Starting Again On Public Sector Statistics?

Talk on police or any performance management ‘statistics’ often turns to their gaming element. I live in a supposedly ‘beacon’ council area with departments like social services ‘achieving’ three-star status and ‘sector-leading’ housing association that ‘replaced’ the Council’s role in housing. In reality, I live on the edge of a dying town and doubt much round here is being done well.  I have seen improvements in the NHS care I need, but most of the rest doesn’t seem adequate and often misses the boat altogether.

The term ‘statistics’ is associated with science, but what we are getting is performance management and there is little science in them.  Even in academe, statistics are routinely fiddled because on one of its performance management criteria – the peer reviewed publication.  Simple stats are used when more complex forms are needed to establish genuine statistical differences.  Ben Goldacre suggests this is the case in half of papers in psychology.  I believe the situation is worse – as most papers with statistical method in them don’t need it at all – there is no point establishing statistical significance in the bleedin’ obvious.  Whist academics write this spurious drivel they ain’t doing what we need from them.

It’s difficult to justify the keeping of annual lists of the numbers of certain crimes if there isn’t much we can do to reduce, prevent and detect the crime people don’t want to experience.  Steve Bennett has done more than any academic I know to point out the gaming problems and we lack a forum to get the real problems expressed in a manner that could really count.  Indeed, we conflate critical evaluation methods with criticism of the police and other bodies.  This runs across the board in our society.  Modern research, based in experiment, has found that whilst we lay claim to welcome innovative ideas, we actually despise creativity, and those trying get the ad hominem in the neck.

The banking industry is a glaring example of the kind of false-accounting that is going on.  The ‘maths’ involved becomes a way of hiding problems instead of expressing them.  We clearly have a system that loses money hand over fist, yet lays claim to have expertise that must be rewarded for its performance.  I’m sorry, but you can play out of your skin and have to take losing pay.

I think we should stop performance management techniques as far as possible and ban the labeling of such as statistics.  Across industry they don’t keep people honest in the way a tackle count (now a complex of modalities) does in rugby league – the tackles become virtual and derivitised to the point we can’t be sure any were ever made.  It’s a bit like telling your coach you didn’t make any in the game, but the hundred on your pillow or in dreams should count – or including toe-nail clipping as an operation.

It is important to know that burglaries are falling, but it’s much more important to establish why this is the case – what any drop or rise is correlated with.  Beyond this, burglary is poorly defined in terms of what matters to me and you, as many other things affect our wealth.  I’m reasonably insured against it and it isn’t a major threat to my wealth and well-being.  The banksters have been a much greater threat to our wealth and democracy. Having to live near druggie, noisy, recidivist scum affects the quality of life of those it’s forced on much more seriously than burglary threat.

Good statistics would be expressed in spreadsheets that anyone could use in promoting public debate.  This is rare.  One thought that crosses my mind is whether the reductions in certain major categories of crime across the West mean anything much beyond us being able to do ‘something’ through focus on these problems.  One problem is that we have no control to measure against, but the lack of a convenient world that otherwise stands still outside our interventions is unavoidable.

One can imagine experiments we could do on typical petty crime.  We could give Bill the burglar £50K a year and send him to university, or even do this with ten evil poor families on one estate and compare them with similar areas with no intervention – I take it a sour taste is arising in you too.  This sounds ludicrous until one realises we ‘credibly’ spend much more in family rehabilitation schemes – then one wonders on the incredulity of what we are actually doing.  What I mean by this type of experiment is that we should be thinking through what links we should be looking for and using research methods we can take to approximate to control ideas.  This is an area of public functional illiteracy.

What we need is not complex mathematical schemes – these are usually the problem.  I can set fairly easy financial problems undergraduates mostly can’t do in class, yet they all appear to be able to do if I issue an out of class assignment – they copy and cheat in the main.  In academic quality assessments I’ve seen departments go from rubbish to excellent just by getting in a performance manager prepared to do the paperwork needed.  Something beyond this kind of bureaucratic lunacy is needed – and indeed the lunacy needs to go.  The QAA collapsed academic standards, though the industry could not see it needed to change.

We need some thinking from the ground up.  No manufacturer is interested in performance statistics that don’t relate to costs, sales and quality.  Trying to transfer techniques from even one related industry sector to another can be difficult and check-list approaches from a generic source usually fail.  We can be smarter than this and we don’t need maths, covariants or Gaussian copulas.  We need something we can’t cheat and we think is useful and fair.  Instead, we are beholden to loads of unnecessary dross that promotes glib argument and political manipulation.  To a scientist, it’s like being stuck having to skew results to suit a dud political theory like a geneticist working for Stalin.

Much as we want thieving druggies to ‘shape up’, we need to understand the effects of an economic system we can’t use to provide enough coppers and other resources like jobs they can do to change the environment around them and the one they grow in.  What difference a couple of trillion wasted on banksters we don’t know, precisely because we keep ourselves ‘free’ of realistic statistics.  Where are the comparisons between what it’s like to live and crime in Sweden, Norway and Britain?  Where is any straightforward statement on crime and immigration?  My burglary was paid for by insurance, but not the much worse financial and quality of life loss of having scum dumped next door.  I lost more in loan insurance ‘legally’ extracted by the bank than in the burglary.  In we had real statistics we’d know more about what crime is and what to do about it.

Every scheme of data recording I’ve seen has been a pain in the arse to use.  In police recording cases could be entered into databases that would print off charge sheets, self-duplicate for secure storage and be interrogation friendly.  The work in data entry should not be an additional burden.  Yet in our incompetence it always is.  Last time I was involved (5 years ago), detectives still had to lug cardboard boxes full of original statements to court, when these were digitised.  We should not try any substantial changes before addressing ‘quill and ink’ attitudes.

We Could Do Better On Crime Statistics

http://macdonaldlaurier.ca/files/pdf/MLI-Crime_Statistics_Review-Web.pdf

This is a link to a thoughtful criticism of Canadian crime statistics – worth a read for ideas on what might be wrong with ours and statistics generally.

Crime has been coming down across most of the EU, Britain, the USA and Canada for a decade.  Citizens generally don’t believe this to be the case.  Yet some of the crime that is down would be hard to dispute – homicide volume is the classic.  It’s hard to think cops around the world have become adept at hiding the bodies.

My own street has been almost completely peaceful since the removal of a dire couple of druggies who have caused trouble in spades wherever they have lived and continue to do so where they are now.  Prison makes no difference, except in the time they are off the street.  In the US people like them spend more time in jail.  Removing them from our society and their children might reduce our crime a great deal and prevent the ‘generational effect’.  Decent statistics (partly as argued in the Canadian article) would give us a clear idea.  The guess is that about 100,000 of these bastard homes exist.  If the significance of each is as strong as the one once next to us and getting rid of them as significant, then incarceration would make a massive difference, unless others would just emerge in the wake.  This could be tested too.

One reason given for the drop in US crime has been legalised abortion – the likely criminals getting an early death sentence.  Crude as this seems we should pursue a structuralist analysis of our offenders.

Across the world, the purpose of police statistics seems to be to tell the public crime is falling.  It’s pretty obvious from education to banking that false-accounting is rife, as is image management.

Police are presumably better organised and using better technology than ever before – this may be building a genuine deterrent effect or set of them.  My feeling is  crime is actually shifting, just as my former neighbours moved and into new categories.  Insurance rates against crime are not falling.  The statisticians don’t deal us a full deck.

Crime Figures

http://www.homeoffice.gov.uk/science-research/research-statistics/crime/crime-statistics-internet/

Our national crime figures often make little sense.  How, for instance, does domestic violence rise by 35% in the course of 12 months?  It doesn’t seem likely that spouses start weiighing into each other a third more often (though even stuff like the local soccer team losing can put up the number of incidents).

The local crime map for my area shows around 330 incidents in May this year – half being antisocial behaviour.  There are 32 officers including the PCSOs – so that’s about 10 incidents each per month.  The ‘map’ is no map at all as no details beyond a set of types of crime and ASB is given.

It would be interesting to know how many of the 330 incidents is actually attended by these local team officers.  Nearly all response I’m aware of is via Response – the poor sods flying around like blue-assed flies.

The average length of police work done would be useful.  If these 330 incidents take an average on an hour, then our local plods are only doing 10 hours each a month.  I know it’s not this simple, but this figure could be accurately defined.

Steve Bennett at thinblueline has written extensively on the gaming involved in police statistics.  The BCS has shown a steady drop in crime across the UK – this survey evades senior officer juking because it’s generated directly from the public.  It generally runs at twice the police recorded level and sensibly it would.  Not all crime is reported to police.

What always gets me about these figures is what I’d do with them if they crossed my desk and were from my area of concern.  I’d toss them in the bin and see if I could get the costs of producing them off my budget.  I expect stats to help me run things – and these just don’t.  They reek of expensive PR and bonus-rigging.

One assumes none of the senior cops who were covering up the phone hacking and dubious police earners now droning 24/7 recorded the crimes they were doing nothing about.  If we could open up stuff being written off as ASB we might well find many more crimes.  The actual number doesn’t matter much – what we need to know is how they are being dealt with and the costs of this execution.  This can be done statistically and it’s precisely not what the public gets.

Most cops don’t believe ‘crime’ is falling.  It has in my parochial world because a criminal family finally moved on after 7 years of the authorities failing to deal with them.  We have hardly seen a police car since.  They went to prison, but are now active in new premises.

With only 330 incidents a month, it would be useful to know who is responsible for those incidents – 10 lousy bastards like our former neighbours or 330?  How many perpetrators are identified?  What happens to them?  What does a fair sample of victims think about police action and final outcomes?  What do those dissatisfied identify as the problem?  What would a research team looking in detail at one month discover?

The whole point of practical statistics is not the numbers, but patterns, correlation and help in control.

If the 330 incidents in May in my neighhood is all our 32 officers are doing, then it seems a very expensive solution.  With on-costs 32 times £30K  then £960K or about £3K per incident is being spent just in the local budget, probably to ‘contain’ a dozen highly dud families.  I’m doing guesstimate – but with some real stats I wouldn’t have to.  What is the actual police cost for a recorded incident?  How much of this is the burden of really serious incidents?

In comparison with any harm from phone hacking, the way lives are wrecked around anti-social criminals is far more damaging.  One would expect police statistics to identify and quantify this.  The point of statistical analysis ion organisations is quality in the broad sense.  If we take a guesstimate on each of my local officers working ten incidents a month, there seems room for substantial focus and prevention of the same people perpetrating.  Yet, as Response (rather than local) turn up, one hears the perps are well known.

I could easily database 330 incidents a month in a manner that would allow reporting to identify repeat victims and offenders, crime concentrations and so on.  Such effort is wasted if it doesn’t lead to better, different actions.  This is what we should expect to hear on police statistics.  In part we do – and the message is that we aren’t stopping the same, sad, druggie prats.

One can understand ‘scrap’ metal thefts going up as commodity prices rise – crime tends to move with other trends and what cops are targeting.  That domestic violence can leap by 35% suggests some change in reporting and classification has occurred.  The ASB rate has been going up – but is this a new set of problems or just a new bin for crime?  The point is that we don’t know.  Where is the statistic on who any of these statistics have actually led to changes other than in the wallets of some senior cops?

Or for that matter on the cost-benefits of taking 100,000 crap families out of our system by putting them in curfewed trailer parks away from the environment they mess up and commit anti-social and other crimes in?  For 500 towns that would be an estate of 200 mobile homes each.  One pilot might be enough to send the message.  Say ten places were reserved for my neighbourhood and the number of incidents dropped from 330 to 200 a month.  This is the kind of statistic I’m interested in.

If we cut policing back to 1973 levels, how much money might be available to give back to ‘deserving welfare’?  I’d like to know this – though I don’t approve of the idea.  Would the amount, for instance, be more than that available from a cull of scroungers?  Or cutting legal bills in half?  One needs some numerate grasp of spending – preferably the scum who get nicked need costing in terms of all the money they put into other people’s pockets and thus for us to know how much they are really stealing from ours as tax-payers and who has interests in maintaining the problems.

Half our neighbourhood team is PCSO.  If they can really do the job, why are we paying for ‘expensive police officers’?  A cruel statement in many ways – but unlike many parts of the private sector, wage costs are not under such easy control.  One also has to wonder at higher-level costs – could we not get decent judges and advocates at much lower cost through specialist training for criminal matters (say undergrad plus one year law school or experience plus same).  Why no statistics on this kind of efficiency gain?  If a cop is paid twice the market rate, she’s costing us as much as some delinquent families if a PCSO can really deliver or could given the powers of arrest.  Once into this kind of economic thinking, a lot of nasty savings crop up.  Everyone wants to be a special case.  What would be the wage spreadsheet on police pay if ‘marked to market’?

The likely means to curbing police pay will be cuts and inflation through other wage rises or the dumping of electronic money as quantitative easing or in exchange markets and tricks like the PCSOs.  They are only the flick of a pen away from ‘full status’.  A recruitment and promotion freeze on other ranks will help swell their numbers.  The history on this isn’t pleasant – the last real police strike led to a big cut in pay (1919-21ish).  Has your force issued real stats on what the cuts mean and how they intend to cope?

These are issues that the public should know about in order to make informed decisions on policing and no great numeracy is now required once spreadsheets are established.  This could be achieved by scrapping the police reported system and putting these resources into a costing system on incident response, actual police activities and some sampling scrutiny on what turns up in the BCS.

Police might well want to start recording more crime now – something that happened in CID years ago when the number of crimes became more important than the fudged detection rate in terms of promoted ranks that could be claimed.  Expect HMG to try to dodge this with some new criteria.  Certainly the stuff put in front of us over many years would only get short-shrift by any executive trying to run the firm.  Does ACPO have another set it works from?