most people only believe crime figures when they are rising, and that how they feel about the level of crime is little to do with what is reported at a Police Authority meeting.

It was a shame though about the soundbite. I feel the real news at the meeting was missed – that after years of never doing it, the Authority had trained a number of members in how to handle pension forfeiture cases, as a number were expected before November. Now there’s a story.


This entry was posted in Perspectives and tagged , , , , . Bookmark the permalink.

13 Responses to Oh dear!

  1. Sam – what is really frustrating is how hard it is to find the real performance data.

    It was not attached to the PA papers – and instead (I assume) was part of the Acting CC’s oral report. When I google ‘Lancashire Police Performance – the first item that comes up is this: http://www.lancspa.gov.uk/news/?ctid=200&edid=1967 – which is an undated press release – but on closer reading refers to 07/08 data. Beyond that – I can find nothing very easily. All that I find is either projections or data that goes back too far.

    So lots of heat – and not a lot of light.

    And we should be tracking trends in performance closer. I have been advocating the use of Statitistical Process Control methods for years (and years) – and whilst there are elements of this in the iQuanta data from the Home Office – it is still not presented in a way that an SPC practictioner would recognise or approve of. With SPC it would be relatively easy to determine whether the factors highlighted by the Acting Chief Constable were inflcting a significant change or not.

    This is not a definitive guide to SPC – but as a ready reference it will do: http://www.flowmap.com/documents/booklets/spc.pdf (particularly as he references three books by Mal Owen – who is a top chap in the field).

    The key question is whether the last month’s data (and as reported widely) was an ordinary (or common cause) variation – or a special one. Did the performance go beyond the ‘control limits’ for example? The co-incidence with a reduction in police numbers may or may not have been behind this special cause variation – if it was special. Perhaps it was the wet weather? (Oh hang on – we are in Lancashire – you always have wet weather!) As we all know (well – apart from the Daily Mail) – correlation does not prove causality.

    But we just don’t know without a whole of heap of analysis which may or may not be happening – but certainly does not seem to be published accessibly (but if someone has the link…)

    And so we are left with rhetoric and newspaper headlines. I guess my question back to you Sam – is had the Actng Chief said that whilst the force has suffered huge losses in police numbers – it was making no difference to crime levels – would you or the Government have promoted that just as much?

    • samchapman says:

      I can’t speak for the Government, but I did a report from the last Police Authority meeting. Something about squirrels, if I remember rightly. You are right that sometimes stats are hidden in oral reports which makes them inaccessible.

  2. And just to add – SPC has it roots in the Deming approach to management as used by Toyota and many other businesses around the world. Deming’s 14 points are still as relevant today as they ever were – see http://www.ifm.eng.cam.ac.uk/dstools/process/deming.html – they are worth a read if you have not read them before / recently.

    Note (for example) that Deming does not agree with a) targets b) performance related pay c) management that uses fear.

    Nine years ago, I published a piece on a backwater of the internet against the use of targets – my old employer then (Office for Public Management) was not prepared to let me publish such dangerous heresy under their banner. But the article is still there – and I stand by it:


  3. samchapman says:

    Reblogged this on Sam4Lancs.com.

  4. Colin Skelton says:

    Dear Sam

    I sympathise with your sceptical view on the reporting of this item. Could we find a correlation between the increase in reported crime and the reduction in Police numbers? With the data available the answer is probably no. And even if we did in this particular case correlation does not always mean, “cause and effect”.

    However, perhaps we should be asking a wider question, does reducing police numbers cause crime to rise? My personal view is yes and the available research evidence also suggests this. The responsivness to crime to police numbers is sigmoidal i.e. “S” shaped. So at one extreme you have very few police and crime is rampant. This is what happened when the Finnish Police went out on strike. No Police, crime through the roof. At the other extreme, once you have a Police officer outside every front door, doubling the size of the force won’t effect the crime rate much. But in the middle, crime does respond to increasing or decreasing Police numbers (to a greater of lesser extent).

    As with everything the devil is in the detail, all the evidence from the US and the UK over the past 30 years shows that extra Police numbers used well and intelligently can have a significant impact on reducing crime. In the UK we are not too bad at using Police Officers as the research suggested (although this is patchy). The bad news is that once these officers are removed, they stop doing all the good things that the research told them to do and the result will be an increase in crime. An example may illustrate this point, the latest research I found on neighbourhood Policing found that well Policed neighbourhoods (the article describes this in detail) had a 12-14% lower crime rate when compared to the control areas (control areas had no neighbourhood Policing). So when faced with a 20% budget cut I would suggest that most new PCC’s will have to redeploy NPT officers to response and so lose neighbourhood Policing. The logical result will be a 12-14% increase in crime.

    One final point would be this, Bill Bratten the fabled New York Police chief takes a lot of credit for driving crime down in New York and cites new practices, COMPSTAT, driving out paperwork and the such as the reason why crime fell so sharply in the 80’s. What he never mentions is that he employed over 7000 new Police Officers. Perhaps that had something to do with it? When he left, Police numbers fell by over 4000 and crime went up. Again, you have to be cautious just looking at this one example but it was a pretty good experiment examining the effect that Police numbers had on crime in a major urban area.

    I realise I have cited research in this post but not provided the references, I have not done so because its difficult to do in a post but if anyone wants the cited articles I will be happy to suppy the references.

    Kind regards

    Colin Skelton

    • samchapman says:

      Many thanks Colin, that’s very helpful.

      P.S. I wonder how many police officers will now be seeking to insert the word ‘sigmoidal’ into interviews and meetings in the wake of your comment.

  5. bunnyson says:

    I have no idea what has being happening in Lancashire since I live and work to the far south.

    There is a big difference between actual crime, what is reported to the police and crucially in this performance-driven era, what the police record. One senior ACPO officer on landing here referred at a meeting to the ‘the problem is the over-recording of crime’ and within months crime figures started to drop. To illustrate there was a series of “car-jackings”, recorded as assault with intent to rob, robbery, assault, theft of motor vehicle and attempt robbery.

    When police officials claim a reduction in crime of 15% plus I have my doubts, especially if the rise is sustained and cannot be attributed to a prolific offender returning to his ‘work”.

    A good indicator of police performance comes from looking at self-initiated work, a couple of examples arrests for Going Equipped (instruments to commit crime), seizure of suspected stolen vehicles and proper Stop & Search – which leads to an arrest. If they are all absent then you have a problem.

    For sometime now anecdote has suggested several metropolitan police areas have seen a rise in crimes of dishonesty, so the Lancashire reference to shop theft of basic foods is of note and the discrepancy between insurance and police figures for commercial and domestic burglary. Yet where I live crime is falling.

    The big change Bratton made in NYPD and I cite a comment by a long-serving, respected NYPD officer ‘When Bratton arrived 10% were motivated to work hard and get results. When he left it was 20% and that is what made the difference’. Yes there had been a rise in police numbers through merging three departments into one (Housing & Transit), more officers and staff were recruited , better courts and jails Etc. Before he arrived NYPD stopped doing labour-intensive tasks, notably gambling enforcement and that had absorbed thousands of officers. Note little is said here about his time in LAPD oddly.

  6. Colin Skelton says:

    Dear Bunnyman

    The problems you mention are real. How crime is reported, how its defined, how senior officers massage the figures, all affect the statistics. Personnally speaking, the British Crime Survey, although far from perfect, has at least been consistant.

    Regarding Bill Bratton and crime in New York, a Prof. Franklin Zimring at the University of California wrote a article called “How New York beat crime”, printed in Scientific American (April 2011). The main upshot is that more Police, used in an intelligent way (which Bratton managed to do) was a significant part of the crime reduction strategy.

    I don’t know how Bratton performed in LA. I might look that up.


    Colin Skelton

  7. Fascinating discussion as ever. The British Journal of Criminology just published a review of COMPSTAT (Compstat and the New Penology, Willis & Mastrofski, Jan 2012) which contains some of the answers to your questions. Once you get social scientists involved, things always turn out to be different to what they appear to be.

    I’m afraid I don’t share John Harvey’s optimism for Statistical Process Control in this context. The data recorded by the police is still not good enough to establish reliable control limits – look at the blatant discrepancies between direction of movement in force statistics and the British Crime Survey. I have yet to see a single confidence interval in any of the monthly force statistics published in North Wales. NB watch out for displacement effects before getting too excited about the results of neighbourhood comparisons Colin!

    When I queried a newspaper report quoting our Chief Constable saying that immediate response times had not changed following his introduction of Response Hubs in May 2011 and that this had been “independently tested”, I was told that what this actually meant was that someone had performed a t-test on independent samples of response times and found no statistical difference.

    Trouble is I just don’t believe that either the Chief Constable or a highly qualified crime analyst could possibly obtain this result from a t-test on 2 x c. 15,000 datapoints – there is bound to be a statistical difference with samples this big (whether the difference is significant in operational terms is another matter). Later on I found that most of the data for May 2011 was missing for operational reasons. Much later when everybody had forgotten what Response Hubs were for, the Chief Constable told the Police Authority that the decline in % incidents attended within 20 minutes from over 90% to under 85% was anticipated. So you can’t win.

    Your best strategy as a prospective PCC in this election is to regard all crime statistics as propaganda!

    • I would never claim that SPC is a panacea – but I do maintain that the ability of SPC to plot trends in real time has much to offer the scrutiny of what works and what does not in this overall context. There is of course a huge amount of noise to contend with and many hidden factors that would make interpreting such trends problematic. I am no statiistician but I understand that tests like a t-test are great for comparisons between one instance and another – but no so good for moving data.

      And of course – we come back to ‘drive out fear’ (one of Deming’s 14 points) – without this – stats will be little more than propaganda as you say. I am reminded of a colleague who was once visiting a biscuit factory which was using SPC to monitor their production and he went up to one of the operators he knew and asked her about her charts – suggesting that everything was running tickety boo. She replied “is it heck! I just put the dots on the sheet to please that bloke in the white coat over there – if they are not in the right place – he gives me grief” etc…

      So I think we can conclude – we need better use of data in policing – indeed we need more evidence based policing – to show whether intervention x works better than intervention y. But there is a whole heap of organisational development needed to get to that point (hence my twenty point plan….)

      As an aside – I was working with one police force once and I heard they had a very useful discussion about driving out fear and their blame culture – and then one person interjected “but whose blommin’ fault was this blame culture in the first place!”

      It was a funny – but serious point.

  8. Colin Skelton says:

    Dear Jon and Richard

    I find myself agreeing. Police and crime data needs to be improved and better focused, otherwise we can never learn from that data or worse, we learn the wrong things. I’m a scientist by training and I was interested in the fact that a t test was done on some response time data. As ever its not that simple, what type of data it was would determine the type of statistical test being conducted (was the data normal in distribution, parametric or non parametric). The students t test is the simpliest type of statictical test and was probably not the right one to use. However, having said that, even if a significant difference was found, so what? Data can be statistically different but that can have no operational relevance or benefit.

    I feel the problem here is one of senior people not understanding data, how to use and interpret it and a general lack of understanding of science (social or otherwise). If we had senior officers who really understood these issues, we would make a significant dent in crime. So I agree with Jon in his second from last paragraph but I disagree that this needs to be a big organisational change; rather senior officers need to employ the odd scientist or two and listen to their advice. All big givernment depts have a chief scientist, could Police forces have the same??

    Or perhaps even better, when developing policy (as Jon says) it just needs to be evidence based.

    Kind regards

    Colin Skelton

  9. Dear Both
    Wow it’s still possible to have a non-dumbed down discussion on crime statistics in the UK. However I fear, John, that you may still be in danger of rushing in with SPC where even statisticians fear to tread. The police face much more basic challenges in recording crime – for a start they only record about 25% of it. So any trends on your SPC chart are actually more likely to be the result of changes in police recording practices, counting rules, public reporting etc.

    In fact it’s worse than that. If you introduce policy X you will find that recording practices around X suddenly improve because staff know that X is of interest. Probable result : an apparent increase in recorded crime due to X on your SPC chart, even if policy X actually reduced crime a bit. Consequence : evidence-based policy, but sadly the wrong one. Beware the Hawthorne effect!

    But let’s stop talking in abstract terms. Suppose X=Police and Crime Commissioner. I think you get the picture – I’d plan for a ‘surge’ in crime in your area after 15th November if I were you. (I’m planning to try and retain … how can I put it, commissioning oversight .. of what goes on in the Joint Control Centre for a while for precisely that reason!)

    The question of whether X is better than Y can only be conclusively determined by an experiment, preferably one with a randomised design (e.g. Randomised Control Trial), which is of course notoriously difficult to do in criminal justice, and there are hardly any reported. Since electronic tagging has been in the news today, you might be interested in a rare Swiss RCT of tagging versus community sentences also reported in British Journal of Criminology (Killias et al. 2010) that found only weak ‘evidence’ in favour of tagging (p<0.1 i.e. 1 in 10 chance that the result is a fluke). But this finding wouldn't even be considered as evidence in a field that takes the evidence-based approach seriously, like medicine, which uses smaller p-values (at least 1 in 20 etc.).

    However I do agree that the standard of police data capture needs to be radically improved until it is good enough for routine and continuous internal evaluation purposes. Here's a little trade secret for you : I'll be using Sequential Analysis methods (originally designed for testing the quality of munitions in the war!), not SPC.

    Colin I agree about operational relevance, but you've given me a brilliant excuse to bore the pants off TopOfTheCops regulars by talking about the Central Limit Theorem (do you remember that from your scientific training?!). You do? Excellent. Basically you will recall that if you've got more than 10 observations in each group, you can perform Student's t-test on the mean response time, whatever the underlying distribution, parametric or otherwise. My point is that the confidence interval on a mean response time with n=15,000 response times over a 4-month period in each group is going to be tiny, and it is therefore just not plausible that no statistically significant difference was detected by North Wales Police if average response declined from 91% to 85% within 20 mins. On the other hand maybe I'm over-estimating their powers of detection. (Groan!)

    The propaganda element is the thing that really bothers me – how come X gets reported as being no different from Y in the local press, and "independently tested" into the bargain? Sounds good doesn't it?!

    Richard Hibbs
    Independent Candidate for Police & Crime Commissioner in North Wales

  10. Whether we are talking Stattistical Process Control, Statistical Quality Control (Shewhart’s original term I believe – maybe it was his student Deming who changed it to SPC), Kanban, Six Sigma, Taguchi, Quality Function Deployment etc etc… (my background is in TQM too…) – the principle is the same:

    We need policing practice that is more evidence based and less anecdotal or (dare I say this) rhetorical or political!

    And you are correct, there is a lot of noise in the system, and a lot of potential for experimenter effects. You are correct, that we need a greater body of knowledge that is based on RCT’s – but, I would argue, not exclusively so. While RCTs offer a pure Popperian (Thatcher’s favourite philosopher you may recall) approach – it has its limits. But without wittering on about whole systems, commercial biases (as Ben Goldacre frequently highlights) and psychology, I will just say we need more science in policing, more humanity in politics and good deal less ‘cost of everything, value of nothing’ accountancy methods that seek to turn the world into one Matrix like spreadsheet.

    But what do I know…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s