Deltoid

Jim Zoes was kind enough to send me the data on English homicide rates that he obtained from the Home Office. I’ve typed it in and included it at the end of this message.

The numbers are certainly higher than those recorded in the WHO Statistical Yearbook. I’ll try to find out why, but for now I think we should consider the Home Office data to be more reliable.

Anyway, there is plenty of data to let us look at the question of whether English homicide rates were lower before the introduction of gun control than after. Fortunately, the answer turns out to be yes and no, so debates here about it can proceed endlessly.

Here the average homicide rates by decade:

1860 means decade starting 1860 i.e. 1860-1869
1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990
 1.7  1.6  1.5  1.1  0.9  0.8  0.7  0.8  0.8  0.7  0.7  1.0  1.2  1.4

The homicide rate in England declined after gun control was introduced in 1920, though this could have been the continuation of a long term trend. The rate increased again after 1970 and is now almost as high as is was in the 1880s.

Compare 1910-19 with 1920-1929 and the rate went down.
Compare 1910-19 with 1990-1993 and the rate went up.
Compare 1857-1919 with 1920-1993 and the rate went down.

Homicide rate per 100k, England/Wales, Home office data
(Technically this is the rate for "offences initially recorded as
homicide" ) 
1857 1.26
1858 1.58
1859 1.42
1860 1.44
1861 1.36
1862 1.62
1863 1.76
1864 1.67
1865 1.96
1866 1.82
1867 1.85
1868 1.71
1869 1.74
1870 1.43
1871 1.78
1872 1.7
1873 1.58
1874 1.69
1875 1.55
1876 1.63
1877 1.53
1878 1.65
1879 1.51
1880 1.52
1881 1.61
1882 1.57
1883 1.57
1884 1.57
1885 1.49
1886 1.56
1887 1.37
1888 1.43
1889 1.07
1890 1.17
1891 1.02
1892 1.08
1893 1.13
1894 1.02
1895 1.13
1896 1.08
1897 0.98
1898 1.04
1899 1.01
1900 0.97
1901 1.05
1902 1.0
1903 0.94
1904 0.94
1905 0.84
1906 0.77
1907 0.77
1908 0.92
1909 0.85
1910 0.81
1911 0.81
1912 0.86
1913 0.91
1914 0.73
1915 0.75
1916 0.69
1917 0.61
1918 0.54
1919 0.79
1920 0.83
1921 0.66
1922 0.64
1923 0.68
1924 0.71
1925 0.82
1926 0.76
1927 0.75
1928 0.72
1929 0.79
1930 0.75
1931 0.72
1932 0.72
1933 0.88
1934 0.86
1935 0.77
1936 0.88
1937 0.74
1938 0.74
1939 0.85
1940 0.7
1941 0.75
1942 0.96
1943 0.76
1944 0.84
1945 1.15
1946 0.81
1947 0.86
1948 0.78
1949 0.68
1950 0.79
1951 0.75
1952 0.91
1953 0.74
1954 0.7
1955 0.63
1956 0.7
1957 0.71
1958 0.58
1959 0.59
1960 0.62
1961 0.57
1962 0.64
1963 0.65
1964 0.63
1965 0.68
1966 0.76
1967 0.86
1968 0.86
1969 0.81
1970 0.81
1971 0.94
1972 0.98
1973 0.95
1974 1.22
1975 1.03
1976 1.15
1977 0.98
1978 1.09
1979 1.28
1980 1.26
1981 1.12
1982 1.25
1983 1.11
1984 1.24
1985 1.25
1986 1.33
1987 1.37
1988 1.29
1989 1.25
1990 1.32
1991 1.42
1992 1.37
1993 1.31

Mark Gibson writes:

And Waller and Okihiro (1978, p. 31) reported that 44% of burglarized Toronto residences were occupied during the burglaries, with 21% of the burglaries resulting in confrontations between victim and offender.

Waller and Okihiro did not have enough money to conduct a full victim survey of Toronto, so concentrated on some high crime areas. Their results do not necessarily generalize to the whole city. A full victim survey of Edmonton in 1987 found an at-home rate of 10%, which is less than that for the US.

That would not explain why at-home burglary rates appear to be inversely correlated with gun-ownership rates.

How on earth can you claim this, Mark? After all, you have dismissed the NCVS (source of the at-home burglary rates) as flawed and the ICS (source of the gun-ownership) rates as fraudulent.

Charles Scripter writes:

BTW, I notice that your web page still seems to purport that your analysis was correct, even though your friends over in sci.stat.edu pointed out that it was not correct;

That’s an interesting interpretation of the discussion.

Perhaps you’d like to correct this “oversight”.

No problem, here’s something from one of my friends in sci.stat.edu:

Barry McDonald writes:

THE ARGUMENT ABOUT AUTOCORRELATION IN THE NSW HOMICIDE STATISTICS One complaint by Scripter about your data was to do with the evident autocorrelation in your data. It was this that was of interest to me. You see sometimes autocorrelation is not real, but just apparent- arising from having omitted a variable from the analysis. The Minitab output shows this is the case for your data. First I should explain that a Durbin-Watson statistic shows significant autocorrelation if D< DL where DL is a number from tables, clear non-significant autocorrelation if D>DU where DU is also from tables, and inconclusive results if DL<= D <= DU. Suppose we just fit a single line to the data:

MTB >  Regress 'homocide' 1 'year';
SUBC>  Constant;
SUBC>  DW.

Regression Analysis

The regression equation is
homocide = 47.3 - 0.0236 year

Predictor       Coef       Stdev    t-ratio        p
Constant       47.30       16.08       2.94    0.006
year       -0.023646    0.008381      -2.82    0.008

s = 0.5666      R-sq = 18.1%     R-sq(adj) = 15.8%

Durbin-Watson statistic = 1.23

Notice that if one just fits a straight line in terms of year then the _overall_ trend is downwards!! but the Durbin-Watson statistic indicates significant autocorrelation at the 5% level (D< DL=1.41) and nearly at the 1% level (1.21). The autocorrelation is cleared up by fitting a more appropriate analysis (see after graphs)

MTB > GStd.
MTB > Plot 'homocide' 'year';
SUBC> Symbol 'x'.

Character Plot


     3.20+
         -         x   x xx
 homocide-
         -                   x          x x
         -                  x      x
     2.40+                    x  x
         -      x
         -            x         x   x                             x
         -                            x              x
         -          x                      x           xx      x
     1.60+                                   x    x x     x
         -   x x                       x                    x x  x
         -                                                 x
         -                                       x
         -        x                           x x
     0.80+
         -
           --------+---------+---------+---------+---------+--------year    
              1904.0    1911.0    1918.0    1925.0    1932.0

MTB GPro.
MTB GStd.
MTB Plot 'FITS1' 'year';
SUBC  Symbol 'x'.

Character Plot


     2.40+   x
         -     xx
 FITS1   -        xxx
         -            xx
         -               xx x
     2.10+                   xx
         -                      xx x
         -                          x xx
         -                              x x
         -                                 x xx
     1.80+                                      xx
         -                                        x xx
         -                                             xx
         -                                                xxx
         -                                                    xx
     1.50+                                                       xx
         -
           --------+---------+---------+---------+---------+--------year    
              1904.0    1911.0    1918.0    1925.0    1932.0

MTB GPro.

Allowing for a change in intercept level with the law change in 1920:

MTB > Regress 'homocide' 2 'year' 'lawchnge';
SUBC>  Constant;
SUBC>  DW.

Regression Analysis


The regression equation is
homocide = - 40.4 + 0.0223 year - 1.18 lawchnge

Predictor       Coef       Stdev    t-ratio        p
Constant      -40.37       26.94      -1.50    0.143
year         0.02233     0.01410       1.58    0.122
lawchnge     -1.1769      0.3111      -3.78    0.001

s = 0.4841      R-sq = 41.9%     R-sq(adj) = 38.6%

Durbin-Watson statistic = 1.63

There is clear evidence _not_ to reject the hypothesis of zero autocorrelation if D>DU=1.52. This is indeed the case so the addition of this extra variable has simultaneously given us a much more believable analysis (see two lines below: not going down this time!!) and removed the apparent autocorrelation. The choice of this cut point (1920) has had a very significant effect (p-value approx 0.001). Note however that the trend in year is not significantly different to zero.

MTB > GStd.
MTB > Plot 'FITS2' 'year';
SUBC> Symbol 'x'.

Character Plot


         -
         -                             xx x
     2.40+                       x xx x
         -                   xx x
 FITS2   -             x xx x
         -         xx x
         -   x xx x
     2.00+
         -
         -
         -
         -                                                     x xx
     1.60+                                                xxx x
         -                                           x xx
         -                                      xxx x
         -                                 x xx
         -
           --------+---------+---------+---------+---------+--------year    
              1904.0    1911.0    1918.0    1925.0    1932.0

If we just assume constant rates of homicides before 1921 and constant after 1920, (i.e. zero slopes) then we retain a significant drop in the level of homicides, and the autocorrelation is still not significant. (though I would be cautious). The conclusion from this test is exactly the same as you were trying to do by a two-sample t-test except that in this analysis we have the added feature of checking whether the autocorrelation is significant.

MTB >Regress 'homocide' 1 'lawchnge';
SUBC>  Constant;
SUBC>  DW.

Regression Analysis


The regression equation is
homocide = 2.28 - 0.753 lawchnge

Predictor       Coef       Stdev    t-ratio        p
Constant      2.2762      0.1078      21.11    0.000
lawchnge     -0.7527      0.1612      -4.67    0.000

s = 0.4941      R-sq = 37.7%     R-sq(adj) = 36.0%

Analysis of Variance

SOURCE       DF          SS          MS         F        p
Regression    1      5.3221      5.3221     21.80    0.000
Error        36      8.7887      0.2441
Total        37     14.1108

Unusual Observations
Obs. lawchnge   homocide        Fit  Stdev.Fit   Residual    St.Resid
  4      0.00     1.0000     2.2762     0.1078    -1.2762      -2.65R 

R denotes an obs. with a large st. resid.

Durbin-Watson statistic = 1.52     

(D= DU=1.52 so there is no proof of autocorrelation ) Of course this analysis does not clear up Scripter’s complaint that (in his eyes) the law change year is irrelevant to homicides and so an adjacent year could be used to give a similar significant result. That is a causal matter that I cannot comment on as a statistician. - except that since a highly significant effect is apparent in the data, it really behooves him to come up with a better explanation than yours, especially as to why he postulates any other year as the change point.

Dr. Paul H. Blackman writes:

The Annest et al. numbers are based on seeking medical treatment in emergency rooms. And the figure for gunshot wounds is roughly 100,000 (which Phil Cook upped to 150,000 for the June JAMA by counting all gun-related wounds including pistol whippings and air/BB/etc. guns).

Prior to that, estimates from (primarily anti-gunners) hovered in the 200-250,000 range (except for a few imaginative folk who assumed all gun-related crimes resulted in injuries).

Kleck would dispute the Annest et al. NEISS-based figure on the theory that criminals try to avoid medical treatment for gunshot wounds, since seeking such treatment tends to invite police inquiry. As a result, Kleck would assume that most minor injuries inflicted on criminals by citizens using guns for protection are statistically unmeasured, and that the lack of enough measured medical treatments does not undercut any Kleck projections on the numbers of criminals wounded by civilians.

1. You seem to agree with me that the Annest number contradicts Kleck and disagree with Lott & Mustard who believe that it confirms Kleck.

2. You write “Kleck would dispute”. Are you sure that this is what he would do? In “Point Blank” he estimates the number of criminals wounded by civilians as 10,000-20,000. The numbers on which this estimate is based have not been overturned by any subsequent research to my knowledge.

Supporting the possibility of Kleck’s being correct would be a couple of medical studies on hospitals treating gunshot wounds and finding that most are treat-and-release — despite a certain medical penchant for keeping gunshot victims overnight as a precaution for possible shock — and finding that most drive-by shootings (which one might suppose are comparable in shooting ability/distance to defensive shootings) involve injuries to extremities (arms and legs), and are thus minor.

I’ll admit to a certain degree of ignorance about conditions in the US, but I find it hard to believe that drive-by shootings are similar to defensive shootings. Surely defensive shootings are not usually done from moving vehicles. Nor would I expect a 50% hit rate from a drive-by.

Criminals might well think, if it ain’t life threatening, avoid doctors. So might non-criminals who note the high number of medically-caused ailments and fear medical research into medicine might be on a par to medical research into firearms, esp. if they note how the Journal of Trauma encourages invasive treatment of gunshot wounds where that might worsen the situation.

If you compare the death rate from gunshot wounds now with that of a century ago it is quite obvious that medical treatment (most notably the use of antibiotics) makes an enormous difference. If criminals never get their wounds treated then you would expect a much higher death rate from gun shot wounds than the 15% estimate that Kleck uses in “Point Blank”. If we conservatively use that number for the death rate, it follows that at least 30,000 criminals are killed by civilians. Where do all the bodies go?

Dr. Paul H. Blackman writes:

Yes, I’m sure [Kleck] would dispute Annest, because he would assume injuries not treated in emergency rooms, and he’d probably stick with an estimate starting around 150,000 — unless I’ve misunderstood non-peer-reviewed conversations.

So how does he reconcile this with his earlier estimate of 10,000-20,000 such woundings? That estimate was computed from his estimate of 1400-2800 self-defence killings and the estimate of 15% wound mortality. Does he know believe that there are 30,000 self-defence killings or that wound mortality for defensive shootings is only 1%? Surely neither figure is credible.

If you compare the death rate from gunshot wounds now with that of a century ago it is quite obvious that medical treatment (most notably the use of antibiotics) makes an enormous difference. If criminals never get their wounds treated then you would expect a much higher death rate from gun shot wounds than the 15% estimate that Kleck uses in “Point Blank”. If we conservatively use that number for the death rate, it follows that at least 30,000 criminals are killed by civilians. Where do all the bodies go?

Criminals don’t “never” get their wounds treated; they avoid treatment whenever possible.

The Annest figures are compatible with at most 10,000 wounded criminals getting hospital treatment. (Assuming that at most 20% of the intentional shooting wounds treated at hospital are of criminals shot in self-defence.) That is, if there are 200,000 such wounded criminals only 5% of them get hospital treatment. It seems unlikely in the extreme that 95% of such wounds are trivial ones that allow the criminal to escape and require no treatment or just basic first aid.

While medical treatment has clearly improved since a century ago, there have also been improvements in non-professional treatment for injuries; general preventive activities not necessarily involving a post-shooting trip to the doctor (tetanus shots, etc.); and today’s bullets are much cleaner than yesteryear’s. In addition, criminals may seek out medical treatment which doesn’t get reported, something possibly akin to backroom abortions of pre-Roe/Wade days.

Even if there is a backroom gunshot wound treatment industry that is larger than the hospital one AND provides treatment every bit as good, then we would expect to see 30,000 dead criminals each year. I ask again: Where do all the bodies go?

[Originally posted to firearmreg on Aug 15 1996]

Daniel Polsby writes:

Lott’s results are highly plausible and internally consistent.

Highly plausible? Lets look at Dade county:

Lott reckons that the carry law caused a reduction of 8% in murders, 5% in rapes, 7% in aggravated assaults and 2% in robberies. For Dade county that translates to 1,500 fewer aggravated assaults, 450 fewer robberies, 65 fewer rapes and 30 fewer murders each year. From Cramer and Kopel’s paper on CCW (TN Law Review v 62p733) one learns that “the police kept track of every known incident involving [Dade] county’s more than 21,000 handgun permitees over a six-year period.” and there were 12 defensive gun uses by CCW holders against persons known to the police over the six year period. That’s two per year.

It does not seem highly plausible that those two uses prevented 2,000 crimes.

[Originally posted to firearmsreg Aug 16 1996]

Daniel Polsby writes:

Mr. Lambert, and for that matter most others on this list, assume that firearms are used defensively when they are brandished. All of the endless back and forth about survey research techniques of establishing how often this sort of thing happens has embedded this assumption.

No, my comments were directed specifically at the deterrence theory. IF concealed weapon permit holders used their weapons frequently (say two or three times a week) and IF these incidents were given wide publicity (so that criminals were made aware of the risk they faced) THEN you would have a plausible mechanism by which a concealed wepons law could cause a large decrease in crime.

Economists’ model of deterrence treats the prospect of meeting armed resistence as affecting, ex ante any misbehavior, the apparent present value of a given crime. The prospect of meeting an armed defender or an armed good Samaritan should be expected to do two things: (1) raise the absolute expected cost of confrontational crimes and thus (2) increase the relative value of other potential uses of ones time, including non-confrontational crimes.

Compare: 2 encounters with armed CCW holders per year in Dade county with 10,000 arrests by armed police for violent crime per year in Dade county. The chance of meeting an armed CCW holder is negligble. In an economic model of deterrence it will be swamped by the 5,000 times greater chance of being arrested.

Incidentally, in case the point has been lost, I have never questioned the integrity or seriousness of McDowall and his collaborators.

Pardon? You accused McDowall et al “of taking the extra steps necessary to locate a subset of the data that would yield a particular result.” It certainly sounds to me that you are accusing them of dishonest data selection.

Suffice to say that no conversation about causation and correlation can be had without a reasonable grasp of David Hume’s argument on this subject, which was, in brief, that “causation” as such can never be demonstrated. How do you know, if you drop a marble, that your dropping of it has “caused” it to go onto the floor? How do you know that this result is not, simply, highly correlated with dropping?

Absolute certainty is of course impossible. Scientists generally accept a replicable controlled experiment as good enough for showing cause and effect for practical purposes. A non-experimental study like Lott’s is merely suggestive of causation. When you have a pile of them and well-understood and verified mechanisms (like for smoking-cancer) then you can maybe make some causal claims. Lott says: “If those states which did not have right-to-carry concealed gun provisions had adopted them in 1992, approximately 1,570 murders; 4,177 rapes; and over 60,000 aggravate assaults would have been avoided yearly.” This statement needs at least two caveats that Lott does not give us.

jmaraldo writes:

Why assume that only the defensive use deters, that the knowledge the intended victim and civilians in the vicinity may be armed does not also deter? It seems to me that a shall-issue CCW system is like a sign which says “Premises Protected By ** Alarm” in that the many more burglars are deterred than actually trigger the alarm and the sign often works even though there is in fact no alarm installed.

But CCW holders do not carry signs saying “Person Protected By Smith and Wesson”. What is it that is telling criminals that they are facing a risk from CCW holders? It can’t be from personal experience in encountering one, or media reports of CCW holders thwarting crime, since these incidents are almost nonexistent.

Tom Tancredo (really Dave Kopel?) writes:

Tim Lambert writes:

It does not seem highly plausible that those two uses prevented 2,000 crimes. “

Sure it’s plausible. First of all, there is some unknown number of DGUs by permitees that were never reported to authorities. It’s possible that 90% may be unreported, since there’s no real need to tell the police that you drew a gun, and the criminal ran away.

If Kleck’s DGU survey is to be believed 64% of DGUs are known to the police. In any case, even if there are another 20 unreported ones, these aren’t going to have much of a deterrent effect since only one or two criminals will be aware of them.

More important, as Kleck has shown in various cases, wide publicity about the prevalence of armed citizens can sometimes have a huge deterrent effect, all by itself.

Firstly, I do not believe that Kleck has shown this at all. Publicity in Orlando and Kansas City was associated with a temporary but not statistically significant reduction in crime. His arguments about Kennesaw are not even internally consistent.

Secondly, if the mechanism to be invoked is that publicity about the introduction of CCW laws reminded criminals of the risks that they faced from armed victims, then Lott has used the wrong model. Such an effect would be as temporary as the publicity, so his model should be for a temporary effect, not a permanent one.

[Originally posted to firearmsreg Aug 19 1996]

Daniel Polsby writes:

McDowall very freely interprets his five-county study as suggestive of causation. I tend to share the view that one should be slow to change public policy on the basis of a single study, though one might say of Lott-Mustard that it amounts to at least 610 McDowall-Loftin-Wiersema studies, as it covers that many times more counties, to say nothing of controlling for a lot more variables. I should say, however, that ordinarily the heavy lifting of causation involves the existence of a theory which models how the world works. The Lott-Mustard results are generally consistent with such a theory (the price theory model).

Qualitatively consistent in that an increase in the price of crime was associated with a decrease in the volume of crime, yes. Quantitatively consistent, no. Arrest is 5,000 times more likely than having a CCW holder pull a gun on you. If the criminal considers these to be equally bad outcomes, then the CCW law increases the cost by no more than 0.02%. It is absurd to expect a 7% decrease in crime from such an insignificant change in the cost.

Even if (contrary to what criminals said in the Wright-Rossi study) criminals are not afraid of police guns but are afraid of victims guns. The change in the cost is insignificant. If you believe Kleck there are about 20,000 DGUs in Dade county each year. If you believe the NCVS it’s more like 500. Either way, the two by CCW holders is insignificant.

A general point about deterrence theory. The theory doesn’t tell us when a person who is supposed to be deterred comes to believe that a threat is credible. It doesn’t assume, in other words, that there will be a certain number of dead criminals before people start believing that if they commit a crime they may be shot. Nuclear deterrence works in exactly the same way; nobody wants to find out how credible that sort of threat actually is.

Ahem. Since no nation has actually attacked NATO it is reasonable to suppose that if one did attack NATO there is a good chance of starting a nuclear war. On the hand with 50,000 or so violent crimes in Dade county each year and only 2 DGUs by CCW holders it is not reasonable for a criminal pulling a violent crime to expect to be thwarted by a CCW holder.