>How would one estimate the accuracy of self-reports of self-defense? I
>know in medical research you can assess the validity of self-reported
>health by doing follow up medical exams or seeing if the respondent dies
>or becomes seriously ill shortly after the survey. 

Well one thing one can do is ask if the survey data make sense in light of other 
sources of data we have. I'm _told_ that if you project the number of certain specific 
types of crimes that were supposedly prevented, according to the survey, that you get 
numbers that are many times larger than the actual number of those crimes committed. 
This doesn't seem plausible given that most people don't carry their guns with them 
when they are out on the street where the vast majority of crimes are committed.  
(Since I don't have a cite for this I'm not claiming its true, just suggesting it as a 

Another thing one can do is compare error rates on verifiable items of a comparable 
nature. For example a lot more people report that they are "managers" than actually 
are (as verified by their employers). Since a lot of people would probably consider it 
heroic to fight of a criminal with a gun I wouldn't be surprised if people engaged in 
a similar sort of wishful thinking on this question. This approach in particular 
suggests that _all_ the reports in the survey could easily be in error (which doesn't 
mean that no one ever uses a gun in self-defense, just that you would need a much 
bigger sample to find them and accurately calculate the true rate).

One could look at published reports of crimes and attempted crimes and look at the 
fraction of reported incidents in which victims were armed. Of course there is going 
to be reporting bias, but isn't this why the whole issue of "brandishing" vs 
"discharging" is important? We expect that if people have to discharge their weapons 
in self defense then we will read about it in the paper and we should be able to get 
an accurate estimate of how important gun use in self defense is from such sources. 
Suppose we never hear about cases where criminals are scared off by someone 
brandishing a gun but we always hear about it when a criminal is shot. My 
understanding is that reports of the latter are very rare. If they account for 25% of 
crimes prevented then there aren't many crimes prevented (4x reports), but if 98% of 
the time all one has to do is show the gun then the number of crimes prevented is 50x 
the number of reports and is considerably more important. Thus the difference between!
 75% and 98% is very very substantive. A difference between 98% and 90% would mean 
1/5th as many crimes prevented. 

>Is self-defense just one of those issues where we'll never have decent

Yes, but that doesn't meant that we can't learn from what data are available. My 
understanding is that depending on how you come at this issue you reach very different 
conclusions. If what I have been told by people I trust on these issues is true, there 
is very little evidence supporting the view that guns are frequently used in defense 
against criminals other than survey data and anecdote. - - Bill Dickens

William T. Dickens
The Brookings Institution
1775 Massachusetts Avenue, NW
Washington, DC 20036
Phone: (202) 797-6113
FAX:     (202) 797-6181
AOL IM: wtdickens

Reply via email to