Administrative and Convenience Samples
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
102 103 104 105 106 107 108 109 110 111 112 113 114 115
A number of administrative data sets have been used or suggested as a
way to study the market for firearms possession and use. In this section, we
describe the administrative data collected as part of the Bureau of Alcohol,
Tobacco, and Firearms’ tracing system, the trace data, and a proposed
addendum on firearms to the National Institute of Justice survey of
arrestees, the Arrestee Drug Abuse Monitoring (ADAM) survey.
BATF Firearms Trace Data: One federal source of information on
firearms related to violence is the firearms trace data compiled by the
Bureau of Alcohol, Tobacco, and Firearms of the U.S. Department of Justice.
Because trace data are quite distinct from the other federal data sources,
and because they have been subject to more criticism than most of the other
systems, we provide a more extensive description of the regulatory background
related to firearm tracing, the nature of the tracing process, and the
uses and limitations of the resulting data.
The Gun Control Act of 1968 established the legal framework for
regulating firearms transactions and the associated record-keeping. The act
was intended to limit interstate commerce in guns, so that states with strict
regulations were insulated from states with looser regulations (Zimring,
1975). To that end, the act established a system of federal licensing for gun
dealers, requiring that all individuals engaged in the business of selling guns
must be a federal firearms licensee (FFL). The FFLs were established as the
gatekeepers for interstate shipments: only they may legally receive mailorder
shipments of guns, and they may not sell handguns to residents of
another state. FFLs are required to obey state and local regulations in
transacting their business.
The Gun Control Act established conditions on the transfer of firearms.
FFLs may not sell handguns to anyone under the age of 21, or long
guns to anyone under the age of 18, nor may they sell a gun of any kind to
someone who is proscribed from possessing one. The list of those proscribed
by federal law includes individuals with a felony conviction or
under indictment, fugitives from justice, illegal aliens, and those who have
been committed to a mental institution. FFLs must require customers to
show identification and fill out a form swearing that they do not have any
of the disqualifying conditions specified in the Gun Control Act. Beginning
in 1994, the Brady Violence Prevention Act required that FFLs initiate
a background check on all handgun purchasers through law enforcement
records; in 1998 that requirement was expanded to include the sale
of long guns as well.
The 1968 Gun Control Act also established requirements that allowed
for the chain of commerce for any given firearm to be traced from its
manufacture or import through its first sale by a retail dealer. Each new
firearm, whether manufactured in the United States or imported, must be
stamped with a unique serial number. Manufacturers, importers, distributors,
and FFLs are required to maintain records of all firearms transactions,
including sales and shipments received. FFLs must report multiple handgun
sales and stolen firearms to BATF and provide transaction records in response
to its trace requests. When FFLs go out of business, they are to
transfer their transaction records to BATF, which then stores them for
tracing (Bureau of Alcohol, Tobacco, and Firearms, 2000a). In essence, the
1968 act created a paper trail for gun transactions that can be followed by
BATF agents.
The tracing process begins with a law enforcement agency’s submission
of a trace request to BATF’s National Tracing Center (NTC). The form
requires information regarding the firearm type (i.e., pistol, revolver, shotgun,
rifle), the manufacturer, caliber, serial number, and importer (if the
gun is of foreign manufacture), the location of the recovery, the criminal
offense associated with the recovery, and the name and date of birth of the
firearm possessor (Bureau of Alcohol, Tobacco, and Firearms, 2000a). This
information is entered into BATF’s Firearms Tracing Center at the NTC
and checked against the records of out-of-business FFLs that are stored by
BATF, as well as records of multiple handgun purchases reported on an
ongoing basis by FFLs. If the gun does not appear in these databases, NTC
contacts the firearm manufacturer (for domestic guns) or the importer (for
foreign guns) and requests information on the distributor that first handled
the gun. BATF then follows the chain of subsequent transfers until it identifies
the first retail seller. That FFL is then contacted with a request to
search his or her records and provide information on when the gun was
sold and to whom.
In 1999, trace requests for 164,137 firearms were submitted by law
enforcement agencies to NTC. Of these, 52 percent (85,511) were successfully
traced to the first retail purchaser. The 48 percent of trace requests
that failed did so for a variety of reasons. Nearly 10 percent of the
guns (15,750) were not successfully traced because they were too old (pre-
1968 manufacture) and another 11 percent (17,776) failed because of
problems with the serial number (Pierce et al., 2002). The majority of the
remaining unsuccessful trace requests failed because of errors on the submission
forms or problems obtaining the information from the FFL who
first sold the gun at retail. It is important to note that, even when a trace
is “successful,” it provides limited information about the history of the
gun (Cook and Braga, 2001). Most successful gun traces access only the
data on the dealer’s record for the first retail sale of the gun. Generally,
subsequent transactions cannot be traced from the sorts of records required
by federal firearms laws.
Beginning in 1993, the Clinton administration was concerned about
the apparent ease with which criminals and juveniles obtained guns. BATF
was charged with initiating a concerted effort to increase the amount of
crime gun tracing, improve the quality of firearms trace data, increase the
regulation of gun dealers, educate law enforcement on the benefits of tracing,
and increase investigative resources devoted to gun traffickers. Comprehensive
tracing of all firearms recovered by police is a key component of
BATF’s supply-side strategy to reduce the availability of illegal firearms. In
1996, BATF initiated the Youth Crime Gun Interdiction Initiative with
commitments from 17 cities to trace all recovered crime guns (Bureau of
Alcohol, Tobacco, and Firearms, 1997). This program expanded to 38
cities in 1999 (Bureau of Alcohol, Tobacco, and Firearms, 2000a) and to 55
cities in 2001 (Bureau of Alcohol, Tobacco, and Firearms, 2002b). Other
jurisdictions have also expanded their use of gun tracing; six states, for
example, have recently adopted comprehensive tracing as a matter of state
policy, either by law (California, Connecticut, North Carolina, and Illinois),
by executive order (Maryland), or by law enforcement initiative (New
Jersey) (Bureau of Alcohol, Tobacco, and Firearms, 2000a).
Understandably, research studies based on analyses of firearms trace
data have been greeted with a healthy dose of skepticism. Although the
quality of firearms trace data has improved over the past decade (Cook and
Braga, 2001), trace data analyses are subject to a number of widely recognized
problems (see Kleck, 1999; Blackman, 1999; Congressional Research
Service, 1992).9 All are based on firearms recovered by police and other
law enforcement agencies, which may not be representative of firearms
possessed and used by criminals. Trace data are also influenced by which
guns are submitted for tracing, a decision made by law enforcement agencies.
Beyond that, not all firearms can be traced. The trace-based information
that results is biased to an unknown degree by these factors.
Furthermore, trace data cannot show whether a firearm has been illegally
diverted from legitimate firearms commerce. Trace studies typically
contain information about the first retail sale of a firearm and about the
circumstances associated with its recovery by law enforcement. These studies
cannot show what happened in between: whether a firearm was legitimately
purchased and subsequently stolen, sold improperly by a licensed
dealer, or any other of a myriad of possibilities. As such, trace analysis
alone cannot reveal the extent and nature of illegal firearms trafficking.
Ultimately, the validity of the conclusions drawn from these data depends
on the application. In general, trace data are not informative about
populations of interest, such as offenders, potential offenders, victims, and
the general population.
Administered until recently by the National Institute of Justice, Arrestee
Drug Abuse Monitoring (ADAM, formerly known as the Drug Use
Forecasting program, or DUF) contains survey data and urine samples from
samples of arrestees charged with felonies and misdemeanors at 35 sites
across the county. Data collection occurred four times a year. Response
rates were relatively high: about 85 percent of arrestees agreed to interview
(http://www.adam-nij.net). ADAM focused on drug use patterns among
criminal suspects and did not regularly collect data on firearms use. However,
in 1996 researchers appended a “gun addendum” to the surveys in 11
sites to study patterns of gun acquisition and use among arrestees (Decker
et al., 1997).
Decker and colleagues (1997) suggested how the addendum might be
used to provide estimates of the frequency and characteristics of arrests in
which the arrested persons owned and used firearms (National Research
9Comprehensive tracing of all firearm recoveries reduces some of the problems in trace data
introduced by police decision making. Jurisdictions that submit all confiscated guns for tracing
can be confident that the resulting data base of trace requests represents the firearms
recovered by police during a particular period of time.
Council, 2003).10 Tracking firearms possession through arrest might also
serve to detect emerging problems in high-risk populations. Quarterly data
collection, such as that conducted by ADAM, permits monitoring of local
trends over short time intervals.
Such data, however, may not be useful for answering many of the
policy-related questions considered in this report. Although the ADAM
samples are representative of the local arrestee populations for which the
surveys were administered, the 35 data collection sites are not a representative
sample of urban areas nationwide. Moreover, a survey of arrestees
cannot be used to infer acquisition and use among criminals or the general
population. The data are not representative of the relevant populations and
can be influenced heavily by police priorities and procedures. Thus, these
data alone cannot be used to infer the effects of guns on crime or the effects
of interventions on gun use or the market for weaponry in general.
Suppose, for example, one found that the fraction of arrestees reported
to possess firearms does not vary by the strength of local regulations. It may
be, as suggested by the data, that regulation has no effect on the market. It
may also be that regulation affects the crime and the ownership rates, but
among the arrested populations the ownership and use rates are unchanged.
And it may be that regulations influence policing and the accuracy of selfreporting
in unknown ways. ADAM data do not reveal the association between
regulation and the behavior of offenders, potential offenders, the crime
rate, or policing. Thus, observing that the prevalence of gun ownership and
use among arrestees changes after some interventions does not reveal how
gun use or crime more generally changed in the population of interest.
Proxy Measures of Ownership
Using proxy measures of ownership raises different issues and questions.
In the proxy approach to measuring ownership (proxy approaches
have not been developed as measures of firearms use) researchers have
sought to find measures that would indicate whether firearms were available.
A variety of these have been proposed, but it appears that the one the
research community has settled on is the proportion of suicides committed
with a firearm (Kleck, 1991; Cook, 1991). This measure has been found to
10This study, for example, reveals that 14 percent of arrestees carried firearms almost all of
the time, that arrestees who tested positive for drugs were no more likely than others to own
or use firearms, that the most frequently cited reason given for owning a gun was the need for
protection or self-defense (two-thirds), that more than half of the arrestees (55 percent) said
that guns are easy to obtain illegally, that 23 percent of arrestees who owned a gun reported
using a gun to commit a crime, and that 59 percent of arrestees reported that they had been
threatened with a gun.
correlate better than other possible proxies with measures of gun violence
(homicide and gun assaults).
As we discuss in Chapter 7, proxies raise two somewhat related but
distinct methodological issues. First, proxies have been used at aggregated
levels, most often the state level, to infer something about the impact of
availability at the individual level on violent outcomes. For example, if the
proxy is correlated with gun homicides at the state level, then it is often
assumed that availability at the individual level of analysis is associated
with individual manifestations of violence. More generally, these studies
are used to infer whether an individual’s probability of access to firearms
explains his or her probability of committing a violent crime or suicide.
Aggregate measures of ownership, however, may or may not be related to
actual availability in the households in which these rare events (homicides
and suicides) occur.
A second issue with proxies is to what extent they are inaccurate indicators
of firearms availability at the geographic level of interest. Proxies
create biases, yet there is almost no research on these statistical problems in
the firearms literature. Without more rigorous evaluations on the impact of
proxies, it is difficult to assess the research on ownership and violence.
Once these biases have been assessed, proxies may be useful because they
are cheaper to collect, their collection is less intrusive, and for other reasons
of economy or design. The research community in this area needs to focus
more attention on assessing the biases created by proxies and on the development
of better direct measures of availability and use.