Stop and Search

[Roadhump]

Crime, especially violent crime, drug-associated crime and crimes against property, tends to take place in poorer, deprived areas. Greenock and port Glasgow certainly qualify as deprived areas and are known for a chronic drugs problem. Most crime, in fact the vast majority of crime, is committed by 14-25 year old males (want to cut the crime stats? Lock up every male between their 14th and 25th birthday) so it makes absolute sense for the police to target young men in deprived areas because there is a statistically much higher chance of them being involved in crime.

When this is transposed to England, it's also a fact that the most deprived areas have higher concentrations of black people, to the extent that we have special police units dedicated to tackling violent crime involving black people but this has nothing to do with race and everything to do with social deprivation.

When you treat violent crime as a disease (as Glasgow very successfully did), using the symptoms to point towards root causes then you have a chance of fighting it. Stop and search addresses the symptoms not the causes.

All the pish peddled about racist police detracts from the real issues. In Glasgow, the community itself was well aware that the problem involved them so the solution had to involve them too. All I ever see from London is attempts to blame an institutionally racist police force for all the ills that befall the black communities.

I’m with you on this. It is far more likely (IMHO) a consequence of the greater disadvantage black people suffer in society generally, resulting in a greater proportion of the black population, through no fault of their own, living in areas where street crime is high, and where overt / proactive police attention is therefore focused, than it is of the police consciously targeting people because of their skin colour.

I could be wrong on that, but it would be interesting to see some analysis of the proportion of stop / searches by location, combined with some analysis of the proportion of people of various racial origins living in areas having different levels of affluence, social deprivation and crime. I know that the proportions of different racial origins living in an area is (or was) provided at police force level, and I suspect that there is information of the kind I suggest is looked at, at a more local level, e.g. council wards, available somewhere, but whether it is public or not, I don’t know. If it was, it might get beneath the headline figures, and provide some meaningful explanations, rather than the usual police knockers v police defenders debate.

 

[swansonj]

The Met Police used to point out that black people were disproportionately involved in stabbings. American police point out that Muslim men are disproportionately involved in terrorism. They didn't understand Bayesian statistics. The (false) argument is "people of this kind are more likely than people of that kind to do bad thing X. Therefore I shall target people of this kind." Not only is this sort of targetting overtly discriminatory, bad policing and misjudged, it's also bad science.

Could you expand on the "bad science" bit please? As you would expect, I agree with you it's bad policing, bad politics, and bad social justice, but your reference to bayesian statistics intrigues me. Are you just talking about the dangers of confirmation bias (which again i would agree with) or are you saying something more?

 

[Profpointy]

Could you expand on the "bad science" bit please? As you would expect, I agree with you it's bad policing, bad politics, and bad social justice, but your reference to bayesian statistics intrigues me. Are you just talking about the dangers of confirmation bias (which again i would agree with) or are you saying something more?

Beat me to it, was about to ask the same thing - although it's a while since I formally studied Bayes' Theorum

 

[srw]

Could you expand on the "bad science" bit please? As you would expect, I agree with you it's bad policing, bad politics, and bad social justice, but your reference to bayesian statistics intrigues me. Are you just talking about the dangers of confirmation bias (which again i would agree with) or are you saying something more?

I hope that you and @Profpointy will forgive me if I don't provide you with a full worked-through proof, or a reference. I may well have read it in a Goldacre book; if not it could be in any one of a couple of dozen books that apply statistics to understanding the world.

By focussing on [group X] for your stop and search you are essentially asserting that P([bad thing] | [group X]) is enough higher than P ([bad thing]) that you will make a significant difference. In fact what you know is that P([group X] | [bad thing]) is higher than P([bad thing]), which isn't the same thing. Once you take into account the negative impact of focussing on [group X] you end up doing more harm than good.

The argument (which I've probably misrepresented) is very similar to the argument that X-ray screening might not be an effective tool for breast cancer detection, or that [expensive drug Y] isn't actually a good choice of treatment for [disease Z], or indeed that helmets aren't an effective intervention for increasing cycle safety.

Feel free to make my argument more robust....

 

[Wobblers]

[QUOTE="srw, post: 4517443, member: 1731"
By focussing on [group X] for your stop and search you are essentially asserting that P([bad thing] | [group X]) is enough higher than P ([bad thing]) that you will make a significant difference. In fact what you know is that P([group X] | [bad thing]) is higher than P([bad thing]), which isn't the same thing. Once you take into account the negative impact of focussing on [group X] you end up doing more harm than good.
[/QUOTE]

You're right in stating that your argument "could be more robust". The highlighted statement is an assumption - to state it as fact is quite, quite wrong. Even if it is reasonable to believe it.

Another issue, which you've completely neglected, but crucialy forms the core of the issue is exactly how "negative impact", bad outcomes and good outcomes are defined and measured. For instance, if over 75% of all knife crime is committed by a small selection of people who happen to wear baseball caps and shellsuits (as was the case in Glasgow 10 years ago) how do you go about quantifying the bad outcomes of possible alienating one section of society with the possible beneficial deterrence effect of it being common knowledge that carrying a knife is likely to result in arrest? And can you be sure that these are the only consequences. This is an issue for society as a whole, and certainly should not be dealt with by any one section, be it police, victims or politicians.

Finally, statistics deals rather well with probabilities. That's not the issue. The issue is are those parameters you're measuring need to be reliable, meaningful and appropriately reflect reality. Your statistical analysis is only ever going to be as good as the model you use. (Hence the endlessly tenditious Helmets thread... ) In a situation like this, where the outcomes are poorly defined and measured, any form of statistical analysis will be of limited use. It's essential to understand the limitations of the analytical tools you use.

 

[srw]

You're right in stating that your argument "could be more robust". The highlighted statement is an assumption - to state it as fact is quite, quite wrong. Even if it is reasonable to believe it.

Another issue, which you've completely neglected, but crucialy forms the core of the issue is exactly how "negative impact", bad outcomes and good outcomes are defined and measured. For instance, if over 75% of all knife crime is committed by a small selection of people who happen to wear baseball caps and shellsuits (as was the case in Glasgow 10 years ago) how do you go about quantifying the bad outcomes of possible alienating one section of society with the possible beneficial deterrence effect of it being common knowledge that carrying a knife is likely to result in arrest? And can you be sure that these are the only consequences. This is an issue for society as a whole, and certainly should not be dealt with by any one section, be it police, victims or politicians.

Finally, statistics deals rather well with probabilities. That's not the issue. The issue is are those parameters you're measuring need to be reliable, meaningful and appropriately reflect reality. Your statistical analysis is only ever going to be as good as the model you use. (Hence the endlessly tenditious Helmets thread... ) In a situation like this, where the outcomes are poorly defined and measured, any form of statistical analysis will be of limited use. It's essential to understand the limitations of the analytical tools you use.

Ouch. That'll teach me to try and be too clever first thing in the morning. If I ever do find the piece I read I'll try and summarise better.

There's a much simpler argument online, which I suspect is closer to what I half-remembered: http://www.slate.com/blogs/moneybox/2013/07/16/richard_cohen_bayesian_inference.html
Even if 75% of knife crime is committed by a bunch of neds, it doesn't logically follow that stopping and searching neds is a sensible or proportionate response - even though that's the argument that the police use. Because you'll stop and search a load of innocent neds as well as a tiny number of neds carrying knives.

 

[threebikesmcginty]

You're right in stating that your argument "could be more robust". The highlighted statement is an assumption - to state it as fact is quite, quite wrong. Even if it is reasonable to believe it.

Another issue, which you've completely neglected, but crucialy forms the core of the issue is exactly how "negative impact", bad outcomes and good outcomes are defined and measured. For instance, if over 75% of all knife crime is committed by a small selection of people who happen to wear baseball caps and shellsuits (as was the case in Glasgow 10 years ago) how do you go about quantifying the bad outcomes of possible alienating one section of society with the possible beneficial deterrence effect of it being common knowledge that carrying a knife is likely to result in arrest? And can you be sure that these are the only consequences. This is an issue for society as a whole, and certainly should not be dealt with by any one section, be it police, victims or politicians.

Finally, statistics deals rather well with probabilities. That's not the issue. The issue is are those parameters you're measuring need to be reliable, meaningful and appropriately reflect reality. Your statistical analysis is only ever going to be as good as the model you use. (Hence the endlessly tenditious Helmets thread... ) In a situation like this, where the outcomes are poorly defined and measured, any form of statistical analysis will be of limited use. It's essential to understand the limitations of the analytical tools you use.

 

[Profpointy]

Neither fair nor helpful.

A reasonable question / challenge has been posed and the challenged has given a fair and reasonable answer. A follow-up, also fair if slighlty picky point has been made. This is sensible enough exchange to understand each others' points - unlike the shyte on a couple of the promising enough threads we've had

 

[Brandane]

Even if 75% of knife crime is committed by a bunch of neds, it doesn't logically follow that stopping and searching neds is a sensible or proportionate response - even though that's the argument that the police use. Because you'll stop and search a load of innocent neds as well as a tiny number of neds carrying knives.

But how many of those "innocent neds" would be carrying a knife if it were not for the deterrent effect of knowing that there was a good chance that the Police are going to search them?
Targeted stop and search (as long as it can be justified in law, i.e. that the Officer carrying out the search has "reasonable cause to suspect" that the person is in possession of a knife) is a perfectly sensible policy. Wasting valuable Police resources by targeting the coffee morning at the local church meeting of "Age Concern" might not be such a good idea.

 

[Wobblers]

Sorry Gramps, I didn't mean to disturb your nap. How about I buy you a packet of Werther's Originals as a peace offering? Should go well with the cardigan and slippers

 

[Joey Shabadoo]

But how many of those "innocent neds" would be carrying a knife if it were not for the deterrent effect of knowing that there was a good chance that the Police are going to search them?
Targeted stop and search (as long as it can be justified in law, i.e. that the Officer carrying out the search has "reasonable cause to suspect" that the person is in possession of a knife) is a perfectly sensible policy. Wasting valuable Police resources by targeting the coffee morning at the local church meeting of "Age Concern" might not be such a good idea.

Added to that is the sad fact that many youths of the same cohort feel the need to carry a knife for self-protection. If the police are stopping and searching all youths in the cohort, it acts as a deterrent for those who can be deterred.

 

[Profpointy]

Added to that is the sad fact that many youths of the same cohort feel the need to carry a knife for self-protection. If the police are stopping and searching all youths in the cohort, it acts as a deterrent for those who can be deterred.

"searching all" - seriously?

 

[threebikesmcginty]

Neither fair nor helpful.

A reasonable question / challenge has been posed and the challenged has given a fair and reasonable answer. A follow-up, also fair if slighlty picky point has been made. This is sensible enough exchange to understand each others' points - unlike the shyte on a couple of the promising enough threads we've had

Tough, if you think that I have contravened the forums's rules, then by all means report by post to the moderators.

 

[Profpointy]

Tough, if you think that I have contravened the forums's rules, then by all means report by post to the moderators.

ooh, get you

 

[EnPassant]

Neither fair nor helpful.

A reasonable question / challenge has been posed and the challenged has given a fair and reasonable answer. A follow-up, also fair if slighlty picky point has been made. This is sensible enough exchange to understand each others' points - unlike the shyte on a couple of the promising enough threads we've had

Agreed that it's neither fair nor helpful, however it is funny.

I am doing my best to follow the discussion as a non student of such things and thus it is very interesting to me from my position of ignorance but not stupidity*.

Further, it is indeed refreshing to witness badinage that doesn't immediately descend to childish point scoring about who is 'right', or simply plain insults.

I can comfortably throw a like on something that made me giggle, without 'taking sides' or denigrating any of the previous enlightening argument. However whilst I like that too, it is harder to put a 'like' on it without either A) appearing to support one stance over another or a clarification or B) adding nothing whatsoever by liking every post. Of course it could be argued that liking the funny picture but not liking any of the original points could of itself be misleading, I'll just have to take that risk....

ETA: *Incidentally, I'm sticking to this "I'm not stupid" line, in spite of all evidence to the contrary.