Matthew draws our attention to a recent statement allegedly by a senior police officer advocating either that the police be armed routinely, or that capital punishment be reintroduced for the murder of police officers. This, of course, was provoked by the recent tragic murder of Pc Sharon Beshenivsky. Now, almost nobody is in favour of murdering police officers (the obvious exception presumably being the murderers themselves), but nor is it obvious that capital punishment is a sensible response. Matthew plots a graph of the number of police officers murdered over time, which doesn't show any very significant upward trend since capital punishment was abolished in 1964.
Here's a slightly more detailed version of the same chart, with the total number of homicides (civilian and police) also plotted: (the data here are for England and Wales only; numbers of police are drawn from this House of Commons Library paper while the number of police killed is from the Police Memorial site, excluding Scottish police and a number of English police killed by terrorists in Cyprus in 1956; and the total homicide statistics are from the Home Office):
Matthew states that the number of police killed doesn't seem to have increased sharply since capital punishment was abolished. The data fit a Poisson distribution pretty well (as you'd expect) and similarly have no serial correlation (the number of murders in year N doesn't depend on the number in year N - 1); therefore we can estimate the rate at which police are killed in any given year by generalised regression, giving something like this over different periods of the last century:
-- naively, there is an increase in rate, but the confidence intervals are pretty broad and overlap heavily (because there isn't that much data), so the estimate of the increase is pretty inaccurate. Varying assumptions make quite a big difference; the change, obviously, becomes a bit smaller (+33% rather than +48%) if we exclude the victims of domestic terrorism, or if we include the slightly more violent inter-war period in the `no hanging' data (+16%, or +5% if terrorism excluded).
(NB that none of this demonstrates an effect of the end of capital punishment, just that something coincided with it.)
But the most striking thing about this is the comparison between the risk of death for on-duty police and the overall homicide rate, which (apart from a spike during the Second World War) remained roughly constant from about 1930 until the early 1960s, and then rose sharply, approximately doubling from about 6/million/year to about 14/million/year (see, e.g., the House of Commons library paper above). In fact, the overall homicide rate is now roughly the same as the risk of death for police officers on duty.
Anyone know how much of this is down to better classification of deaths into homicide and non-homicide? Whether or not a police officer is killed on duty is a fairly unambiguous question: either they are or they aren't. That should mean that the rate of murder of police officers shouldn't be affected much by (e.g.) improvements in medical science or inquest practise. Less so for the general population, where some methods of killing (poisoning etc.) may be put down as accidental death or natural causes. So perhaps improved detection could be responsible for the difference in rates? However, I can't find any positive evidence for this (in particular, long-term data for number of homicides by method) -- any suggestions?
Comments
Posted by Liz Upton, Saturday, 26 November 2005 21:41 (link):
Of course, even though the chap spouting this rather worrying opinion is now a member of the House of Lords, after the police force's recent lobbying of parliament over the 90 days proposal, this just gives me horrible separation of powers twitches. It was Lord Stevens, who was head of the Met until January, who told the News of the World that he thought this case should result in the reinstatement of the death penalty, and he's in...interesting company.
Posted by Peter Clay, Sunday, 27 November 2005 13:39 (link):
I have to say that if I were looking to reintroduce the death penalty I'd deploy that quite striking graph as part of the argument. It's intereting that the general murder rate has risen to that of police officers. I suppose to investigate it further it would be useful to have a breakdown of the general murder rate by relations of the victim to the murderer (was it a big rise in drug-related killings?)
Posted by Alex, Sunday, 27 November 2005 14:43 (link):
That's precisely what I thought - it seems to be just after the 1971 Misuse of Drugs Act and the boom in drugs criminalisation that the rate breaks out of the range it was "trading" in postwar.
Posted by Chris Lightfoot, Sunday, 27 November 2005 18:22 (link):
Well, the other thing that's usually quoted (see Levitt etc.) is the number of young men in the population, since that's the segment of the population which does most of the murdering. That gives a plot like this:
(population data from NDAD). That looks reasonable, but note that the fit for the interwar period is pretty poor, and similarly for the last few years. Alas, the NDAD population estimate is a bit different to the one from the 2001 census, so I can't plot comparable recent population data to see whether that's anomalous or not.
Posted by Harry Hutton, Sunday, 27 November 2005 17:18 (link):
According to the British Medical Journal the current murder rate might not be comparable to the 1960s murder rate due to improvements in trauma care.
Posted by Harry Hutton, Sunday, 27 November 2005 17:22 (link):
Another point is that it is harder to murder the police now that they no longer do foot patrols.
Posted by Dan Hardie, Sunday, 27 November 2005 18:29 (link):
'the rate of murder of police officers shouldn't be affected much by (e.g.) improvements in medical science': You're wrong. As trauma care improves (and radio communications, faster ambulances and the use of helicopters make casualty evacuation much faster) one would expect that, of the total number of police officers injured in violent incidents, a lower number would die.
Certainly this is what has happened to US or UK military casualties over the last century. There were particularly big downward falls in fatalities as a proportion of casualties following the introduction of antibiotics in the last years of the Second World War, and the use of helicopters for casevac in Korea and other '50s conflicts; then a slower downward trend between the '50s and the '90s, probably reflecting better trauma surgery and faster helicopters or ground vehicles; then a really sharp fall for US forces in Iraq, probably due to the massively improved training of US Army medics, and the introduction of QuikClot hemostatic sprays and antibiotic beads. Before QuikClot, for example, a severed femoral artery was likely to be fatal even if you did get the casualty into a trauma room within 'the Golden Hour'. (The British Army hasn't been issued with this stuff, due to Mr Tony's passionate concern for the troops he sends into battle, but a lot of medics buy it themselves.)
Body armour should also have a pretty significant downward pull on the police murder rate: most cops in London now seem to have stab-proof vests on.
Obviously civilians don't have body armour, but casevac and improved trauma care ought to be decreasing the proportion of violently-assaulted civilians who end up dead. Maybe there's a slightly less acute downward pull on the non-police murder rate because all police officers are issued with radios linked to professional dispatchers, so they presumably have a better chance of summoning help than civilians, who may not have a phone when injured. You would also have thought that mobile phones have cut the interval between civilians being injured and their arrival in casualty departments. If that is happening, and the non-police murder rate is rising regardless, that is very disturbing.
Posted by Chris Lightfoot, Monday, 28 November 2005 01:31 (link):
This is completely speculative, but there may be a little bit of evidence for it; for instance, here are the 2003/4 homicide statistics, by means of killing (from this Home Office report), and for comparison those from 1993:
So, looking at 1993, the sum of `burning', `poison or drugs', `other' and `not known' (as methods where there's obvious scope for ambiguity/error) is 42, 7%, or about a fifth of the difference between the then homicide rate and that in ~1960; in 2003/4 it's harder to say, since there's a large number for which the method is not known but will presumably become so later, but excluding those we get 98 by the above methods, or 13% of the 728 killings for which the method is known; again, comparing the total number of homicides to ~300 in ~1960, that accounts for about a fifth.
I haven't found any timeseries data on killings by method for a longer period, though, so I don't know whether this is actually an important effect.
Out of interest, how well do these work? Last time I read anything about this it seemed that everybody recognised it as a Hard Problem, and that was about it.
Posted by Dan Hardie, Monday, 28 November 2005 10:15 (link):
No idea how well stab-proof vests work. I have only ever worn military body armour, which is basically a huge thick vest with dirty great ceramic plates front and back, meant to keep bullets and shrapnel out.
In his collection of essays 'Against the Grain' (really the silliest title for a book by a paid up member of the quangoing class) Simon Jenkins quotes a conversation with an un-named Chief Constable who says that rises in the murder rate are due to the fact that the police now go out and search for 'prostitutes and tramps' who are reported missing. I don't like the attitude behind this statement, but maybe there's something to it: it could be that a higher proportion of people originally reported as Missing Persons now end up on the murder statistics, if a) the police are more likely to undertake a search when someone is reported missing or b) they have the manpower, procedures or technology (eg helicopters, infra-red cameras, scuba equipment) to search more effectively than hitherto.
The only test I can think of for this hypothesis would be the frankly rather balls-aching one of going through reports on murder victims and calculating what percentage of them per year were discovered within 24 hours of the murder, 24-48 hours, and so on at greater time intervals.
Posted by Eben, Monday, 28 November 2005 22:35 (link):
Interesting stuff on this manufacturer's website, which caught my eye for the very long list of 'unsupported' stabbing implements, and the rather cheerful warning that 'to state the obvious, getting stabbed ALWAYS carries some risk'.
I recall reading somewhere that a major problem with the vests issued to the British police is that the ceramic plates which make a big contribution to their stab resistance render them, as you say, pretty much unwearable, especially if you happen to be of the female persuasion. Most therefore leave them out, with predictable consequences.
Posted by Dan Hardie, Monday, 28 November 2005 10:34 (link):
Also, your figures are nice, but I don't know what the hell to make of at least a couple of categories: 'not known' and 'other'. Why the massive rises- 242.9% rise in 'other' and 2000% (two thousand per cent!) in 'not known'.
Possibly, re 'other', there are problems with classifying attacks where more than one type of instrument is used :eg Assailant A kicks and punches the victim, Assailant B uses a crowbar and Assailant C stabs him: the victim subsequently dies of massive trauma. One would imagine attacks using multiple weapon types are quite a large percentage of total violent crime. Going by the fact that all the murder numbers you quote are whole numbers, the Home Office doesn't have a system in place to enter one third of this murder as being blunt instrument, one third stab, etc, but counts them as 'other'.
Re 'not known', possibly that's due to the fact that coroners' courts had yet to pronounce on the cause of death: quite likely, if we're talking about people murdered in 2004 and statistics only recently released. Maybe 'not known' would be much lower for the 2002/3 murder stats. If, on the other hand, the police and forensics people really don't know what caused that number of violent deaths, we're in trouble...
Posted by dsquared, Monday, 28 November 2005 21:14 (link):
I would guess (and it is no more than that) that the largest single category of "other" would be defenestrations and similar shove-off-high-place-or-into-the-path-of-something-dangerous murders since they are not separately listed.
Posted by dsquared, Monday, 28 November 2005 21:17 (link):
also quite possibly industrial accidents recorded as (negligent) homicides?
Posted by Chris Lightfoot, Monday, 28 November 2005 23:47 (link):
I believe you're right about multiple causes being classified as `other'; there are fewer and fewer `not known' cases in earlier years, so most of that category must indeed be cases where they don't yet know but will later find out. The remainder are presumably (e.g.) bodies so badly decomposed as to reveal no forensic evidence. Unfortunately I can't find anything to confirm that yet, though the Home Office report (link above) does contain a few footnotes on the subject, of which the oddest is this (on the `shooting' heading):
-- I read that as meaning that a person who is battered to death with a handgun counts as having been `shot' for purposes of the statistics, which seems completely mad.Posted by Dan Hardie, Tuesday, 29 November 2005 11:13 (link):
I don't think it is all that mad. Plenty of us are likely to vigorously resist any assault with fists or blunt instruments, or at least to run away. On the other hand, if threatened with a handgun, even a hard case is likely to stay still and do nothing, which gives the assailant plenty of opportunity to get in the first couple of stunning blows to the head, followed by a little leisurely battering to death. Handguns, then, along with all their other delights, may make it easier to club people dead, precisely because the victim is less likely to offer initial resistance. It might therefore make sense to class this as a 'gun killing' rather than a 'blunt instrument killing'. Besides, are there enough of these cases for them to be statistically significant?
Posted by Chris Lightfoot, Tuesday, 29 November 2005 11:58 (link):
From the next chapter of the Home Office 2003/4 report: used to threaten in 64% of cases; fired in 32%; used as a blunt instrument in 4% (the latter two are read off the graph). There were about 10,000 cases in which firearms (rather than airguns) were used, of which about 2,400 resulted in any injury (from `slight' to `fatal'). If we assume that a single incident is only counted once (e.g. if somebody threatens a person with a gun, hits them with it, then shoots them, that would count only as `fired' rather than towards all three totals) something like 250 of those would be from guns used as blunt instruments. That's tiny compared to the overall level of violent crime (~2 million incidents / year) or recorded level (~1 million / year), but still a significant fraction of gun crime.
Posted by Roy Badami, Thursday, 8 December 2005 23:48 (link):
FYI, you need to be careful with terms like detection rate. In the context of crime, AIUI, the word detection is used in a very specific technical sense. It's what a detective does, ie loosely speaking figuring out 'whodunnit'. So the detection rate is actually the proportion of (reported) crimes which are solved (aka the clear up rate), which doesn't appear to be what you're talking about here...
Incidentally, I have always presumed that this technical sense is what is intended when legislation uses phrases such as the prevention and detection of crime.
-roy
Posted by Chris Williams, Monday, 28 November 2005 21:46 (link):
Quite by chance, I happen to study this sort of thing for a living. I know far too much about this to have time to comment on all of it right now. First, though, I'd like you all to go away and read two articles:
Taylor, H., 'Rationing crime: the political economy of criminal statistics since the 1850s', Economic History Review, 1998a, 51, 3, p.569-590.
summary: 'The murder rate was artificially held down until about 1930.'
Morris, R.M., '"Lies, Damned Lies, and Criminal Statistics": Reinterpreting the criminal statistics in England and Wales' in Crime, Histoire & Sociétiés / Crime, History & Societies, Vol. 5 no. 1 (2001), 111-127.
summary: 'It might have been, although not in the way that Howard thinks it was.'
Once you've done that, you also ought to factor in the ending of the year-and-a-day rule, and also check whether or not the Police Memorial people are using the same criteria of 'homicide victim' as the Home Office were.
More later, especially if the ESRC funds our proposal to look more closely at the way that the HO treated crime stats in the mid C20th - a proposal which, by an amazing co-incidence, also namechecks Jon Agar's _The Government Machine_.
Posted by Chris Lightfoot, Tuesday, 29 November 2005 01:38 (link):
OK, haven't got the Morris paper yet, but Taylor's argument essentially is that the money available for prosecutions was limited and this had the effect of limiting the total number of prosecutions. Murders for which resources were not available for prosecutions were reduced to lesser offences: (p586)
or were ignored entirely: (p587) and:So (modulo the second paper you recommend) this all seems pretty plausible (from my position of complete ignorance of the subject), but it'd be nice to have some specific examples of incidents of the type in the first quote where the police reduced the charges in a particular murder case to `wounding, assault etc.', rather than just the assertion offered that such incidents must have taken place (Taylor cites Bodkin's evidence to the Royal Commission on Police Powers and Procedures for the statement about rates vs. Vote funding for prosecutions, but no reference for the `reduce charges' statement).
(As an aside, I think Horatio Nelson Hardy must come a close runner-up to Sylvanus Percival Vivian in the Bonkers Victorian Name Award.)
Posted by Chris Lightfoot, Tuesday, 29 November 2005 01:58 (link):
But here's the abstract for the Morris paper, from here:
Posted by Chris Williams, Tuesday, 29 November 2005 10:56 (link):
Hey, you gave away the ending.
Note though, that Bob _doesn't_ distance himself from Howard's assertion that murder figures could be massaged down substantially at local level, for example by coroners and local police forces simply failing to investigate many suspicious deaths, in the interests of a quiet life.
Posted by Dan Hardie, Tuesday, 29 November 2005 11:22 (link):
'Howard's assertion that murder figures could be massaged down substantially at local level, for example by coroners and local police forces simply failing to investigate many suspicious deaths, in the interests of a quiet life.'
Yes, but these coverups were always potentially subject to challenge from local newspapers, councillors and MPs. Furthermore, the likelihood of such challenges will have increased following a) the greater spread of adult literacy after the various Education Acts, esp. that of 1870 (more influential local newspapers, greater awareness of one's rights) and b) the spread of the franchise after the Representation of the People Acts, up to 1918 but especially 1867 and 1884 (more of an incentive for elected representatives to listen to even poor constituents).
I shall read Howard's paper, but if he doesn't account for these possibilities his argument seems pretty seriously weakened to me.
Posted by Chris Williams, Tuesday, 29 November 2005 15:58 (link):
If I were you, Dan, I wouldn't set vague expectations against definite evidence. But each to their own. Check it out and see what you think.
Posted by Dan Hardie, Tuesday, 29 November 2005 16:49 (link):
If I were you, Chris, I wouldn't pretend that there was anything 'vague' or indeed 'expected' about saying that a massive shift in popular literacy and the gradual enfranchisement of the entire male population exerted a massive influence on how agents of the state interacted with the populace. Given the subject under discussion, you might, for example, like to inform yourself about the date of the establishment of the Court of Appeal (Criminal Division), and find out how much that had to do with the popular press.
...But each to their own.
Posted by Eben, Tuesday, 29 November 2005 18:18 (link):
If I were you, Dan, I wouldn't set vague expectations against definite evidence. But each to their own.
/me starts inserting ceramic plates :)
Posted by Chris Lightfoot, Monday, 19 December 2005 22:08 (link):
OK, so after a pathetic exhibition of time management I've finally managed to get to the library and find a copy of Morris's paper.
Essentially his argument re. the homicide rates is that: the budgetary control which Taylor perceives did not in fact take place; the decentralised structure of the criminal justice in England and Wales did not lend itself to a conscious and wide-ranging effort to ration murder prosecutions; and that there is no evidence for centrally-imposed controls on the numbers of prosecutions in the variation of the homicide rate around an average of ~150/year in the years before 1966, and that had such control been imposed, then you'd expect the annual variations to be smaller than the historical `tight band [in Taylor's words] of 20%' around the mean.
On that last point... the total recorded homicide rate -- I don't have stats to hand for the murder rate alone -- between 1898 and 1966 is not Poisson-distributed, whereas naively you'd expect that the total number of homicides would, as rare random events, be so distributed. The total homicide rate averaged ~300/year, ±35/year, whereas if it were Poisson you'd expect it to be less variable: 300±17/year. Similarly, this statement from Taylor quoted by Morris,
is also not consistent with the murder rate being Poisson -- in that case you'd expect only one or two years in which the rate strayed outside those bounds. A reasonable control mechanism would be likely to reduce the variance from its putative Poisson level. This leaves open the question of why the variance was so high at all.Year-and-a-day rule: that was abolished in 1996, which is too late to make much difference to the conclusions of the above (such as they are).
Differences in the definitions of homicide between the police and general population series: hmmm... not sure about this one. But the relevant thing would be changes in the definition within each series, wouldn't it?
Posted by dsquared, Tuesday, 20 December 2005 10:31 (link):
Are the outliers mainly positive, mainly negative or evenly distributed? If they're mainly negative and occur in war years I'd chuck them out.
Posted by Chris Lightfoot, Tuesday, 20 December 2005 11:17 (link):
Dunno about the murder rate, but for total homicides, it's not just the outliers -- excluding both wars you get 308±31/year, and a distribution like this:
Posted by Chris Williams, Tuesday, 20 December 2005 16:18 (link):
Cheers for looking that lot up, Chris.
As for the non-poisson-ness of the distribution, could this be explained by the Shipman effect? If someone gets convicted in year X for killing 4 people, those murders will act as a blip which pushes up the total, and artificially lowers it for the years in which the murders were actually carried out. I note that there are a number of 'high' years on the graph, which would be consistent with this, although I'd guess that most of them are accounted for by the end of the period, when the rate was increasing.
Posted by Chris Lightfoot, Tuesday, 20 December 2005 19:08 (link):
I don't think that effect would do it (and some trivial numerical experiments confirm that) but I'd have to work it out properly to confirm that that was the case.
Posted by dsquared, Tuesday, 20 December 2005 21:48 (link):
gosh, quite bizarre. I'm out of ideas. Serial correlation in British murder rates? It actually looks like a Gaussian.
Posted by Chris Lightfoot, Wednesday, 21 December 2005 18:22 (link):
I'd assumed there'd be no serial correlation. In fact, there is. Here's the murder rate, 1901-1965 (sadly I only have annual population statistics -- I've interpolated linearly, which is Bad and Wrong):
It's not very significant, but it's there. It'd be astonishing if this were present in the series for actual homicides; it must be an effect of the prosecution process and how long it takes, I guess.
Posted by dsquared, Thursday, 22 December 2005 10:43 (link):
Might not be so astounding; all sorts of things are correlated to the business cycle (property crime certainly is) and given that the correlogram looks to me like an AR(4) process (Box-Jenkins analysis, how I love it so) which would usually be a reasonable first stab at a model of employment or GDP, it could be that some form of procyclicality is driving the departure from the Poisson model. But I remain completely baffled.
Posted by Chris Williams, Thursday, 22 December 2005 19:36 (link):
Reported violent crime certainly correlated with the business cycle in postwar Britain. See: 'Trends in crime and their interpretation: a study of recorded crime in post war England and Wales', S Field - 1990 - London: HMSO
Posted by Roy Badami, Friday, 2 December 2005 21:52 (link):
I wonder what the EU's response would be to a member state re-introducing the death penalty. Isn't abolishion considered part of the acquis communautaire?
-roy
Posted by Chris Lightfoot, Tuesday, 20 December 2005 11:35 (link):
Well, with any luck, we'll soon get to see how they treat secret CIA prisons in Eastern Europe, which should be some sort of a precedent. (As I understand it the threatened sanction in that case is to remove voting rights in the Council of Ministers; I'm not sure how much more they can do. Ban the cross-border trade in rope?)
Posted by Roy Badami, Friday, 23 December 2005 02:40 (link):
Actually, that sounds pretty serious in any case, but exactly how serious depends whether its restricted to competencies that require qualified majority voting, or applies to all competencies, including those that require unanimity... ie does the member state lose its veto?
Post a new comment.
Comments copyright (c) contributors and available under a Creative Commons License. See also the comments policy.