rev. 10/3/99


RETURN TO "Wintemute" PAGE

The gun control crowd seeks to eliminate the cheapest handguns.  The reason they give publicly is that these guns are preferred by criminals or some class of criminal.  This is the gun controllers' "gun of choice" argument.  The implication is that the guns must have some special value to certain criminals that they don't have for the general gun-owning public, because of some unspecified criminal features or something.

The gun control advocates point to information about the frequency with which various guns are traced by the US Bureau of Alcohol, Tobacco and Firearms (BATF), for police nationwide, as evidence that cheap guns are used more in crime than other guns are.  This claim works well on the public even though it ignores the facts that:  (1)the high incidence of a specific firearm in the list could be because far more of that particular firearm are in circulation; and, (2)the BATF traces are not all of guns used in crime and are not traces of all guns used in crime.

Part of the reason this claim has been successful with the public is that some police and police organizations have been taken in by it and, as a result, have supported bans on the cheap guns.  The public expects that police are qualified to know about such things.  They don't realize that police recruits have, on average, about the same intelligence as office workers, bank tellers and salespeople.  (Sep 9 '99 LA Times article dealing with the fact that a Connecticut police department has a policy of not hiring people that are too intelligent.)

Although the gun controllers have not had to use it, they have already prepared for the possibility that the public might come to understand the deficiency of the BATF trace numbers.  One of their chief medical advocate researchers, Dr. Garen Wintemute, has done some taxpayer funded analysis of records on firearms held in evidence in California, and compared the numbers with a crude measure of how many of the same firearms were in circulation in the state for the time period in which the firearms were seized by police.  His analysis determined that some of the cheap handguns made by the California handgun makers are two to three times as prevalent in police evidence rooms as they would be based on the numbers sold in the state.

The problem is that there is a basic deficiency in the doctor's analysis (besides the inexact measure of prevalence among the general public).  That problem is that the analysis actually "proves" something that should be obvious, and it is not that criminals prefer certain handguns.  The problem is simply attaching an incorrect significance to the results of the analysis.  More about this later.

So, what would it take to indicate almost with certainty that criminals prefer a certain handgun more than the noncriminal public, thus implying that such handgun must have some special value or appeal for criminals?

First, identify a large group of randomly selected criminal handgun users/possessors, or criminals of a special class if the alleged preference is one of that class of criminal.  Then identify a group of randomly selected noncriminal handgun possessors that is exactly like the criminal group except for the criminality and gun acquisition or possession.  That is, match characteristics of the individuals in the two groups.  This second group is called a "control" group.

Then compare the rates at which the different kinds of guns are acquired or possessed in the two groups.  If these groups differ significantly in the rates of acquisition or possession, this would indicate almost certainly (lacking any other explanation) that criminals have a handgun preference that is different than the preferences of noncriminals.

It is not possible to get a control group that is exactly like the criminal group--that is, the same in all respects.  And, it is not actually necessary that the two groups be exactly the same.  Valid results can be obtained without identical groups in either of two ways.

The groups can just be the same with respect to any factors that would affect the outcome being investigated....acquisition or possession rates in our case.  Or, if the groups are large enough and the differences are recognized and quantifiable, the differences can be taken into account using mathematical (statistical) methods.

Analysis involving groups that aren't identical is only possible if one can identify the characteristics or factors that are likely to have a bearing on the result being investigated.  In our case, we would have to have an idea about what would affect specific gun model acquisition rates.  Such factors would include features of the guns, their prices, the buyers' financial status, and maybe some other things, just as for an analyis regarding other products.  The usual statistical analyses of such things typically account for subject demographics such as race, sex, age, residence urban nature, country region, and possibly other readily available data just in case any of these have an impact on the results.

Because the objective of Dr. Wintemute's analysis was to determine if cheap guns are preferred by criminals, meaning that gun cost is the topic, it is obvious that financial status of the subjects is one thing that should have been accounted for.  That is, it is likely that income levels of the criminals and gun buyers could be the cause for any difference observed between criminals and noncriminals.

Well, Dr. Wintemute did not consider buyer financial status in his analysis.  And, the fact is that criminals tend strongly to have incomes that are low in relation to the noncriminal public or the public at large.  (Data on both jail and prison inmates show this.) So, it is natural that criminals would tend to select cheaper handguns, just as other low-income people do.  What he actually proved is the obvious:  people with little money select cheaper guns, just as they select cheaper products in general.

The problem with Wintemute's work, for the purpose of acquiring valid information as opposed to politically useful information, is that he compared criminal use rates to measures of possession rates of the general public.  What he should have done is compare to possession rates of noncriminal handgun owners that are otherwise like the criminal group, especially where financial status is concerned.

But what about the overrepresentation in BATF traces or observations by police?  Well, these "measures" suffer from the same basic problem as Wintemute's.  They are both observations of frequency relative to the total population of handguns, not just the population possessed by people that are like the criminals (as in income) except for the criminality.