Self-protection
Between 1987 and 1990, David McDowall found that guns were used in defense during a crime incident 64,615 times annually. This equates to two times out of 1,000 incidents (0.2%

that occurred in this time frame. For violent crimes (assault, robbery, and rape), guns were used 0.83% of the time in self-defense. Of the times that guns were used in self-defense, 71% of the crimes were committed by strangers, with the rest of the incidents evenly divided between offenders that were acquaintances or persons well-known to the victim. Of all incidents where a gun was used for self-defense, victims shot at the offender 28% of the time. In 20% of the self-defense incidents, the guns were used by police officers. During this same time period, 1987 and 1990, there were 46,319 gun homicides, and the National Crime Victimization Survey estimates that 2,628,532 nonfatal crimes involving guns occurred.
The findings of the McDowall's study for the American Journal of Public Health
contrast with the findings of a 1993 study by Gary Kleck, who finds that as many as 2.45 million crimes are thwarted each year in the United States, and in most cases, the potential victim never fires a shot in these cases where firearms are used constructively for self-protection. The results of the
Kleck studies have been cited many times in scholarly and popular media.
McDowall cites methodological issues with the Kleck studies, claiming that Kleck used a very small sample size and did not confine self-defense to attempted victimizations where physical attacks had already commenced. The former criticism, however, is inaccurate — Kleck's survey with Marc Gertz in fact used the largest sample size of any survey that ever asked respondents about defensive gun use — 4,977 cases, far more than is typical in national surveys.
A study of gun use in the 1990s, by David Hemenway at the Harvard Injury Control Research Center, found that criminal use of guns is far more common than self-defense use of guns. By the Kleck study, however, most successful preventions of victimizations are accomplished without a shot being fired, which are not counted as a self-defense firearm usage by either the Hemenway or McDowall studies.
Hemenway, however, also argues that the Kleck figure is inconsistent with other known statistics for crime, citing that Kleck's figures apparently show that guns are many times more often used for self-defense in burglaries, than there are incidents of bulgaries of properties containing gun owners with awake occupants. Hemenway concludes that under reasonable assumptions of random errors in sampling, because of the rarity of the event, the 2.5 million figure should be considered only as the top end of a 0-2.5 million confidence interval, suggesting a highly unreliable result that is likely a great overestimate, with the true figure 10 times, or more, less.