Deltoid

Steve D. Fischer writes:

There are two kinds of lies to worry about. The first is making up defensive gun usages (DGUs) which never occurred, or which did not occur within the 1 year time period. The second is concealing DGUs that did occur because you feared that your DGU might have been of question- able legality. We’ve talked at length about the second circumstance. Let’s look only at the first, then.

Kleck reported 213 DGUs in a sample of 5,000. Assuming 59 million gun owners, that leads to about 2.5 million DGUs per year.

The NCVS estimated 80,000 DGUs in that same gun-owning population, which leads to about 88 DGUs in a sample size of 65,000. If the NCVS estimate were “correct” then Kleck should have only seen about 7 DGUs in his smaller sample size. Kleck’s “excess” 206 DGUs must have been lying, then. This amounts to 4.1% of Kleck’s sample.

But wait a minute, if both studies took good random samples of the population, then wouldn’t we also expect 4.1% of the people who responded to the NCVS to make up DGUs also? There is no reason to assume that people who make up DGUs are not randomly distributed in the population, because roughly 48% of all households own at least one firearm. Where, then, are the (roughly) 2670 liars in the NCVS sample?

Good question. It is possible that 4% of the population make things up. If you ask them if they’ve used a gun for defence, they’ll say “yes” and invent some details. If you ask them if they’ve seen a UFO, they’ll say “yes” and invent some details. If you ask them if they’ve been burgled, they’ll say “yes” and invent some details. So the 2670 liars in the NCVS sample would invent crimes, but not necessarily invent DGU as one of the details.

The NCVS interview is preceded by a phone verification of the location of the residence and persons who live there. This gives the respondents time to make up a good story, if that is their intent.

True, but a story about crime victimization does not have to include DGU.

In the Kleck poll, there in no pre-interview, and respondents must make up a lie contemporaneously, in such a way that the details will be relatively consistent.

Though not too consistent. The numbers saying somebody would have died, the number saying police knew of the incident, the number saying they wounded the perp, the number saying that they killed the perp are all impossible. I also note that in 85% of Kleck’s cases the person involved was the person interviewed (rather than someone else in the household). As Kleck notes (p165) this is surprising since most households contain more than one adult. Kleck suggests that this is because people may be reluctant to “inform” on other household members. Alternatively, it may be because only made stuff up about themselves rather than others.

Steve D. Fischer writes:

Now, you’ve got 100 lines to convince me WHY I should risk getting myself into trouble with the law, when all I wanted to do was report a crime that happened to me.

The trouble is that Kleck would have you believe that this accounts for the discrepancy between his survey, that is, 97% of DGUs are by people who think they would get in trouble for it AND that these people would tell his interviewer but not one working for the NCVS. That is 97% of the population simultaneously believe that the government will lie to them (by promising confidentiality) and that the government would not lie to them (by pretending to be an interviewer working for Kleck). This is absurd.

I think the 97% figure is way too high, also. I have no problem with 50% though.

If it is 50% then the NCVS underestimates by a factor of 2 i.e. this implies 160,000 DGUs.

It’s all just guesswork. The only way to get at this is to build in some kind of “lie scale” into the survey. You have to find some way to get at the respondent’s level of trust of the government - you can’t do that in a government survey, however, like the NCVS. As for “making up” DGUs, I have no crystal ball for estimating the magnitude of this. Perhaps you have some suggestions?

85% of Kleck’s DG users reported that they personally (rather than some other household member) used the gun. The average household size is just slightly under 2, so only 51% of uses should be personal rather than household ones. The discrepancy may be because people are more likely to make up DGUs about themselves rather than others. If this is true, the difference between 85 and 15 (the number that would be equal to the number of DGUs by other household members) are “made up” DGUs. I.e 70% of Kleck’s DGUs are made up, implying 450,000 DGUs after correcting for “made up” ones.

160,000 vs 450,000 is still an enormous gap, but at least they’re closer….

You see, the NCVS has a problem too. Kleck reported 213 DGUs in a 5000 interview sampling. The NCVS interviewed something like 65,000 people as I recall, and “found” 80,000 DGUs. If you assume the NCVS is God’s honest truth, then Kleck should have found only 7 DGUs in his sample of gun owners. That means 206 people lied.
That’s only 4.1% of Kleck’s sample. Lying on that scale is certainly not hard to accept, but this still poses a problem for the NCVS.

The NCVS should have found something like 88 DGUs if you assume, as Kleck did, that there are 59 million gun owners in the USA. If you also assume that the same number of people lied to the NCVS as lied to Kleck (4.1%), then 2665 people should have reported a DGU to the BJS! Both surveys used random sampling, and the number of households owning guns (48%) is large. There is no reason, a priori, to assume that more people would invent DGUs to a greater degree in one poll than the other.

So where are the missing liars in the NCVS poll?

Are we to believe that the NCVS filters out 97% of the liars who wrongly report DGUs? I find that rather hard to believe, and I’ll bet you do too?

They may lie and and make up crimes, but not necessarily invent the detail of gun defence in their fictional crime.

Some people will keep quiet about DGUs, but they’re not going to tell Kleck either.

Well, now that’s the crux of the matter, isn’t it? It all hinges on two differences between face to face polls versus phone polls.

No, because the NCVS is mostly conducted by phone. (Some interviews are face to face but the great majority are over the phone.)

kebarnes writes:

Are Kleck’s numbers concerning the self-reporting of robbery and burglary incidences from this survey out of line with the comparable NCVS results, for instance? Rs to Kleck’s survey reported that 5.5% (274/4977 Rs) had been a burglary victim within the past year, and 2.5% (124/4977 Rs) had been a robbery victim within the past year. This would imply (if I’m correct) some 242,600 robberies and 533,800 burglaries.

You dropped a decimal: it’s 2.4 million robberies and 5.3 million burglaries. And 500,000 (20%) of robberies where a gun was used for defence. And 850,000 (16%) of burglaries where a gun was used for defence. That last percentage is especially remarkable. With only 14% of burglaries occupied ones and 50% of US households having guns a generous upper bound on the percentage of burglaries where a gun is used for defence is 7% which is considerably less than 16%.

How does this square with other sources for 1993?

According to the 1992 NCVS figures (the latest available to me), the percent of households reporting a burglary in the past year was 4.2%, and the percent of households reporting a robbery in the past year was 1.0%. An approximate ratio for burglaries to robberies in 1993 according to the FBI’s Uniform Crime Reports is about 3.7 to 1, which isn’t substantially different from the NCVS reporting (if you exclude robberies and burglaries of commercial buildings and banks, which Kleck’s Rs are unlikely to report). If the NCVS numbers on the incidence of burglary and robbery are reliable, then there might be some argument for ‘telescoping,’ but this depends whether these differences are within the sampling error of the survey.

A 95% confidence interval for percentage burgled is 4.5%-6.7%, for percentage robbed it is 1.8%-3.4%, so the difference are not within the sampling error of the survey. (In case anyone wants to check my calculations, the sample size was 1660 not 4977 since the full questionnaire was only given to 1/3 of Kleck’s respondents.)

Another way to estimate DGUs is to ask criminals how often they were thwarted by armed victims. This is what Wright & Rossi did. 34% of the criminals that they surveyed admitted this. These criminals had an average of 10 prior arrests, so if the 34% who had been thwarted had an average of two thwartings each, it would appear that being arrested is roughly 15 times as likely for these criminals as being stopped by an armed victim. Combine this with about 988,000 arrests per year (this number from Kleck’s “Crime Control” paper) and we get a (very rough) estimate of 70,000 crimes stopped each year with guns. The uncertainties in this calculation are large: it might be off by a factor of two or three, but a factor of thirty is implausible.”

Lambert’s really grasping at straws with this analysis, since he’s making the assumption that the criminals who are actually incarcerated are not only a representative sample of the criminals involved in DGUs with victims,

No, I’m assuming that criminals involved in a DGU are no more than two or three times less likely to be imprisoned. Is it your contention that a DGU makes it LESS likely that the criminal will be captured?

but are honest enough to admit to something which would be personally embarrassing to them.

They seem to have admitted to a bunch of other embarrassing things. Maybe 50+% of Wright and Rossi’s respondents lied. Maybe 4% of Kleck’s lied. Take your pick.

What gets me is that Kleck actually cites the 34% statistic as SUPPORTING his study when in fact it contradicts it!

He’s also assuming that criminals who are arrested (using Kleck’s 988,000 figure) are representative of the criminals involved in DGU incidents with victims, or are anywhere near a reasonable estimate of the number of actual violent crimes attempted each year.

You are confused. The comparison is based on often these criminals were arrested versus how often they faced an armed victim.

We can based on the number of violent crimes if you prefer: They averaged 50 violent crimes each in their careers versus .67 times they faced an armed victim. That’s 75 to 1. Dividing 10 million violent crimes by 75 gives an estimate of 130,000 DGUs. Again, nowhere near Kleck.

Steve D. Fischer writes:

I have no problem accepting the idea that respondents lie about reporting incidents to the police. From my own experience, I know that people tend to disbelieve a report of a DGU if you say you did not report it. The tendency to lie on this question is high. Because one lies about reporting to the police, it does NOT mean they made up the DGU.

Yes, just because Kleck’s DG users gave untruthful answers on the questions of whether the incident was known to the police, on whether the perp was wounded, on whether the perp was killed, on whether someone would have died, and on whether the incident happened to the person first contacted (rather than some other household member) it doesn’t necessarily follow they gave untruthful answers to other questions. On the other hand, it doesn’t exactly inspire confidence in the veracity of the other answers.

The number of woundings is still an open question. There is a reference in the Kleck paper, which I don’t have in front of me now, that suggests the number is greatly under-reported anyway. A criminal is only likely to seek hospital treatment for wounds that he/she considers life threatening.

Kleck has used the estimate that 15% of gunshot wounds are fatal. If criminals don’t get treatment, the percentage will be even higher. So there should be 30,000 dead criminals every year. Where do all the bodies go?

Respondents call anything from a mere scratch to a near fatal shooting a “wounding.” They could also simply be mistaken about having wounded the attacker (”I couldn’t have missed him from THAT distance!”).

Kleck’s own estimate of the number of woundings associated with DGUs is 10,000-20,000. It would seem that 400,000 people fire at the perp and only 2.5%-5% actually hit. Mighty poor shooting.

The 97% figure is a bit deceptive, like your use of 90%. In actuality, 213 liars would constitute only 4.1% of Kleck’s sample.
However, if 5.1% of the NCVS respondents suppressed a DGU, that would represent the 650 un-reported cases I referred to earlier. Therefore, a modest rate of lying or suppression could lead to widely divergent results in both polls.

This is very misleading. Whether you believe Kleck or the NCVS, most people have not had a DGU, so could not possible conceal it from the NCVS. The correct way to measure the percentage of liars is:

(Number who lied) / (Number who could have lied)

If Kleck is right and the NCVS is wrong that ratio is

(29,000) / (30,000) or 97%.

If the NCVS is right and Kleck is wrong then Kleck should have found 7 DG users and that ratio is

(194-7)/(4977-7) or 4%.

4% liars or 97% liars, take your pick.

Steve D. Fischer writes:

more threatening than a govt official sitting across a desk from you?

(1) not sitting across a desk

Irrelevant. The key is that they can look you in the eye and read your body language.

Yet one more time: The NCVS used to be face to face but now most interviews are conducted by phone (except for interviews of people without phones, of course).

(2) 97% of the people cooperate with the census NCS survey

Co-operate does not mean tell truth. It simply means they agree to answer the questions.

The sensible thing to do if you want to conceal something from the NCVS is not to answer ANY of the questions. Some may not think of this, by a 97% response rate is inconsistent with a wide degree of mistrust of the NCVS.

(3) NCS gives legal guarantee

Which they may have every intent of honoring, but that won’t prevent someone from hacking into their database, or bribing someone in the hierarchy into providing names, addresses, etc. Besides, the degree to which one believes the guarantee is dependent upon the trust of any particular individual.

And neither can Kleck prevent someone from hacking into his database, or finding out phone number, name and address.

I’m surprised that people are still maintaining that Kleck’s survey was anonymous and that DG users were called back to verify that they had used a gun. If it was anonymous how could they call them back?