rev. 5/2/03

In early 2000, the media ran stories that some democrats (especially schumer) were calling for the Bureau of Alcohol, Tobaco and Firearms (BATF) to go after "suspect" gun dealers and that a study had shown that the suspects are: (a)some FFL holders who sold no guns over some period; and (b)about 1000 dealers which are a small portion of the total but are "responsible" for a very large portion of the BATF "crime" gun traces.  The media of course had no idea that there are some legitimate reasons to have FFLs other than to sell guns--like being able to handle transfers for gun club members or gunshow attendees in a state where private transfers require government permission (background checks).  Clinton apparently ordered the BATF to start increased checking of dealers that had more than a threshold level of traces.

If one doesn't think about it too much, it seems reasonable that a dealer must be crooked if a large portion of crime gun traces are for guns sold by that dealer.  But, if you think carefully about it, it turns out that this is faulty (though simple) logic.

The problem with the logic is that having a lot of guns turn up in BATF traces is not an indication of wrongdoing by a dealer, because there are much more significant reasons for the numbers of BATF traces associated with a dealer.  Besides the fact that traces are done for all kinds of reasons, some not even associated with crime, the trace frequency is most likely the result of other factors.

Besides the possibility of a dealer being or employing a crook or careless person, the frequency at which a given dealer's sales end up being used in crime--or simply possessed by someone who maybe did the "crime" of trying to protect someone where the government doesn't allow this--is determined basically by:  (1)the number of firearms the dealer sells, and (2)by the portion of potential gun buyers in her market area who are criminals or become criminals.

So, a proper analysis to identify potentially crooked or negligent dealers would look at the numbers of guns (preferably only those actually used in violent crime and preferably by year in which sold by the dealer) over a period of time, divided by the dealer's sales volumes for the different years.  That is, it would look at portion of sales per period that are "used in crime."

A shortcoming of this is that it would allow high-volume dealers to "sneak through" a few illegal sales without detection.  This is a statistical limitation.  To catch such transactions, the authorities simply cannot use statistics as a valid basis.  The law enforcement would just have to get out of their offices and investigate.  If one is going to use some statistical analysis to identify "suspects" to dump resources into investigating, the analysis must be a reasonably valid one in order that the resources aren't wasted and unnecessarily diverted from really useful efforts.

A proper analysis would also account for the portion of the people (or, better yet, gun buyers) in the dealer's market area who are criminals or are likely to become criminals.  You see, the risk a dealer has that she will innocently sell a firearm to a criminal or person who will become a criminal is proportional to the portion of the dealer's market who are criminals or will become criminals.  The more criminals there are coming to a gun store, the more likely it will be that the store will innocently sell a gun to one.  Alternative measures (proxies) that could be used to account for this might include the violent crime rate in the market area, the poverty rates of the citizens, or some combination of factors known to relate to criminality level in a region.

But, if the risk of selling to criminals is increased by a dealer being in a high-criminality neighborhood, why not just prohibit or severely monitor sales in such neighborhoods?  The reason is that these are the same neighborhoods where people have the greatest need to have a firearm for personal protection.  To do net good we need to do things to keep criminals from getting firearms to a greater extent than we keep potential victims from getting firearms.

Neither the BATF nor anyone in the clinton administration did a valid analysis of the situation--or they're just dishonest.

On Feb 4 the NRA had a press release about the clinton directive.  In this press release, the NRA took the tack that the new policy was just an attack on dealers who sell too many guns or too few guns.  In the Aug. 4 issue of the journal of the American medical assn, garen wintemute had a letter with info about a little analysis he had done to check out the claims that some had made (referencing the NRA press release) that the number of traces for guns from a dealer "simply reflects the dealer's sales volume."  This letter also said that clinton's administration "is now requiring dealers linked to more than 10 firearm traces in 1999 to undergo more intensive surveillance."

Wintemute's simple analysis did find that 1998 traces per California dealer was correlated with the dealer sales volume for '96 through '98 to an extremely high degree.  He found that there was significant variation between dealers for the ratio of traces to sales volume, although the variation would not be as great if he were to use a valid trace number.  His ratio, which he called a "crime gun trace ratio" (CGTR), would have been a better measure had he excluded the traces of guns that were not first sold in one of the three years for which the sales volume was used.

But, wintemute basically confirmed that sales volume was the most important determinant of the trace frequency.  Maybe with a gun controller telling the BATF this, the BATF will change their harassment criterion to something more valid and reasonable.  Now, if someone can show them that criminal population density in the dealer's market area is also highly significant, maybe the BATF could be convinced to use fully valid criteria--although the criteria would probably point them at very few dealers.  Although wintemute found that the BATF criterion was not valid, he did not have the integrity to specifically say so in his letter to the jAma.


What would be the effect of "going after" (harassing) "suspect" dealers on the faulty basis of simple numbers traced per year?  In the first place, it would harass innocent dealers, increase their costs, and possibly discourage some of them out of the business while increasing the BATF bureauracracy and costs paid by us taxpayers.  It would tend to unjustifiably eliminate highest-volume dealers.  This in conjuction with the "crackdown" on dealers who don't sell "enough" would eliminate dealers at both ends of the sales volume spectrum.

Because the highest rates of criminal use or possession of guns are in the high-crime communities and neighborhoods, and these "just happen" to be the communities and neighborhoods where people tend to have low incomes, the policy would discriminate against the poor and the minorities that tend to be among the poor (by eliminating or increasing costs for the dealers that serve the poor areas).

And the faulty policy would ensure that the real crooks and sloppy dealers won't be investigated.  A criminal dealer who had only one gun used in crime would be overlooked even though it was the only one he sold!

BATF should not be funded to pursue "suspect" gun dealers unless BATF can demonstrate that they have the sense and integrity to determine a legitimate basis for identifying the "suspect" dealers.  So far, it appears they aren't competent to do the job, or honorable enough to do it in an honest way rather than a political way.