ArticlePDF Available

What Do CDC’s Surveys Say About the Prevalence of Defensive Gun Use?

Authors:

Abstract and Figures

In 1996, 1997, and 1998, the Centers for Disease Control and Prevention (CDC) conducted large-scale surveys asking about defensive gun use (DGU) in four to seven states. Analysis of the raw data allows the estimation of the prevalence of DGU for those areas. Data pertaining to the same sets of states from the 1993 National Self-Defense Survey (Kleck & Gertz, 1995) allow these results to be extrapolated to the U.S. as a whole. Possible sources of error in surveys of DGU are reviewed, and the results of previous surveys compared. CDC’s survey data confirm previous high estimates of DGU prevalence, disconfirm very low estimates derived from the National Crime Victimization Survey, and indicate that defensive uses of guns by crime victims are far more common than offensive uses by criminal offenders.
Content may be subject to copyright.
What Do CDCs Surveys Say About the Prevalence
of Defensive Gun Use?
Gary Kleck
1
Received: 1 June 2020 / Accepted: 12 August 2020/
#Southern Criminal Justice Association 2020
Abstract
In 1996, 1997, and 1998, the Centers for Disease Control and Prevention (CDC)
conducted large-scale surveys asking about defensive gun use (DGU) in four to seven
states. Analysis of the raw data allows the estimation of the prevalence of DGU for
those areas. Data pertaining to the same sets of states from the 1993 National Self-
Defense Survey (Kleck & Gertz, 1995) allow these results to be extrapolated to the U.S.
as a whole. Possible sources of error in surveys of DGU are reviewed, and the results of
previous surveys compared. CDCs survey data confirm previous high estimates of
DGU prevalence, disconfirm very low estimates derived from the National Crime
Victimization Survey, and indicate that defensive uses of guns by crime victims are
far more common than offensive uses by criminal offenders.
Keywords Defensive gun use .Self-protection .Firearms .Gun control .Surveys
Introductionx
The debate over gun control in the U. S. heavily revolves around the issue of the costs
and benefits of widespread gun ownership. The principle costs are death, injuries, and
property loss due to criminal, suicidal, or accidental uses of firearms. The benefits
include recreation-related uses of guns, but the most serious and consequential benefits
are arguably deaths, injuries, and property loss prevented by defensive use of guns, or
criminal attempts deterred by the potential for such use. The magnitude of these
benefits are partly a function of how often guns are used for self-protection. Thus,
the frequency of defensive gun use (DGU) is an important part of the American gun
control debate.
https://doi.org/10.1007/s12103-020-09562-0
*Gary Kleck
gkleck@fsu.edu
1
College of Criminology and Criminal Justice, Florida State University, Tallahassee, FL 32306
1273, USA
Published online: 10 September 2020
American Journal of Criminal Justice (2021) 46:401–421
At least 21 national surveys have asked large probability samples of the U.S. adult
population whether they had used guns defensively, including 17 privately sponsored
surveys before 2000, and at least four more since 2000 (Kleck, 2001b; Roper Center,
2018). As will be discussed later, the private surveys have generally yielded annual
estimates of the number of DGUs by adults against other persons in the one-to-three
million range. The estimates vary considerably because the surveys addressed different
subsets of the universe of DGUs or pertained to different time periods with differing
crime rates (Kleck, 2001b).
Gun control advocates assert that DGUs are rare, and cite estimates of DGU
frequency derived from the National Crime Victimization Survey (NCVS). This survey
has yielded annual estimates of just 64,615 DGUs (McDowall & Wiersema, 1994),
only about 3% of the typical result of 21 other surveys. The extremely low NCVS
estimates are uncritically cited by advocates of stricter gun control, so it is worth
considering why the NCVS generates such deviant results. First, the NCVS is a
nonanonymous survey - the identities of respondents (Rs) are known to researchers.
Second, the survey is conducted by an agency of the federal government, the U. S.
Bureau of the Census. Third, respondents (Rs) are told that the information generated
by the survey will be provided to the U.S. Department of Justice the law enforcement
branch of the federal government. Fourth, Rs are asked about their self-protective
actions only after they have stated where the incident occurred. In most states, for all
but the few people who have carry permits, it is illegal to possess a firearm off their
own property, and thus most Rs could not report a DGU carried out in a public place
without confessing to the crime of unlawful carrying. Under these circumstances, there
is a sound basis for doubting whether Rs would be willing to report incidents in which
they had pointed a gun, or even shot at, another human being, regardless of the
justification. Fifth, the Rs are only asked an open-ended nonspecific question about
what they might have done to protect themselves during a crime incident, but no Rs in
the NCVS have ever been specifically asked about defensive use of a gun. Victim use
of a gun can only be reported in the NCVS if the R chooses to volunteer that specific,
controversial detail.
The Centers for Disease Control and Prevention (CDC) has also conducted surveys
in which large probability samples of the adult population were asked about DGU, as
part of their Behavioral Risk Factor Surveillance System (BRFSS). The DGU questions
were asked of representative samples of adults in various sets of four to seven states as
part of an optional module of firearms-related questions. To my knowledge, CDC never
reported the results of those surveys, and does not currently report on their website any
estimates of DGU frequency.
I only discovered that CDC had ever asked about DGU in their BRFSS surveys
while searching through the questionnaires used in the surveys for questions on other
topics. Once I found the key question in the questionnaire for 1 yearsBRFSS,I
searched the questionnaires for all the other years, from 1984 through 2016, and found
the DGU question had been asked in the 1996, 1997, and 1998 surveys. It was included
as part of Optional Module 18 asking a variety of questions about firearms. Individual
states could include these questions in their BRFSS surveys if they wanted to do so, in
addition to the standard set of questions asked in all states. Seven states chose to do so
in the 1996 survey, seven in the 1997 survey, and four in the 1998 survey (see Table 1
for lists of participating states).
402 American Journal of Criminal Justice (2021) 46:401–421
The timing of CDCs addition of a DGU question to the BRFSS is of some interest.
Kleck and Gertz (1995) conducted their National Self-Defense Survey in February
through April 1993, and estimated that there were about 2.5 million annual DGUs.
They privately circulated their estimates of DGU in 1993 and 1994, formally presented
them at the annual meetings of the American Society of Criminology in November of
1994, and published the results in the Journal of Criminal Law and Criminology in the
Fall of 1995 (Kleck & Gertz, 1995).
CDC added a question about one specific subtype of DGU to their 1994 Injury
Control and Risk Survey (ICARIS), a national telephone survey of U.S. adults fielded
between April 28 and September 18, 1994, just months after preliminary results of the
Spring 1993 Kleck/Gertz survey were circulated. ICARIS interviewers asked Rs in
gun-owning households about incidents in which they retrieved a gun because they
thought an intruder was in, or trying to get into their home (Ikeda, Dahlberg, Sacks,
Mercy, & Powell, 1997), actions which roughly correspond to burglary-related DGUs.
CDC then added a DGU question to the BRFSS the first year they could do so after the
1995 publication of the Kleck/Gertz survey, in their 1996 survey.
CDC was not the only federal agency during the Clinton administration to field a
survey addressing the prevalence of DGU shortly after the results of the Kleck/Gertz
survey became known. The National Institute of Justice (NIJ) financed a national
survey devoting even more detailed attention to estimating DGU prevalence. It was
fielded in November and December 1994 (Cook & Ludwig, 1996). Then, NIJ and
CDC helped co-finance a national survey conducted by David Hemenway and Deborah
Azrael, which was fielded in the Spring of 1996 (Hemenway & Azrael, 2000, pp. 259,
272). Finally, CDC funded a second national survey by Hemenway that purported to
correct some of the problems in his 1996 survey. This one was fielded in the Spring of
1999 (Hemenway, Azrael, & Miller, 2000, pp. 263, 267). In sum, although neither
CDC nor NIJ had ever financed a single survey asking about DGU before 1994, but then
they supported at least seven of them in 1994 through 1999, immediately following the
Spring 1993 survey by Kleck and Gertz.
Surveys by CDC are of special interest because the agency has often been criticized
by gun owner organizations like the National Rifle Association (NRA) as being
antigunand for awarding research grants on firearms and violence only to researchers
with strong anti-gun or pro-gun control publication records (see remarks of the NRAs
chief lobbyist, Cox 2017). Belief in this anti-gun bias was so strong among pro-gun
forces that the NRA got Congress to slash CDCs budget by an amount exactly equal to
the budget for its program that studied firearms violence, and to insert a rider in the
funding bill that read: Provided further that none of the funds made available for injury
Table 1 Groups of states that asked DGU question in BRFSS
Survey
Group Year States Included
A 1996 AK, KY, LA, MD, NH, NY, WV
B 1997 CO, HA, MS, NH, NJ, ND, OH
C 1998 LA, MT, NJ, PA
403American Journal of Criminal Justice (2021) 46:401–421
prevention and control at the Centers for Disease Control and Prevention may be used
to advocate or promote gun control(Jamieson, 2013). Of particular relevance to the
present topic, CDC has helped finance surveys on defensive gun use (DGU) by David
Hemenway and others that their authors interpreted as indicating that DGU was
actually quite rare (Hemenway & Azrael, 2000, p. 272; Hemenway et al., 2000,p.267).
To the extent that the NRAs complaints have been widely circulated among gun
owners, the CDCsreputationasanti-guncould discourage Rs in surveys sponsored
by CDC from reporting firearms-related behaviors, such as the ownership, carrying, or
defensive use of guns. There is direct evidence of this concerning estimates of the
prevalence of gun ownership. A national poll by Gallup in 1993 found that 49% of
households reported gun ownership, and a 1994 Los Angeles Times national poll found
the share to be 45% (Roper Center, 2018), but only 34% of households reported gun
ownership in CDCs 1994 Injury Control and Risk Survey (ICARIS)(Ikeda et al., 1997,
p. 366). If Rs are as reluctant to report DGUs to CDC interviewers as they appear to be
regarding gun ownership, CDC surveys may significantly underestimate the prevalence
of DGUs. Their estimates may for this reason be regarded as conservative.
CDCs 1994 ICARIS included a question on DGU but only in connection with what
CDC personnel called intruder-related firearm retrievals.Researchers asked those
who reported any guns in their household: During the past 12 months, how many
times did you or any other household member get a firearm because there might be an
intruder in or trying to get into your home?The researchers then established whether
those retrieving a gun actually saw an intruder and believed the intruder was fright-
ened away because of the gun,which presumably implies that the intruder saw the gun
and was threatened with it. Of the 34% of Rs reporting household gun ownership, 6%
contained at least one person who, in the previous 12 months, retrieved a gun, saw an
intruder, and believed the intruder had been scared away because of the gun
(Ikedaet al., 1997). The researchers estimated that there were 497,646 incidents in that
year in which an intruder was reportedly scared away by a gun. Kleck and Gertz (1995)
had found that 33.8% of DGUs were linked with burglaries, implying that the total
number of DGUs is 2.96 times the number of burglary-linked DGUs (1/0.338 = 2.96).
If CDCsintruder-related firearm retrievalsare interpreted as burglary-linked DGUs,
these survey results indicated that there were about 1.5 million total DGUs in 1994
among persons reporting a gun in their household ( 2.96 × 497,646 = 1.5 million).
It is, however, unclear how many of these experiences constitute DGUs, since it was
not explicitly established that the gun user attacked or threatened the intruder. Conse-
quently, it is debatable whether CDC generated any DGU estimates with this survey. In
contrast, they clearly did ask about DGU in their 1996, 1997, and 1998 BRFSS surveys.
CDCs Behavioral Risk Factor Surveillance System Surveys
The BRFSS surveys are high-quality telephone surveys of very large probability
samples of U.S. adults, asking about a wide range of health-related topics. Even just
the subset of four to seven state surveys that asked about DGU in 19961998
interviewed 31974500 adults, depending on the year. This is more people than were
asked about this topic in any other surveys, other than the National Self-Defense
Survey conducted in 1993 by Kleck and Gertz (1995), who asked DGU questions of
404 American Journal of Criminal Justice (2021) 46:401–421
4977 people. Sample sizes were much smaller in all the rest of surveys on the topic
(Kleck, 2001b).
The wording of the DGU question in the BRFSS surveys was also excellent,
avoiding many problems with the wording that afflicted the DGU questions used in
other surveys. The exact wording was:
During the last 12 months, have you confronted another person with a firearm,
even if you did not fire it, to protect yourself, your property, or someone else?
Respondents (Rs) in these surveys had previously been instructed not to report firearm
uses associated with an occupation that requires and authorizes you to use a firearm.
Thus, the question excluded uses by military, police, security guards, and others with
firearm-related jobs. Further, the question appropriately excluded uses against animals
(“…another person…”), asked about a specific, recent recall period (“…during the last
12 months…”), covered uses by any type of firearm (not just handguns), covered uses
regardless of where they occurred (not just uses in the home), and explicitly told
respondents that they should report uses even if they did not fire a gun. In sum, the
surveys used an excellent, carefully worded DGU question, in contrast to the wordings
used in so many other surveys (reviewed in Kleck, 2001b).
The most important shortcomings of the BRFSS surveys regarding DGUs were that
(1) the DGU question was asked only in four to seven states, and (2) the DGU question
was asked only of Rs who had reported guns in their household at the time of the
survey. In the present analysis, procedures have been developed to adjust for these
limitations.
Results - What Did CDCs Surveys Indicate About the Prevalence
of Defensive Gun Use?
I downloaded the BRFSS datasets for 1996, 1997, and 1998 from the BRFSS website
(CDC, 2018a) and obtained frequencies on the DGU question. Three different combi-
nations of states asked the DGU question in the BRFSS surveys in the 3 years it was
asked. Table 1displays which states asked the DGU question in each of the 3 years.
Alaska asked the DGU question in 1996 and Hawaii asked it in 1997, but these results
could not be used because the NSDS did not cover those states, and findings from the
NSDS were needed to adjust the BRFSS results.
The weighted and unweighted frequencies used to compute the prevalence of DGU
in the three BRFSS surveys and in the NSDS are shown in Table 2. Readers should
note that figures derived from the BRFSS surveys cannot be compared across years
since they were based on three different subsets of states. Thus, they reflect the
prevalence of gun use in differing subsets of the nation. Some of the states are places
where gun ownership is higher than average and one would therefore expect more
DGUs, while other states are lower in gun ownership and thus presumably lower in
DGU. These year-to-year differences should not be interpreted as either inconsistencies
or as reflecting real changes over time in the prevalence of DGU.
It is the weighted frequencies in Table 2that are meaningful because they adjust for
differences in the probability of a given person being selected into the sample. Weighted
405American Journal of Criminal Justice (2021) 46:401–421
frequencies from the BRFSS are based on data weighted by the FINALWT weight (see
CDC, 2018a for a thorough explanation of the weighting). Weighted frequencies from
the NSDS are based weights documented in an unpublished 1995 research memoran-
dum by Kleck (Weights used in the National Self-Defense Survey datasets.College of
Criminology and Criminal Justice, Florida State University.) Nevertheless, unweighted
frequencies are shown because some readers may be skeptical of weighting procedures
and worry that they could somehow distort DGU estimates.
The unweighted frequencies are, however, relevant to the question of how much the
BRFSS results are subject to sampling error. As previously noted, the sample sizes used
in the BRFSS surveys, even just for the subsets of states that asked the DGU question,
are enormous, ranging from 3197 to 4500. Such large sample sizes minimize the degree
to which DGU estimates could be influenced by random sampling error. For illustrative
purposes, if we use the unweighted frequencies from the 1996 survey (46 DGUs among
4500 total Rs) and assume random sampling, the 95% confidence interval estimate of
the fraction of adults who had a past-year DGU for the 1996 survey would be 1.022%
± 0.294%, or 0.731.32%. Note that the more meaningful weighted DGU prevalence
was 1.329% in the 1996 survey, but readers are again cautioned that this pertains only
to the six states that asked the DGU question and therefore cannot, without further
adjustment, be directly compared with estimates for the U. S. as a whole.
Table 2also includes weighted and unweighted frequencies from the NSDS con-
ducted in the Spring of 1993, for both the continental U. S. as a whole, and for each of
the three subsets of states used in the three BRFSS surveys. These figures allow us to
estimate how much more common or less common DGU was in the nation as a whole
compared to how common it was in each of these subsets of states. Knowing these
relative levels in turn allows us to extrapolate the BRFSS estimates of DGU prevalence,
based on the subsets of states, up to the nation as a whole.
Table 2 Numbers used to compute relative prevalence of DGU in U.S. as a whole vs. groups of states
BRFSS 1996 1997 1998
Unweighted past-year DGU cases 55 29 33
Unweighted total cases 5484 4189 3197
Weighted (by FINALWT) DGU cases 69,387 51,976 61,360
Weighted total cases 5,336,378 5,835,973 5,869,842
Weighted % with DGU 1.300 0.891 1.045
NSDS
Weighted (by fwt) past-year DGU cases, U. S. 66 66 66
Weighted total sample cases, U. S. 4969 4969 4969
Weighted % with DGU, U. S. 1.326 1.326 1.326
Weighted past-year DGU cases, BRFSS states 10 8 4
Weighted total cases, BRFSS states 337 535 513
Weighted % with DGU, BRFSS states 2.967 1.495 0.780
Ratio, U. S. % DGU over BRFSS states % DGU 0.447 0.887 1.700
Abbreviations: BRFSS = Behavioral Risk Factor Surveillance System (CDC surveys); DGU = Defensive gun
use; NSDS = National Self-Defense Survey (fielded in 1993 by Kleck and Gertz)
406 American Journal of Criminal Justice (2021) 46:401–421
The NSDS-based estimated ratio of nationwide DGU prevalence over DGU prev-
alence in the subset of states that asked the DGU question in a given BRFSS survey is
shown in the last row of Table 2.
The numbers needed to estimate the number of U.S. adults who used a gun
defensively in each of the years from 1996 to 1998 are shown in Table 3. The estimated
prevalence of DGU for these subsets of states, based on the BRFSS, is reported in line
(4) of this table. To produce national estimates of the prevalence of DGU from the
BRFSS surveys, however, requires adjusting these prevalence estimates for the effects
of two limitations of those surveys. First, they only cover the populations of four to six
states (excluding Alaska and Hawaii), not the entire nation. Second, the DGU questions
were only asked of people reporting current household gun ownership. Thus, our
strategy was to project BRFSS results pertaining only to some states up to the nation
as a whole, and to then adjust for the omission of DGU uses by persons who did not
report household gun ownership at the time they were interviewed.
The process of adjusting BRFSS-based DGU estimates is summarized in Table 3.
We began by estimating how many adults live in gun-owning households, because
this is the population that was asked a DGU question in the BRFSS surveys. Row
(1) displays Census Bureau estimates of the size of the U.S. adult (age 18+)
residential population for years from 1996 through 1998 (U. S. Bureau of the
Census, 1998), while Row (2) shows the percent of U. S. households reporting
gun ownership in national surveys. Multiplying row (1) (divided by 100) times row
(2) yields the estimated number of U. S. adults living in gun-owning households,
Table 3 Estimating the frequency of defensive gun use based on CDCs BRFSS surveysa
1996 1997 1998
(1) U.S. resident population age 18+ (in 1000 s) 196.044 198.156 200.296
(2) % U. S. households reporting guns (Gallup Poll) 42 42 42b
(3) Number of adults in gun-owning households [(1)× (2)/100] 82,338,480 83,225,520 84,124,320
(4) BRFSS: % of adults in gun-owning households with DGU in
group of states asking DGU question (Table 2)
1.300 0.891 1.045
(5) NSDS: Ratio of U. S. DGU rate over rate in BRFSS states
(from last row of Table 2)
0.447 0.887 1.700
(6) BRFSS: Estimated % of all U. S. adults in gun-owning households
with DGU [row (4) times row (5)]
0.581 0.790 1.776
(7) BRFSS: Estimated number of all U. S. adults in gun-owning
households with DGU [row (6) times row (3)]
478,387 657,482 1,494,048
(8) BRFSS: Total U. S. DGUs, all adults [1.266 times row (7)] 605,638 832,372 1,891,465
Annual average of estimates in row (8) = 1,109,825
Notes:
a. DGU prevalence rate is defined as the estimated percent of U.S. adults (age 18+) who used a gun for self-
protection against a person, not including uses in connection with military, police, or security guard duties, in
the 12 months before the date of the survey interview. BRFSS =Behavioral Risk Factor Surveillance System,
NSDS = National Self-Defense Survey (Kleck & Gertz, 1995)
b. Gallup estimate of household gun ownership prevalence was 42% in both 1996 and 1997. Gallup did not
ask the question in 1998, so their 1997 figure was used
407American Journal of Criminal Justice (2021) 46:401–421
shown in row (3). (It was assumed that average household size is the same in
households with or without guns).
Row (4) reports the weighted percent of adults in gun-owning households who
reported a DGU in the various sets of states that asked a DGU question in the 1996,
1997, and 1998 BRFSS surveys (listed in Table 1). The raw frequencies used to
compute these percentages were reported in Table 2. We cannot directly apply these
estimates to the U.S. because the population of these sets of states do not constitute a
probability sample of the U. S. population. The prevalence of DGU could be far higher
in some states than in the nation as a whole if the states have higher-than-average rates
of gun ownership or crime, or could be far lower if the set of states had lower gun
ownership or crime rates. The substantial differences in the figures in row (4) are
certainly consistent with this view. The NSDS, however, provided information that
allows us to extrapolate the BRFSS figures applying to the various subsets of states to
the U.S. as a whole, since the NSDS provides both an estimate of DGU prevalence in
the nation as a whole and its prevalence in any given subset of states.
This NSDS-based ratio of the DGU prevalence for the U. S. as a whole over the
prevalence in the subset of surveyed states (seen in the last row of Table 2)isagain
reported, for convenience, in row (5) of Table 3. This adjustment ratio (row 5) is
multiplied times the figures in row (4) to yield BRFSS-based estimates of the percent of
adults in gun-owning households with a DGU in the U. S. as a whole, shown in row (6)
of Table 3.
The BRFSS asked the DGU question only of people living in households that
reported guns at the time of the interview. This excludes (1) DGUs by people who
used a household gun that was no longer in the household by the time they were
interviewed in the BRFSS, (2) DGUs by people who used a gun belonging to a person
who was not a member of their household, and (3) DGUs by people who falsely denied
having a gun in their household. This is not a trivial matter, since Kleck and Gertz
(1995, p. 187) found that 21% of persons who reported a DGU had denied having a
gun in their household at the time of the interview. To adjust for this difference, the
DGU prevalence estimates based on BRFSS surveys were multiplied by 1.266
(1/0.79 = 1.266). The figures shown in row (7) were therefore multiplied by 1.266,
yielding the final BRFSS-based DGU estimates shown in row (8).
These three CDC-based estimates average 1,109,825 DGUs per year for the period
19961998. This puts the CDC results squarely within the range of DGU estimates
typically produced by the many private surveys (Kleck, 2001b). This CDC-based
estimate, however, is 18 times larger than the number of DGUs supposedly implied
by the NCVS (McDowall & Wiersema, 1994). Thus, even other federal government
surveys, and not just private surveys, suggest that the NCVS-based estimate of DGU
prevalence is far too low.
What little difference there is between the BRFSS estimates for 19961998 and
those of earlier surveys could be due to declining rates of violent crime, the crime type
that accounts for most DGUs. For example, the U.S. murder rate was 9.3 per 100,000
population in 1992, but just 6.3 in 1998 (Federal Bureau of Investigation, 1999,p.64).
Thus, the murder rate declined by 32% from 1992, the period covered by the Kleck and
Gertz past-year DGU estimates, to 1998, the year when the last of the three BRFSS
surveys with a DGU question was fielded. With fewer occasions for self-defense in the
form of violent victimizations, one would expect fewer DGUs. Other things being
408 American Journal of Criminal Justice (2021) 46:401–421
equal, one would expect that a 32% decline in violence rates would result in a roughly
similar proportional decline in DGUs. Thus, if there were 2.5 million DGUs in 1992,
we would expect about 1.7 million in 1998. The CDC-based estimate for 1998 of
1.9 million is therefore quite consistent with the NSDS estimate for 1992.
A Review of Errors in Surveys of Defensive Gun Use
There is currently no feasible way to measure the prevalence of DGU other than with
surveys. Certainly police data cannot provide meaningful estimates given the unwill-
ingness of most crime victims to even report their victimizations to the police (Hart &
Rennison, 2003), never mind the controversial fact that they had threatened or attacked
another person with a firearm during the crime event. Counts of news accounts of
DGUs would be even more incomplete because news outlets would normally know
only about the subset of DGUs known to the police.
All surveys are flawed, some more than others. The BRFSS surveys, however, are
among the better ones, using large probability samples, carefully crafted question
wordings, and skilled interviewers. Is it nevertheless possible that even the BRFSS
yields DGU estimates that are too high? This could happen either because BRFSS
samples include too many people who had a DGU experience (sample bias) or because
too many people reported DGUs who did not actually have such experiences (response
errors). CDCs assessment of sample biases, however, indicate that they contribute to
an underrepresentation of persons likely to have a DGU. Their samples underrepresent
males, nonwhites, and low-income people that is, people who are more likely to
become crime victims and thus have occasions to use firearms for self-protection(CDC,
1998). These sample biases would all tend to make BRFSS-based estimates of DGU
prevalence too low.
Surveys can also be subject to coverage bias, that is the use of sampling procedures
that preclude the inclusion of some parts of the target population. For example, both the
CDC surveys and almost all other surveys asking a DGU question are telephone
surveys, so their sample frame is the set of all possible telephone numbers with known
area codes and prefixes. People without telephone service therefore have zero chance of
inclusion in the sample. Likewise, all these surveys interview only adults age 18 or
over, so no adolescents are interviewed. Surveys that ask about DGUs involving any
member of the household might indirectly capture some adolescent DGUs known to
adult household members, but reports by proxies are never as complete as reports
directly provided by the target individuals. Finally, national surveys almost always
exclude the homeless, and usually exclude persons in institutions at the time of the
survey (Dillman, Smith, & Christian, 2009). All of these excluded subpopulations are
of lower average income, and at greater risk of being a victim of the types of crime
which guns are used to defend against.
Thus, both sample bias and coverage bias tend to make estimates of DGU preva-
lence too low. Any upward bias in the estimate would therefore have to come from
response errors respondents giving, intentionally or unintentionally, inaccurate an-
swers to the DGU question. Critics of higher DGU estimates like David Hemenway
(1997) have speculated about reasons why respondents in these surveys might give
inaccurate answers, but these discussions are misleading because they focus solely on
409American Journal of Criminal Justice (2021) 46:401–421
false positiveresponses inaccurate claims of DGU experiences by persons who had
no such experiences. Thus, they address only response errors that tend to make DGU
estimates too high, while ignoring well-established sources of response errors that
would tend to make estimates of controversial behaviors too low. No one disputes that
false positive responses occur, but false positive responses cannot lead to an overesti-
mate of DGU prevalence unless they outnumber false negative responses people
saying noto the DGU question when the accurate answer would have been yes.
Hemenway and the other critics of higher DGU estimates (e.g. McDowall, Loftin, &
Presser, 2000) have had nothing to say about the frequency of false negative responses,
and thus can have nothing to say about the relative balance of these two kinds of
response error.
Unfortunately, there is little hard evidence bearing directly on response errors in
reporting DGUs in particular since it is ordinarily impossible to know for certain
whether a DGU really happened. In two national surveys, interviewers were asked
for their impressions as to whether each R they questioned was either concealing a
DGU (false negative) or inventing one (false positive), based on Rstoneofvoice,
hesitation in responding, extraneous remarks, and so on. Both surveys found that
interviewer-suspected false negatives greatly outnumbered suspected false positives
(Kleck, 2001b, pp. 253254, based on the NSDS; secondary analysis of NSPOF data,
Police Foundation, 1994). These judgements, however, could be dismissed because
they are necessarily subjective and indirect.
There is, on the other hand, considerable hard evidence bearing indirectly on the
issue of response errors because it concerns elements of a DGU. We can begin with the
fact that most DGUs occur away from the victims home (Kleck & Gertz, 1995,p.185)
and that, in the 1990s, it was unlawful for anyone to carry a gun off their own property
unless they were among the few (under 1% back then Kleck 1997, Chapter 6) who
had a carry permit. Therefore, a survey respondent had to be willing to confess to a
crime (unlawful possession of a firearm) if they wanted to report a DGU that occurred
in a public place. In the NSDS, it was found that 63% of DGUs occurred in places other
than the defenders home (Kleck & Gertz, 1995, p. 185). Likewise, people forbidden to
possess guns regardless of location, such as convicted criminals, would have to confess
to a crime to report DGUs that occurred anywhere, even those in their own home. Thus,
even if the gun use itself involved lawful self-defense, most DGUs probably involve the
crime of unlawful possession of a weapon.
The technical literature on self-report surveys of offending consistently indicates that
few people report crimes that they did not commit, and many deny committing crimes
that they did commit. That is, false negatives greatly outnumber false positives, and
consequently response errors in surveys, on net, contribute to the underestimation of
the prevalence of criminal offending (reviewed in Kleck, 2001b).
Regardless of the location of the DGU, in order to report using a gun for protection,
one must also be willing to admit to possessing a gun. Research on survey reporting of
gun ownership has found that large shares of even law-abiding gun owners falsely deny
having guns, or refuse to answer a question about their gun ownership (Rafferty
Thrush, Smith, & McGee 1995; Kellermann, Rivara, Banton, Reay, & Fligner,
1990). For example, Rafferty and her colleagues found that 9% of registered handgun
owners refused to answer the gun ownership question when interviewed in a telephone
survey, and of those answering it, 13% denied having a gun in their household. On the
410 American Journal of Criminal Justice (2021) 46:401–421
other hand, I am not aware of any evidence of any significant numbers of false positive
responses regarding gun possession.
Finally, in order for survey respondents to be willing to report using a gun to protect
themselves against crime, they must be willing to report the victimization attempt itself.
Without a crime victimization experience, there can be no defense against crime.
Research on the reporting of crime victimization has long indicated that most crime
victims fail to report the victimization to police. The National Crime Victimization
Survey indicates that no more than a third of crimes are reported to the police (Hart &
Rennison, 2003). Further, reverse-record checks conducted by the Census Bureau
found that even among assaults that were reported to the police, only half were reported
to survey interviewers (Turner, 1972,p.26).
To summarize, for a person who had experienced a typical DGU to be willing to
report it to a surveyor, she or he would have to be willing to report (1) a crime they
committed (unlawful carrying), (2) possession of a gun, and (3) a crime victimization
experience. Research consistently indicates that false negative responses are common
in surveys asking about these topics, while false positives are rare. Therefore, as best
we can tell from empirical evidence currently available, the net effect of response errors
in surveys asking about DGU is likely to be an underestimation of DGU prevalence.
CDC Survey-Based Estimates Compared to those of Other National
Surveys
One way to judge the validity of DGU prevalence estimates derived from CDCs
surveys is to compare them with estimates derived from national surveys. Table 4
summarizes the results of 21 national surveys that asked a DGU question (recall that the
NCVS did not ask a DGU question). All were based on probability samples of the
national adult (age 18+) population and were conducted by professional survey orga-
nizations. All were telephone surveys except the Pew 2017 survey, which was an
Internet (Web-based) survey.
The surveys differed in the subsets of DGUs asked about, the subsets of the survey
sample asked the DGU questions, and in other respects, so their estimates had to be
adjusted if they were to be even moderately comparable with each other. They were
adjusted so as to produce standardized estimates of the share of U.S. adults who used any
kind of gun defensively against a human (rather than an animal) in the preceding
12 months, not including uses in connection with military, police, or security guard duties.
The adjustments are described in detail in the Appendix. Readers who are skeptical
about these adjustments may focus just on the surveys that did not require any
adjustments - the NSDS fielded in 1993 (Kleck & Gertz, 1995) and the NSPOF fielded
in 1994. Both yielded unadjusted DGU estimates that fall in the middle of the range of
adjusted estimates 2.5 million and 2.7 million respectively.
Original (unadjusted) DGU prevalence percentages from surveys that required
multiple adjustments were multiplied by as many adjustment factors as necessary.
For example, the 2000 Washington Post survey summarized in Table 4required
adjustments2,3,and4(seeAppendix). Its initial unadjusted DGU prevalence per-
centage was 8%, which was multiplied by adjustment factors 0.1628 × 0.90984 ×
0.806, yielding an adjusted DGU prevalence of 0.96%. We did not adjust for the
411American Journal of Criminal Justice (2021) 46:401–421
Table 4 National surveys of defensive gun use in the U. S.a
Survey: Cambridge Reports DMIa DMIb Hart Time/CNN Mauser Gallup
Time of Interviews: AprilMay 1978 MayJune 1978 December 1978 October 1981 December 1989 MarchApril 1990 May 1991
Sample Size: 1500 1500 1010 1228 605 344 1.002
Population covered: Adults Registered voters Registered voters Registered voters Firearm ownersAdults Adults
Gun Type Covered: Handguns All guns All guns Handguns All guns All guns All guns
Recall Period: Ever Ever Ever 5 years Ever 5 years Ever
Excluded Uses Against Animals? No No Yes Yes No Yes No
Excluded Military, Police Uses? No Yes Yes Yes Yes Yes No
DGU question asked of: Protection hgun owners All All All Gun owners All Rs in handgun
households
DGU question refers to: R Household Household Household R Household R
Unadjusted % adults with DGUb315749 3.798
Adjustments applied: 25,11 2,3 2 1,5 2,3,8 1 24,12
Adj. % with DGU 1.62 2.22 1.14 2.01 4.50 1.52 2.36
Population 18 or over 146.5 [76.030 m] [76.030 m] [82.368 m] ((59.5 m)) [93.347 m] 185.0 m
Implied number of DGUsd2.4m 1.7m 0.9m 1.5m 2.7m 1.4m 4.4m
Survey: Kleck & Gertz Gallup L.A Times U.S NEWS CDC ICARIS NSPOF Time/CNN
Time of Interviews: Feb.-April 1993 December 1993 April 1994 May 1994 April-Sept. 1994 Nov.-Dec. 1994 May 1995
Sample Size: 4997 1014 1682 1000 5238 2568 600
Population covered: Adults Adults Adults Adults Adults Adults Gun Owners
Gun Type Covered: All guns All guns All guns All guns All guns All guns All guns
Recall Period: 1 year Ever Ever 5 years 1 year 1 year Ever
Excluded Uses Against Animals? Yes No No Yes Yes Yes No
412 American Journal of Criminal Justice (2021) 46:401–421
Table 4 (continued)
Survey: Cambridge Reports DMIa DMIb Hart Time/CNN Mauser Gallup
Excluded Military Police Uses? Yes Yes Yes Yes Yes Yes No
DGU question asked of: All Gun Owners All All Rs in handgun households All Gun Owners
DGU question refers to: R R R R R R R
Unadjusted % adults with DGUb1.326 11 8c1 0.63 1.44 10
Adjustments applied: None 2,3,8 2,3,6 1 7,9 None 24,8,10
Adj. % with DGU 1.326 2.74 3.18 0.40 2.36 1.44 8.39
Population 18 or over 188.0 m 188.0 m 190.3 m 190.3 m 190.3 m 190.3 m ((62.4 m))
Implied number of DGUsd2.5m 5.1m 6.1m 0.8m 4.5m
d2.7 m 5.2
Survey: Hemenway & Azrael Hearst Hemenway Gallup Washington Post CNN Pew (Internet)
Time of Interviews: MayJune 1996 August 1997 Spring 1999 May 2000 June 2000 Sept. 2014 March 2017
Sample Size: 1906 2016 2474 1031 1068 1014 3844
Population covered: Adults Adults Adults Adults Adults Adults Adults
Gun Type Covered: All guns All guns All guns All guns All guns All guns All guns
Recall Period: 5 years Ever 5 years Ever Ever Ever Ever
Excluded Uses Against Animals? Yes Yes Yes No No No No
Excluded Military Police Uses? Yes No Yes No No No Yes
DGU question asked of: All All All All All Gun owners All
DGU question refers to: R R R R R R R
Unadjusted % adults with DGUb0.73 5 1.15 7 8 20 7
Adjustments applied: 1 2,4 1 242424,8 2,3
Adj. % with DGU 0.29 0.66 0.46 0.84 0.96 1.33 1.04
Population 18 or over 193.7 m 196.0 m 200.4 m 202.6 m 202.6 m 239.9 m 250.1 m
413American Journal of Criminal Justice (2021) 46:401–421
Table 4 (continued)
Survey: Cambridge Reports DMIa DMIb Hart Time/CNN Mauser Gallup
Implied number of DGUsd0.6m 1.3m 0.9m 1.7m 1.9m 3.2m 2.6m
Abbreviations: DMI = Decision Making Information; R = respondent; Hgun = handgun; m = million; DGU = defensive gun use; CDC = Centers for Disease Control and Prevention;
NSPOF = National Survey of the Private Ownership of Firearms
Notes: a. Table 4covers surveys of probability samples of the general U.S. population that directly asked Rs about DGU. It excludes the survey reported in McDowall et al. (2000),
which was instead based on samples of commercial lists of likely gun owners(p. 8), and the NCVS, which never asks Rs specifically about DGU
b. This percentage is the share of persons or households who reported a DGU for whatever recall period was used, for whatever subset of gun types or circumstances that happened to be
specifiedinthesurveys original question. Thus, these figures are generally not even minimally comparable across surveys without adjustment. Household-based prevalence figures
assume just one member of the household had a DGU
c. This survey inquired only about DGUs outside the home
d. Implied DGUs is for DGUs in connection with all types of crimes, projected from the surveys estimate for burglary-linked DGUs only
Population Bases to Which Adjusted DGU prevalence percentages were applied: Figures in the penultimate row for most surveys display the number of U.S. residents age 18 or over,
but those in brackets refer to number of households, and those in double parentheses refer to the estimated number of adults who personally own guns
Sources: Cook and Ludwig (1996); Hemenway, Azrael, and Miller (2000); Kleck (2001b); Mauser (1996,p.397);RoperCenter(2018); iPoll Databank online database of surveys
414 American Journal of Criminal Justice (2021) 46:401–421
different time periods in which surveys were fielded because we wanted the figures to
reflect changes over time. We could not adjust for any possible difference in DGU
involvement between registered voters and the adult population as a whole, because
there are no data on which an adjustment factor could be based.
Once the DGU prevalence percentage was adjusted, it was in most cases -
multiplied times the size of the U.S. adult population (age 18+) at the time of the
survey. For example, this population numbered 202.6 million in 2000 when the
Washington Post fielded their survey. Applying this surveys adjusted DGU percentage
of 0.96% times 202.6 million yields an estimated number of adult DGUs in 2000 of
about 1.9 million. For surveys of the population of registered voters, it was assumed
that DGU prevalence was the same in the entire adult population, including those not
registered to vote, as among adults registered to vote. For the two surveys whose
samples were confined to persons who personally owned guns at the time of the
interview (the Time/CNN surveys of 1989 and 1995), it would be implausible to
assume that DGU prevalence was the same among people who do not own guns as
it was among those who do. For these surveys, the adjusted DGU percentage was
multiplied times an estimate of the number of people who personally own guns. This
procedure yielded estimates of DGUs only among persons who reported gun ownership
at the time of the interview, estimates which are therefore incomplete and not compa-
rable with the other surveys. All final estimates of numbers of DGUs are conservative
in three respects: (1) they assume only one DGU-involved adult per household
reporting a DGU, (2) they assume only one DGU incident per DGU-involved person
in the recall period, and (3) they do not include any DGUs by adolescents, persons
institutionalized at the time of the survey, or persons without telephone service.
The annual number of DGUs implied by the 21 national surveys summarized in
Table 4ranges from 0.6 million to 6.1 million, averaging 2.2 million. Figure 1visually
illustrates the distribution of estimates, displaying the number of surveys that yielded
estimates in each range.
Fig. 1 Distribution of Survey Estimates of Annual Defensive Gun Uses (DGUs)
415American Journal of Criminal Justice (2021) 46:401–421
The estimates we derived from the three CDC surveys of 0.6, 0.9, and 1.9 million
(Table 3) all fall within the range of estimates generated by previous surveys, though
the agencysanti-gunreputation among some gun owners may have biased its DGU
estimates to the low side. The CDC surveys confirm what private surveys have found
defensive use of firearms by crime victims is common in the U.S.
The sources for these survey results (such as the Roper iPoll database) generally do not
report standard errors or confidence intervals. Further, most of them use complex
sampling designs that depart from simple random sampling, so introductory textbook
formulae for computing standard errors are not strictly applicable. Nevertheless, for the
sake of giving readers a rough idea of the magnitude of sampling error, estimates were
computed, using the textbook formula, for the two most recent national surveys. The 95%
confidence interval estimate of the percent of the adult population with a DGU was 0.625
2.035% for the 2014 CNN poll and 0.8761.204% for the 2017 Pew survey. These
estimates imply 1.54.4 million annual DGUs based on the CNN poll and 2.23.0 million
for the Pew survey. The intervals are wide, but even the lower limits imply numbers of
DGUs much larger than the number of crimes committed with guns. In any case, the
sampling error in any one survey becomes less important if you have 21 surveys, with a
collective sample in the tens of thousands, all yielding large DGU estimates.
These estimates indicate that defensive use of firearms by crime victims is far more
common than the highest estimates of offensive uses by criminals based on victim
surveys. The National Crime Victimization Survey estimated that there were 416,350
violent crime incidents committed by offenders with firearms in 2016, including crimes
not reported to the police (U.S. Bureau of Justice Statistics, 2017). By comparison, the
average of the estimates of annual DGUs based on the 2014 CNN poll and the 2017
Pew survey was 2.9 million (Table 4)seven times larger than the number of crimes
committed by gun-armed offenders in 2016.
Discussion
Some cautions are in order about the CDC estimates of DGU prevalence. First, there is
no way to adjust for any reluctance of gun owners to report defensive gun uses to CDC
interviews or to even participate in a CDC-conducted survey, since there is no way to
calculate how much NRA efforts to characterize CDC as anti-gunhave influenced
the population of gun owners.
Second, the factors we used to extrapolate from results based on four to six states up
to the nation as a whole depend on small numbers of persons reporting DGUs in those
subsets of states in the NSDS. The number of persons residing in those states who were
asked the DGU questions was substantial (between 337 and 535), which minimized
sampling error, but the number reporting a DGU in those subsets of the NSDS was
small, which makes DGU prevalence estimates in those subsamples sensitive to
response error. As previously noted, the dominant response error is likely to be false
negatives, which would tend to make these DGU prevalence estimates too low. This is,
however, true for Rs in both the national NSDS sample as a whole, and for Rs in the
BRFSS residing in states that asked the DGU question.
Third, there are limitations to the BRFSS surveys that parallel those of the NSDS.
Although the numbers of persons asked the DGU question was very large in the
416 American Journal of Criminal Justice (2021) 46:401–421
BRFSS states (between 3197 and 5484) and this minimized sampling error, the raw
numbers of Rs reporting a DGU in the BRFSS surveys was just 2955 in each of the
three surveys, which makes the estimates of DGU prevalence more vulnerable to
response error of the sort discussed in the previous paragraph. Again, in light of prior
evidence suggesting that false negative responses are more common than false positive
responses, the effect of response errors is likely to be predominantly one that makes the
BRFSS-based estimates of DGU prevalence too low.
Conclusions
In sum, even when CDC, an organization perceived by some to be strongly anti-gun,
devised and conducted the surveys, their survey results implied huge estimates of
defensive gun uses over a million per year, far more than the number of violent
crimes in which offenders used guns. Although the CDC routinely reports results of the
BRFSS regarding a wide variety of topics on their website, including results pertaining
to subareas of the nation (CDC, 2018b) and even results pertaining to individual states
(CDC, 2018c), the CDC has not reported their DGU results.
These estimates are relevant for the scholarly purpose of describing the full set of
consequences of widespread gun ownership in America, but do they imply anything
about public policy on the regulation of firearms? If one were considering only
moderate controls that would not disarm significant numbers of noncriminal Ameri-
cans, such as background checks, DGU estimates would be of little relevance since
such policies would presumably not substantially reduce defensive use of firearms by
noncriminals. The number of DGUs is, however, highly relevant to the relative costs
and benefits of prohibitionist controls aimed at disarming everyone, including the
noncriminal majority.
Organizations like the Brady Campaign to Prevent Gun Violence, the Violence
Policy Center, and Everytown for Gun Safety insist that they do not favor
prohibition but their denials are always carefully phrased so as to pertain only
to the policies that they are currently actively advocating policies that indeed do
not include prohibition (Kleck, 2001a). Given current political realities, and public
opinion polls indicating that most Americans presently oppose banning guns, or
even just handguns, it would be futile to lobby for gun prohibition. The October
2017 Gallup poll found that only 28% of U. S. adults favor banning the possession
of handguns (Roper Center, 2018). Further, any admission of future prohibitionist
ambitions would play into the NRAsslippery slopeargument that proposals for
moderate controls must be opposed because they would incrementally lead to
banning guns altogether (Kleck, 2001a).
None of the major gun control advocacy organizations, however, have publicly
committed themselves to never favoring prohibition in the future. There is considerable
evidence that the leaders and activist members of the major advocacy groups do think
that banning gun possession in the general civilian population would be a good idea,
and that they would favor it if its achievement ever became politically feasible (Kleck,
2001a). To the extent that the fight over guns eventually becomes a fight over gun
prohibition, the incidence of victims using firearms to defend themselves will become
increasingly relevant to policy.
417American Journal of Criminal Justice (2021) 46:401–421
AuthorsContributions No applicable.
Availability of Data and Material NSDS Data are available from author; CDC data are public.
Compliance with Ethical Standards
Conflicts of Interest/Competing Interests None.
Code Availability Not applicable.
Appendix
The following adjustments were applied to the Table 4DGU prevalence percentages in
order to make them more comparable with each other.
Adjustment 1 was applied to surveys inquiring about a five-year recall period, in
order to produce an estimate pertaining to a one-year recall period. The NSDS
obtained a DGU prevalence of 1.326 when using a one-year recall period, and
3.315 when using a five-year recall period (Kleck & Gertz, 1995, p. 184).
Therefore, adjustment 3 consisted of multiplying a prevalence figure based on a
five-year recall period by 0.40000 (1.326/3.315 = 0.40).
Adjustment 2 was applied to surveys using a lifetime recall period (have you ever.
.. ), in order to produce an estimate pertaining to a one-year recall period. A
California poll was unique in asking a DGU question for both a one-year recall
period and a lifetime recall period, and indicated that the same survey yielded a 1.4
percent prevalence figure for a one-year recall period and an 8.6 percent figure for
lifetime period (Field Institute, 1976), so the adjustment consists of multiplying an
ever usedestimate by 0.16279 (1.4%/8.6%=0.16279).
Adjustment 3 was applied to surveys that failed to exclude uses of guns against
animals, in order to produce an estimate pertaining only to uses against humans.
The NSDS (Kleck & Gertz, 1995) indicated that of 244 Rs initially reporting
DGUs, 22 had used guns only against animals, so the adjustment consists of
multiplying a humans-plus-animals estimate by 0.90984 (222/244).
Adjustment 4 was applied to surveys that failed to exclude incidents linked with
military, police, or security guard duty. An experiment by McDowall et al. (2000)
showed that when Rs were not instructed to exclude such incidents, 30 of the 155
Rs who initially reported a DGU were found to be reporting these kinds of duty-
related experiences. The adjustment consisted of multiplying the DGU prevalence
percentage by 0.806 (125/155 = 0.806).
Adjustment 5 was applied to surveys that asked only about DGUs committed with
handguns. Kleck and Gertz (1995, p. 185) found that 79.7% of DGUs were
committed with handguns, so the adjustment consisted of multiplying a
handgun-only DG prevalence percentage by 1.255 (100/79.7 = 1.255).
Adjustment 6 was applied to a survey that inquired only about DGUs occurring
outside the Rs home. Kleck and Gertz (1995) that only 62.7% of DGUs occurred
outside the users home, so the adjustment consisted of multiplying the DGU
prevalence percentage by 1.595 (100/62.7 = 1.595).
418 American Journal of Criminal Justice (2021) 46:401–421
Adjustment 7 was applied to a survey that asked only about burglary-linked DGUs.
Kleck and Gertz (1995) found that 33.8% of DGUs were linked with burglaries, so
the adjustment consisted of multiplying the DGU prevalence percentage by 2.959
(100/33.8 = 2.959).
Adjustment 8 was applied to surveys that asked the DGU question only of Rs who
reported personally owning a gun or handgun, in order to produce an estimate
pertaining to the entire population and thus reflecting uses among those who do not
report current ownership of a gun. The NSDS indicated that only 59.5% of Rs
reporting DGUs reported current personal ownership of a gun (Kleck & Gertz,
1995, p. 187), so the adjustment consists of multiplying a gun-owners-only
estimate by 1.681 (1/0.595 = 1.681).
Adjustment 9 was applied to surveys that asked the DGU question only of Rs who
lived in households reporting current gun ownership, in order to produce an
estimate pertaining the entire population and thus reflecting uses among those
who do not live in a household reporting current ownership of a gun. The NSDS
indicated that only 79.0% of Rs reporting DGUs reported current household
ownership of a gun (Kleck & Gertz, 1995, p. 187), so the adjustment consists of
multiplying a gun-owning household-only estimate by 1.26582 (1/0.790).
Adjustment 10 was applied to the 1989 Time/CNN survey that only asked about
DGUs in which the gun was fired. Kleck and Gertz (1995, p. 185) found that only
23.9% of DGUs involve the gun being fired, even as a warning shot, so the
adjustment consisted of multiplying the guns-fired DGU prevalence by 4.184
(100/23.9 = 4.184).
Adjustment 11 was applied to the 1978 Cambridge Reports survey, which only
asked the DGU question of Rs who reported personally owning a handgun for
purposes of protection.Analysis of the NSPOF dataset shows that only 27.69%
of DGUs were by Rs who personally owned a handgun primarily for self-protec-
tion, so the adjustment consisted of multiplying the unadjusted DGU prevalence by
3.6111 (100/27.69 = 3.6111).
Adjustment 12 was applied to the 1991 Gallup poll, which only asked the DGU
question of Rs who reported household handgun ownership at the time of the
survey. Analysis of the NSPOF dataset shows that only 40.46% of DGUs were by
Rs who lived in a household with handguns, so the adjustment consisted by
multiplying the unadjusted DGU prevalence by 2.4717 (100/40.46 = 2.4717).
References
Centers for Disease Control and Prevention (CDC). (1998). 1997 BRFSS Summary Quality Control Report.
https://www.cdc.gov/brfss/annual_data/1997/pdf/1997summarydataqualityreport.pdf. Accessed 14
Dec 1998.
Centers for Disease Control and Prevention (CDC). (2018a). Behavioral Risk Factor Surveillance System
website at https://www.cdc.gov/brfss/annual_data/annual_data.htm. Accessed 14 Feb 2018.
Centers for Disease Control and Prevention (CDC). (2018b). MMWR Surveillance Summaries. Web page at
https://www.cdc.gov/brfss/publications/ssummaries.htm. Accessed 14 Feb 2018.
Centers for Disease Control and Prevention (CDC). (2018). BRFSS facts sheets. Web page at https://www.
cdc.gov/brfss/factsheets/index.htm. Accessed 14 Feb 2018.
419American Journal of Criminal Justice (2021) 46:401–421
Cook, P. J., & Ludwig, J. (1996). Guns in America: Results of a comprehensive national survey on firearms
ownership and use. Washington, DC: Police Foundation.Cox, C. 2018. Why we canttrusttheCDC
with gun research.Politico,https://www.politico.com/agenda/story/2015/12/why-we-cant-trust-the-cdc-
with-gun-research-000340.Accessed8Dec2018.
Dillman, D. A., Smyth, J. D., & Christian, L. M. (2009). Internet, mail, and mixed-mode surveys. New York,
NY: Wiley.
Federal Bureau of Investigation. (1999). Crime in the United States, 1998. Washington, DC: U.S. Government
Printing Office.
Field Institute. (1976). Tabulations of the findings of a survey of handgun ownership and access among a
cross section of the California adult public. San Francisco: Field Institute.
Hemenway, D., & Azrael, D. (2000). The relative frequency of offensive and defensive gun uses: Results from
a national survey. Violence and Victims, 15, 257272.
Hemenway, D., Azrael, D., & Miller, M. (2000). Gun use in the United States: Results from two national
surveys. Injury Prevention, 6,263267.
Hemenway, D. 1997. Survey research and self-defense gun use: an explanation of extreme overestimates.
Journal of Criminal Law and Criminology, 87, 14301445.
Hart, T. C., & Rennison, C. (2003). Reporting crime to the police, 19922000. Bureau of Justice Statistics
Special Report. At https://www.bjs.gov/content/pub/pdf/rcp00.pdf. Accessed 14 Dec 2003.
Ikeda, R. M., Dahlberg, L. L., Sacks, J. J., Mercy, J. A., & Powell, K. E. (1997). Estimating intruder-related
firearm retrievals in U.S. households, 1994. Violence and Victims, 12,363372.
Jamieson, C. 2013. Gun violence research: History of the federal funding freeze.American Psychological
Association website at http://www.apa.org/science/about/psa/2013/02/gun-violence.aspx.Accessed29
Nov 2013.
Kellermann, A. L., Rivara, F. P., Banton, J., Reay, D., & Fligner, C. L. (1990). Validating survey responses to
questions about gun ownership among owners of registered handguns. American Journal of
Epidemiology, 131,10801084.
Kleck, G. (2001a). Absolutist politics in a moderate package: Prohibitionist intentions of the gun control
movement. In G. Kleck & D. B. Kates (Eds.), Armed: New perspectives on gun Control (pp. 129172).
New York, NY: Prometheus Books.
Kleck, G. (2001b). The frequency of defensive gun use: Evidence and disinformation. In G. Kleck & D. B. Kates
(Eds.), Armed: New perspectives on gun Control (pp. 213284). New York, NY: Prometheus Books.
Kleck, G. 1997. Targeting Guns: Firearms and their Control. NY: Aldine.
Kleck, G., & Gertz, M. (1995). Armed resistance to crime: The prevalence and nature of self-defense with a
gun. Journal of Criminal Law & Criminology, 86(1), 150187.
Mauser, G. (1996). Armed self-defense. JournalofCriminalJustice,24,393406.
McDowall, D., Loftin, C., & Presser, S. (2000). Measuring civilian defensive firearm use. Journal of
Quantitative Criminology, 16,119.
McDowall, D., & Wiersema, B. (1994). The incidence of defensive firearm use by US crime victims, 1987-
1990. American Journal of Public Health, 84, 19821984.
Police Foundation. 1994. National study of private ownership of firearms in the United States, 1994.Ann
Arbor, MI: [distributor], 1998-10-08. Computer dataset. https://doi.org/10.3886/ICPSR06955.v1.
Rafferty, A. P., Thrush, J. C., Smith, P. K., & McGee, H. B. (1995). Validity of a household gun question in a
telephone survey. Public Health Reports, 110,282288.
Roper Center. (2018). Roper Center for Public Opinion Research website at https://ropercenter.cornell.
edu/ipoll-database/. Accessed 12 Dec 2018.
Turner, A. G. (1972). The San Jose recall study. In R. G. Lehnen & W. G. Skogan (Eds.), The national crime
survey: Working papers, volume I: Current and historical perspectives (pp. 2227). Washington, DC:
U.S. Government Printing Office.
U.S. Bureau of the Census. (1998). Statistical abstract of the United States, 1998. Washington, DC: U.S.
Government Printing Office.
U.S. Bureau of Justice Statistics. (2017). Criminal victimization, 2016. Washington, DC: U.S. Government
Printing Office.
PublishersNote Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
420 American Journal of Criminal Justice (2021) 46:401–421
Gary Kleck is the Emeritus David J. Bordua Professor of Criminology and Criminal Justice at Florida State
University. His research has focused on the impact of firearms and gun control on violence, deterrence, crime
control, and violence. He is the author of Point Blank: Guns and Violence in America, which won the 1993
Michael J. Hindelang Award of the American Society of Criminology. He also wrote Targeting Guns (1997)
and, with Don B. Kates, Jr., The Great American Gun Debate (1997), Armed (2001), and (with Brion Sever)
Punishment and Crime (2017). Kleck has testified before Congress and state legislatureson gun control issues,
his work has been cited by the U.S. Supreme Court, and he has worked as a consultant to the National
Research Council, National Academy of Sciences Panel on the Understanding and Prevention of Violence,
and as a member of the U.S. Sentencing Commissions Drugs-Violence Task Force.
421American Journal of Criminal Justice (2021) 46:401–421
Reproduced with permission of copyright owner.
Further reproduction prohibited without permission.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
There is a vigorous debate over the frequency with which private citizens resort to the use of firearms for self-defense. No information has been previously available about how often firearms are used defensively outside of the United States. This article estimates the frequency with which firearms are used for self-protection by analyzing three telephone surveys of the general public in Canada and a fourth survey of the general public in the United States. Canadians report using firearms to protect themselves between 60,000 and 80,000 times per year from dangerous people or animals. More importantly, between 19,000 and 37,500 of these incidents involve defense against human threats. The results of the American survey confirm estimates about the frequency of firearms used for self-protection in the United States (Kleck, 1988, 1991). In comparison with the number of households with firearms, the frequency with which Canadians use firearms to defend themselves against human threats is somewhat less than that of Americans. Policymakers in both the United States and in Canada should be aware the private ownership of firearms has benefits as well as costs for society. Firearm bans may cost more lives than they save.
Article
Estimates of the incidence of victim gun use from the National CrimeVictimization Survey (NCVS) are consistently lower than are those fromother studies. To examine the divergence, we conducted a survey that gaugedthe impact of methodological differences between the NCVS and the otherstudies. For half of the sample, we asked questions from the NCVS, followedby questions from the other surveys. For the other half of the sample, wepresented the questions in the reverse order. We examined two hypotheses:(1) survey methods account for the divergent results, and (2) the questionscover unrelated activities. The results provided some support for the firsthypothesis, but respondents also reported many more defenses to thequestions from the other surveys than to the NCVS questions. Consistent withthe second hypothesis, this suggests that the NCVS and the other surveysmeasure responses to largely different provocations.
Article
While survey-based data on firearm ownership are essential for epidemiologic studies of the relation between gun ownership and injuries, the validity of respondent answers to questions about gun ownership has not been confirmed. In order to assess the accuracy of interview data about firearms, in June to August 1987, the authors attempted to contact residents of 75 homes in the cities of Memphis, Tennessee and Seattle, Washington listed as the address of the owner of a recently registered handgun. Despite problems with inaccurate registration data, contact was ultimately made with 55 households, 35 of which consented to a general interview that included a series of questions about gun ownership. Respondents in 31 of these 35 households (88.6%) readily acknowledged that one or more guns were kept in their home. Respondents in three of the remaining four households (8.6%) stated that guns were recently kept in their homes but were no longer kept there. Only one respondent (2.9%) denied categorically that guns of any kind were kept in her home. The authors conclude that, at least among registered gun owners, respondent answers to questions about gun ownership are generally valid and that survey data of this type can be utilized with confidence.