The Harvard Injury Control Research Center recently published a report about this very thing. First they described how the survey questions might be presented.
Approach 1: ask everyone about gun use;
Approach 2: ask only those who first report that someone tried to commit a crime against them--and assume that a negative response to this screener question means that the respondent could not have experienced either a genuine self-defense or criminal gun use. For example, a preemptive strike would not be considered a self-defense gun use.
Approach 2, which obtains gun information from only a subset of those asked in Approach 1, yields much lower estimates of both types of gun uses (see McDowall et al. 2000 for the differences with respect to self-defense gun use).
The results were as follows:
Approach 1: 10 million (Box A criminal) 2.5 million (Box B defensive)
Approach 2: 800,000 (Box C criminal) 80,000 (Box D defensive)
One of the most fascinating things for me is that approach 1, maintains the famous 2.5 million DGUs but puts the criminal figure at four times that, 10 million.
No one likes Approach 2, which is the one used by the NCVS.
The Harvard researchers go on to point out another important factor, one which I find of particular interest. Some, they say most, of the 2.5 million supposed defensive uses of guns are nothing of the sort.
Regular citizens with guns, who are sometimes tired, angry, drunk, or afraid, and who are not trained in dispute resolution, have lots of opportunities for inappropriate gun uses. People engage in innumerable annoying and somewhat hostile interactions with each other in the course of a lifetime. It should not be surprising that inappropriate, socially undesirable "self-defense" gun uses by people who believe they are law-abiding citizens outnumber the appropriate and socially beneficial use of guns.
Although most of the reported self-defense gun uses from Approach 1 surveys seem more like criminal uses, even if one believed they were all genuine socially beneficial uses, the number of criminal gun uses would still vastly exceed the number of self-defense gun uses in the United States. No survey using similar methodology to determine both criminal and self-defense use has ever found otherwise.
What's your opinion?