Was the Dunblane Inquiry Misled?

The justification for the handgun ban proposals of the government and opposition parties was a claimed link between the general level of firearms ownership and firearms violence. The key international evidence of this link was submitted by the Home Office to the Dunblane inquiry. Whilst Lord Cullen acknowledged that the quality of this evidence was open to severe criticism, he allowed that it "shows that there is a relationship… when considered overall" (Cullen, 9.21).

It was only after the closure of the evidence to the Dunblane inquiry that Lord Cullen received James Hawkins's reductio ad absurdum demonstration of the statistical weakness of this notional "relationship": over the same set of countries, firearms homicide correlates closer with car ownership than it does with firearms ownership.

The Home Office, however, had further evidence of the relationship between firearms violence and firearms ownership levels in the UK itself, which it did not submit to Lord Cullen. The Home and Scottish Office's own figures on armed crime and firearms ownership by constabulary area show a negative correlation. In Britain, in areas where legal firearms ownership is higher, armed crime is lower: and this is the case both for total offences and for every individual category of armed crime. Moreover, factoring out the differences between urban and rural areas yields the same result: it is not just the total, but the proportion of crimes committed with firearms that is lower where licensed firearms ownership is higher.

It is regrettable that the Home Office should have presented Lord Cullen with a study which (setting aside its specific flaws and weaknesses) suffered from the inherent fragility of any international statistical comparison, whilst failing to present him with the United Kingdom's own data which pointed to diametrically opposite conclusions. The failure to present this evidence to Lord Cullen has fundamentally undermined the firearms aspect of his report, and the validity of the political decisions associated therewith.

P H Jackson, J A G Hawkins, Prof. A R Horrocks, R A I Munday; 9 November 1996

Notes:

The Home Office submitted the Killias research in support of the claim that firearms homicide rates are related to firearms ownership levels in the countries surveyed. As can be seen from the scatter plot of the data (overleaf, top), the quality of fit of the data was not good, and this is reflected in the trivial correlation coefficient. To indicate how trivial this correlation is, the firearms ownership data is replotted (right), together with car ownership data from the same sixteen countries. Car ownership yields a greater correlation with firearms homicide than does firearms ownership.

Domestic data (right): firearms homicide and attempted murder and other acts endangering life per thousand population plotted against numbers of firearms licences per thousand population, for each police force in England & Wales, for the period 1992-94. Unlike the Killias data, this data set is complete, contains no estimates and uses statistics recorded in a consistent manner. It also gives a quite different result. Within the United Kingdom, areas of high legal firearms ownership have a low incidence of firearms homicide and attempted murder.

The graph plots all firearms offences excluding criminal damage (mostly airgun offences) for the same period, and shows a similar inverse correlation. With the much larger number of incidents in this category, it is clear that the relationship is stronger than linear, as indicated by the inverse-square curves shown. These are Home Office figures, and are directly relevant to Britain.

The Home Office also claimed, on the basis of the Clarke & Mayhew studies, that in countries with high firearms ownership, a higher proportion of homicides were committed using firearms. If we plot England & Wales data on the same basis, we see that the rate of use of firearms in all categories of crime is lower in areas of high legal firearms ownership.

Scotland: recognising that crime categories and recording policies vary between jurisdictions, we have analysed Scottish Office data separately. The Scottish analysis yields very similar results to those for England & Wales.

The authors invite critical review. The full set of data and calculations summarised in this paper is available as crimstat.zip in the form of Excel spreadsheet files bundled into a Zipfile.

Send mail to Peter at:

Peter H Jackson <pjackson@forge.demon.co.uk>