Administrative and Convenience Samples

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.