Deltoid

brian.m.leary said:

In the five months after the passage of the mandatory gun ownership law in Kennesaw, Georgia the residential burglary rate was down 89% from the same period the year before. Does this prove the law worked? No - proof is difficult in these matters.

However, is it clear that the law had no effect? Hardly.

The source for this claim appears to be Kleck’s paper in “Social Problems” v35p15, where he states there were five reported residential burglaries in the seven months after the law, while there were 45 in the corresponding seven months of the preceding year.

As you have noted, this isn’t enough data to conclude if the law did or did not have an effect. Sure, there is a big reduction, but how do we know that before year was not unusually high?

If the sequence is 7 43 4 5 6 45 5

we certainly wouldn’t conclude that the law caused the reduction.

However, there is a MUCH more serious problem with this 89% reduction claim. From the “Criminology” v29p541 paper by McDowall et al (which provides monthly totals) I computed the total number of burglaries in the seven months after the law (23) and for the seven corresponding months of the previous year (37). Something is wrong here. It’s possible that 5 out of 23 burglaries after the law were residential but it sure as hell ain’t possible for 45 out of 37 burglaries before to be residential. One dataset must be wrong.

McDowall et al’s data come from the FBI’s Uniform Crime Reports. Kleck’s data comes from a telephone conversation somebody else had with the Kennesaw police chief.

I think the UCR data is likely to more accurate.

brian.m.leary said:

The residential burglary rate in Kennesaw, Georgia dropped sharply after a city ordinance requiring heads of household to keep at least one firearm in their homes was passed. The law passed early in 1982. In 1986 the rate was still down 85% compared to 1981. (1)

This statistic is essentially meaningless. If the crime rate fluctuates, then by picking the right two years to compare, you can get any result you want. To make a credible case, you need to provide data for at least the ten years 77-86. I haven’t seen Kleck’s book, but in his “Social Problems” paper, he uses figures for residential burglary that are contradicted by the official UCR data.

What about that study that shows no drop in burglary in Kennesaw? Well strangely, the authors used the total burglary rate not the more relevant residential burglary rate.

Does the UCR data break burglaries down into residential and non-residential ones? If not, then in the absence of reliable data on residential burglaries, it is best to use total burglaries. Since this includes residential burglaries, it would be surprising if a reduction in the residential rate did not appear as a reduction in the total rate.

In another strange error(*), the authors did not account for Kennesaw’s seventy percent population increase over the period. (2)

The population would have to increased by over 500% between 81 and 85 for there to be an 85% drop in the population adjusted rate, and no reduction in the actual number of burglaries.

The population increase would occur gradually over the period and could not mask an abrupt change in the burglary rate, so dividing the number of burglaries by the population will not change the result of the interrupted time series analysis: Here are the burglary rates 76-86, assuming a population growth of 70%, normalized to the 76 population value. Each x represents 5 burglaries.

xx             xx             xx
xx       xx x  xx x  x        xx
xx xx xx xx xx xx xx xx xx xx xx
xx xx xx xx xx xx xx xx xx xx xx
76 77 78 79 80 81 82 83 84 85 86
                  ^
                  Law        

The rate does not appear to be lower after the law.
Average 76-81 is 29
Average 82-86 is 26
The average after is a little lower. To test the significance of this, I use Student’s t-test. I compute t=-0.13 with 9 df, p=0.51, that is, 51% of the time chance will produce a change this large. This obviously not significant.

In any case, a population increase of 70% weakens any evidence for a deterrent effect. Such rapid growth means Kennesaw changed significantly during the period in question, so even if there was a significant reduction in burglaries, we don’t know whether one of the other changes in Kennesaw caused it.

In other words, the NCS only counts defensive uses against crimes.

Andy Freeman said:

Wrong. NCS doesn’t get into defensive uses unless the victim thinks that a crime occurred even if it was successfully self-defended against. As in “Have you been the victim of a crime?” Someone who successfully self-defended against an armed robber might well answer “no”. They weren’t a victim. Yet, that’s a self-defense against crime, one that the NCS wouldn’t ever find out about.

In fact the NCS does not ask “Have you been the victim of a crime?”. Andy appears to have made that question up. The relevant questions to Andy’s example are “Did anyone TRY to rob you?” and “Did anyone threaten you with a weapon?”

Let’s suppose for the moment that Andy’s claim is true, and that this explains the enormous discrepancy between the NCS estimate of defences (80,000), and the 1,000,000 figure that some people here seem to believe. What are the implications?

(1) The people who designed the NCS surveys are incompetent. NCS is supposed to measure attempted crimes, but only finds out 10% of them.

Wrong. That isn’t an implication. NCS is interested in victims. These people don’t necessarily consider themselves victims, and NCS isn’t interested in finding the ones that don’t.

Can you provide some evidence for this remarkable claim? Like an NCS document where they state “we’re only interested in people who consider themselves victims”. Or perhaps a screening question where they ask “do you consider yourself a crime victim?” OR did you make this one up too?

(2) NCS data on completion rates of crimes is utterly worthless since it is based on only 10% of attempted-but-not-completed crimes.

Wrong again. Some people may be victims even though the crime wasn’t completed.

So? How does this address the point I raised?

People who successfully self-defend are not necessarily like everyone else who was a target of an unsuccessful attack.

Yes, so if completion rates are lower for “self-defence with a gun” you cannot infer that guns are more useful for self-defence.

(BTW - Consider the crime “attempted murder”.)

Certainly. A victim survey on murder would discover that the completion rate for murder was 0%, irrespective of self-defence measures used. Because the survey is biased (towards people who are still alive) it tells us nothing useful about the success rate of would-be murderers. Similarly, if the NCS is, as you claim, biased away from successful defenders, it tells us nothing useful about crime completion rates.

NCS data can be quite useful for many purposes without being able to tell us about everything.

Certainly, but you want to reject the bits that do not support your political beliefs and accept the bits that do.