I wonder why Dr. Lambert writes "So you haven't actually seen their paper?" on 4/20, when my message of 4/3 states "I haven't read the Ludwig & Cook paper ..."  Well, I don't really wonder -- it is painfully apparent.

One problem with using statistics by many not properly trained in both mathematics and the scientific method is the misperception that correlation is evidence of causality or that statistics can prove causality by itself.  You might suspect causation, but to prove it you need a controlled experiment (see:
for a discussion suitable for laymen). 
 
A common misuse of statistics is imputing significance to non-zero sample correlations (people get more colds when they wear warmer clothing, so warmer clothes cause colds).  My personal favorite from the 1950's and 1960's is the idea that skirt lengths are a predictor of the stock market direction (A Random Walk Down Wall Street, Burton G. Malkiel).
 
While I haven't read the Cook and Ludwig paper cited by Dr. Lambert, I have read the working paper "Does Gun Prevalence Affect Teen Gun Carrying After All?", Working Papers Series, SAN03-04, August 15, 2003, where Cook and Ludwig assert: 
"Of course there is a possibility of reverse causation, which would be true if teen gun carrying affected the household demand for gun ownership. To rule out the possibility of reverse causation we estimated our gun-carrying regressions using two-stage least squares, where the second-stage estimate is based on the variation in county-level gun prevalence in the 1990s that is explained by variation in the proportion of the state population that lived in rural areas in 1950. It turns out that “percent rural in 1950” is strongly predictive of county-level gun prevalence in 1995."

Here you can see how Cook and Ludwig deal with issues of reverse causation by appealing to “percent rural in 1950” as a parameter determining gun prevalence (by arguing “percent rural in 1950” determines gun prevalence, they are able to assert prevalence is independent of the crime and pre-determined by another parameter). 

Since Cook and Ludwig claim “percent rural in 1950” STRONGLY predicts gun prevalence and gun prevalence predicts burglary rates, believers in their work must conclude burglary rates in 1995 are predicted by proportions of residents who lived in rural counties in 1950.  You read it here first -- hicks cause burglaries.
 
The operative word is "believers" as in the faithful.  What we are dealing with is religion, not science. 
 
Phil Lee


-----Original Message-----
From: [EMAIL PROTECTED]
[
mailto:[EMAIL PROTECTED]]On Behalf Of Tim Lambert
Sent: Tuesday, April 20, 2004 12:53 PM
To: [EMAIL PROTECTED]
Subject: Re: Unintended (negative) consequences of UK gun control


"Phil Lee" <[EMAIL PROTECTED]> writes:

> "Tim Lambert" <[EMAIL PROTECTED]> writes:
>
>> "Phil Lee" <[EMAIL PROTECTED]> writes:
>>
>> > Lambert's case, as illustrated in his 3 points below, illustrates an
>> > approach he regularly argues which I'll call the "tobacco company
> defense."
>> > The argument, which has valid elements, is that correlation between
> events
>> > (smoking and cancer, hot burglary rates and lack of firearms) do not
> prove
>> > there is a causal relationship.
>>
>> You turn the truth on its head.  I have pointed out before
>> that there is no correlation between hot burglary rates and gun
>> ownership.  See Ludwig and Cook's study.
>
> Cook and Ludwig state "Rather, our analysis concludes that residential
> burglary rates tend to increase with community gun prevalence" -- that is,
> they assert a correlation between gun prevalence and burglary rates.  It did
> seem that by citing that paper Lecturer Lambert was arguing community gun
> prevalence was correlated to residential hot burglary rates.

No, by citing the paper in which Cook and Ludwig wrote:

   The results from a cross-sectional analysis of NCVS data from
   1994-1998 are presented in Table 10. The first column of estimates
   are of the effect of gun prevalence on the likelihood that someone
   is at home when there is a burglary (22 percent on average). While
   the coefficient estimates are negative in every case (suggesting a
   deterrent effect), they are not discernibly different from zero,
   and in all but the simple regression have t-statistics less than
   one. While null findings may sometimes simply reflect a lack of
   statistical power, in this case the NCVS data support a quite
   precise estimate, with the standard errors all less than one-tenth
   of one percentage point.

I was referring to their result that the hot burglary rate (the
proportion of burglaries that are hot) was not significantly
correlated with gun prevalence.

> 3. None of these measures account for "chicken and egg" issues.  I've not
> seen analyses determine whether gun prevalence being high is due to high
> burglary rates rather than the other way around.

So you haven't actually seen their paper?

    The new results reported here suggest that if there is such a
    deterrent effect, it may well be swamped by other factors
    associated with gun prevalence most likely, it seems to us, the
    fact that guns are particularly attractive loot. Cross-section
    analysis of the NCVS and panel-data analysis of the UCR yield
    quite similar findings: a 10 percent increase in our measure of
    gun ownership increases burglary rates by 3 to 7 percent. These
    results are not likely to be due to reverse causation: Among other
    evidence on this matter is the findings from our
    instrumentalvariable estimates, which are consistent with the OLS
    results. Most important, we find that gun prevalence has little
    effect on the fraction of residential burglaries in which someone
    is at home, and that the hot-burglary victimization rate tends to
    increase with gun prevalence. These results are robust to
    alternative specifications and data sets.


>>
>> Again you turn the truth on its head.  The evidence tells us what they
>> think others believe.  It is speculation to argue that they believe
>> the same thing themselves.  You can speculate if you want but you
>> should accurately describe what Wright and Rossi's survey found and
>> not make false claims about their findings.  They DID NOT conclude
>> that the higher hot burglary rate in Britain was caused by American
>> burglars fear of being shot.
>
> Lecturer Lambert is certainly confused about who turns truth on its head
> with Wright and Rossi's results.  Wright and Rossi found that gun control
> laws have relatively little effect on violent criminals;

No. They didn't even study the effects of gun control laws.

> they found more than half of the convicts would not attack a
> potential victim known to be armed;

Nope. Again, this was a statement about what other criminals would
do.

> more than half feared encounters with armed citizens more than
> encounters with police;

Nope. This was another statement about what other criminals felt.

> and that approximately one-third of them had been
> scared off or defeated by an armed citizen.

Closer.  That should be "armed victim".

And Wright and Rossi DID NOT conclude that the higher hot burglary
rate in Britain was caused by American burglars fear of being shot.
Please try to accurately report their results.

>> (BTW, how come you give Olson a title and not me?)
>>
> I hope that by finding your title and using it, you are mollified at my
> previous gaucherie.  You can continue to use Lee with no title.  Although I
> have a PhD in Mathematics, I'm happy with the reputation that attaches to my
> name and it need no inflation.

I did not ask you to give me a title.  I asked: "how come you give
Olson a title and not me?"  You don't seem to have answered my
question.  Not giving me a title is not gauche, but giving me an
incorrect title over and over again, is.  I don't really care if
whether you give me a title or not, but if you do, "Dr" or "Mr" is
correct.

--
Tim
_______________________________________________
To post, send message to [EMAIL PROTECTED]
To subscribe, unsubscribe, change options, or get password, see
http://lists.ucla.edu/cgi-bin/mailman/listinfo/firearmsregprof
_______________________________________________
To post, send message to [EMAIL PROTECTED]
To subscribe, unsubscribe, change options, or get password, see 
http://lists.ucla.edu/cgi-bin/mailman/listinfo/firearmsregprof

Reply via email to