The '91 GAO Report GAO/PEMD-91-9


"Many Deaths and Injuries Caused by Firearms
Could Be Prevented"


It would be a rare gun control advocacy organization that does not somewhere in their propaganda about protecting "the children" have a statement that some number or portion of accidental shooting deaths (and injuries) of children would be avoided if only we would require gun makers to include a "loaded" indicator and an automatically engaging child-proof safety.  These statements of "fact" are generally attributed to "the GAO" or refer to an unspecified '91 GAO report (or one with the above title) as the source of the "information."  Not wanting the public to actually see the report, or maybe just out of ignorance (common there), they generally do not give the number of the GAO report.

In all its reading of materials by people and organizations that speak out against gun control, has never come across anything questioning or challenging these gun controller claims or the General Accounting (now "Accountability") Office study/report.  In effect, anti-control advocates have just let the gun controllers have this issue by implicitly accepting the claims.  It took some time, but we finally got a copy of the report and analyzed it.

One of the reasons it has taken a while to get the report is that it is not widely known just what it takes to obtain a GAO document.  By visiting the GAO website eventually learned that anyone can obtain one free copy of any GAO document (except some for which access is restricted) simply by calling the GAO at (202)512-6000 and asking for it.  The GAO can look up the document in various ways other than the document number, including title and key words, but you should have as much information as possible, like the year of the report, to help the customer service rep if necessary.  And don't be surprised if it takes 60 days to actually get what you order in spite of customer rep claims to the contrary.

Are the gun controller claims valid?  Do they properly represent what the report says?  Was the study addressed by the report valid?  Is the report valid in how it represents the results of the study?

No, the gun controller claims are not valid, mostly because they do not honestly represent what the report says—possibly because the report itself does not honestly and correctly represent what GAO found.  The remainder of this document addresses what was done in the GAO study, what the results of the study were, what the GAO said the results were, the difference between the actual and claimed results, the actual significance of the study results, and some comments about the study and report.  You don't have to take our word about the study.  Order a copy of the report and check the following explanation against the report itself.


For their study the GAO randomly selected some large jurisdictions and some small jurisdictions, then obtained and examined case records on accidental gunshot deaths during 1988 and 1989 in those jurisdictions.  They examined 107 such cases.  They used specific criteria to determine which of those deaths would have been prevented by each of two firearm design features.  They also did a separate, nonrandom analysis to estimate how many people suffer accidental gunshot injuries in comparison to the number that die from accidental gunshots.  This analysis was not random because it relied upon searching for police jurisdictions that kept suitable records of gunshot injuries.

Several details about the method used in the accidental death analysis are important to the question of validity.  First, the two design features for which the analysis was done were:

  1. a "child proof" device that would automatically engage on the firearm and prevent it from firing under any circumstances until disengaged in a manner that could not be done by any child under 6 years old; and,
  2. a "loaded" indicator that would be visible to and understood by any idiot that might happen to pick up the firearm.

Note that the GAO analysts had no clue as to how many murders might be facilitated by victims fumbling with the "child proof" safety.  Note too that no "loaded" indicator exists that meets the criteria of the study, and that probably no safety exists that meets the criteria for the "child proof" safety.  They did an analysis about hypothetical, perfect design features.

The GAO analysts/authors did not represent their criteria as expressed above.  They represented the criterion for the "child proof" safety as being about the child being able to pull the trigger, but they actually tallied as "preventable" with such a device all deaths in which a child under 6 had the gun, regardless of whether or not the gun fired as a result of the child pulling the trigger.  They represented the criterion for the "loaded" indicator as being just about whether or not the shooter claimed to have thought the gun was unloaded.  It was only near the end of their report that they acknowledged that the validity of their numbers depended upon anyone who would ever pick up any gun recognizing and understanding the significance of the indicator that existed on the particular firearm.

The analysts considered that a death would have been prevented by a "child proof" safety if the shooter was less than 6 years old.  This applied to 9 of the deaths.  They considered that a death could not have been prevented by either of the two design features in 52 of the deaths, and sited examples of such deaths as including:

  1. mostly those deaths that resulted from the firearm being dropped or knocked down; plus,
  2. 10 deaths resulting from playing "Russian roulette."

Note that the report did not determine whether any of those children under 6 years old had fired the gun by pulling the trigger, or if any of them had fired the gun by dropping it or banging it on something—a circumstance that would have put the death into the "unpreventable" category rather than the "childproof" category.  That is, an unknown portion of the 9 death in which the gun was held by a child under 6 may have occurred regardless of whether or not the trigger could have been pulled by the child.  The basic problem of the GAO analysis here was that they did not use criteria that were mutually exclusive.

Note also that the 10 Russian roulette deaths should never have been included in a count of "accidental" gun deaths.  They are more appropriately called suicides—suicides that started out as attempted suicides with some chance of failure if the "game" had some limit to the number of tries.

The analysts tallied 25 of the deaths as being preventable by the firearms having a "loaded" indicator.


The GAO report claimed under its "findings" that 8% (all of children under 6 years old) of the "accidental" gun deaths they examined "could have been prevented by a child-proof safety device."  It claimed that 23% of the gun deaths could have been prevented had the firearms all had "loaded" indicators.  The first claim is clouded by the fact that the analysts did not consider the possibility of the safety device not being perfect or of a child firing a gun in some manner other than pulling the trigger.  The second claim is clouded by the fact that a gun simply having a "loaded" indicator is by no means a guarantee that a person holding the gun will observe the indicator or understand what it means.  It would be interesting to know how many of the 25 cases in which people "didn't know the gun was loaded" involved guns that did have "loaded" indicators.  Both of the claims are clouded by the fact that the 10 "Russian roulette" deaths should not have been even included in the cases being examined.

Based on the fact that the analysts found that 8% and 23% of the deaths they examined could have been avoided with the presence of the two design features, they projected in the Executive Summary and in the Chapter 2 "Estimates of Preventable Deaths" that these same percentages of all accidental firearm deaths could be prevented by having these design features in what they later acknowledged would have to be all firearms and only if the "loaded" indicator were understood by everyone who ever holds or possesses a firearm.  Specificallly, of the 1501 accidental firearm deaths in 1988, they estimated that 113 ± 64 could have been prevented by a [perfect] "child proof" safety device and 345 ± 99 could have been prevented by the firearms having [universally noted/understood] "loaded" indicators.

Other than the deficiencies earlier noted, the GAO analysts actually made a technical mistake that artificially reduced the predicted impact of the two hypothetical, magical and perfectly functioning design features.  Specifically, they projected the nationwide impact based on the percentages they had found of all the cases they examined.  But those 107 cases included 21 for which the case files had insufficient information for the analysts to make a determination of preventability.  Some of those cases very likely would have fallen into the "preventable" categories had the analysts had sufficient information.  So the GAO estimates are actually conservative estimates, even if they are estimates relating to nonexistent, ideal safety devices that are assumed to work perfectly.

In the part of the study that dealt with injuries as opposed to deaths, the analysts found that firearm injuries coming to the attention of the police are in the neighborhood of 100 times as common as firearm deaths in urban areas.  It should be kept in mind that a high proportion of firearm injuries, being minor enough not to impel the victim to seek medical attention, would not come to the attention of police and would therefore be over and above the 100 ratio observed in this study.


In their report the GAO analysts tended to say that they found that the two types of design features could (i.e., might) prevent firearm deaths.  They didn't actually say the features would.  However, most people reading the statements would not pay any attention to the distinction.  The estimates were actually estimates of maximum conceivable impact of using perfect design features.

The analysts also referred to the two types of design features as being things that actually exist and would accomplish their intended functions by definition.  It was only near the end of the report that the analysts acknowledged that people wouldn't necessarily pay any attention to or understand the significance of "loaded" indicators.  The authors were honest only in that eventually, in the bowels of the report, they came clean.

The study came about because it was requested by then Senator howard metzenbaum, a devout gun banner who once said that he didn't care about reducing crime but wanted only to get rid of guns (and implied that John Glenn had never had a job, having been only a Marine).  The senator requested the study by the GAO as chair of the Judiciary committee's Subcommittee on Antitrust, Monopolies, and Business Rights—which you should note has nothing to do with guns, health or safety.  In other words, he misused his position to commission a study about something he had no business getting involved in.

It should be no surprise then:  that the GAO, in its letter transmitting the report to metzenbaum, referred to accidental deaths and injuries from gunshots as being a "public health problem"; that the report recommended that the Consumer Product Safety Commission be given authority to regulate firearms; and that the report bibliography was 3+ pages of citations of articles, etc. predominantly by kellerman, wintemute, and other doctors who have been responsible for shamefully deficient and biased "research" about guns "as a public health problem."

RETURN TO weil-hemenway '92 "STORAGE" STUDY PAGE