In reporting on the Firearms Licensee Survey, we use “retailer” only to refer to an individual person.
Identifying the Study Population
A roster of all federal firearms licensees, including manufacturers, importers, and collectors, as well as gun dealers and pawnbrokers, is available online from the Bureau of Alcohol, Tobacco, Firearms and Explosives (ATF) and is updated monthly.16
Entries contain the name of the licensee (who may be an individual person or a corporation), the license type and number, the business name and telephone number, and both premises and mailing addresses. A separate license is required for each fixed location at which firearms transactions are conducted; some large chain-store corporations hold dozens of licenses. In such cases, the mailing address for all locations is typically that of the corporate headquarters, but the premises addresses are specific to individual retail establishments. We used the list for February 2011 to create a file with records for all 55,020 licensees who were classified as retail establishments: gun dealers and gunsmiths, who hold Type 01 licenses, and pawnbrokers, who hold Type 02 licenses.
Retail licensees may sell firearms only occasionally or not at all, and we wished to restrict the study population to those who were known to be more than occasional sellers of firearms. Licensees do not report their sales in most states; the data for a direct determination do not exist. However, the Federal Bureau of Investigation (FBI) collects data on background checks performed by its National Instant Criminal Background Check System (NICS checks) on firearm purchasers at the request of licensees. NICS checks are required for the great majority of firearm sales made by licensed gun dealers and pawnbrokers. Records for individual requests are retained for 90 days after the requests are made.
The number of NICS checks requested by a licensee does not equal the number of firearms that licensee sells, for several reasons. (1) In so-called Brady alternative states, purchases by holders of permits to carry concealed firearms and similar permits are exempt from the requirement.17
There were 18 such states when the study was conducted. (2) Only one NICS check is required per transaction, regardless of the number of firearms involved. (3) A small percentage of sales are denied (the nationwide average is 1.8 %18
), and the sale does not proceed. (4) Background checks are also required when firearms are redeemed from pawnbrokers. Nonetheless, NICS checks may be considered a rough proxy measure for firearm transactions for both dealers and pawnbrokers.
In response to a Freedom of Information Act (FOIA) request, the FBI provided a tabulation of NICS check requests, specific to individual licensees, for the 88-day period November 13, 2010, through February 9, 2011. These dates were not chosen deliberately; the data were compiled as soon as possible after the FOIA request was approved. The tabulation was provided as a 723-page PDF document. We manually reviewed the document to identify all licensees having Type 01 or Type 02 licenses and ten or more NICS checks. This number was chosen to represent, very roughly, 50 firearm sales and/or redemptions per year (approximately one per week): approximately 40 transactions [10
(365/88)], with an allowance for transactions involving multiple firearms.
We identified 9,720 licensees having ten checks or more and prepared a sampling frame containing their records from the ATF roster. These licensees accounted for 17.7 % of all retail licensees on the roster. A random sample of licensees, stratified by license type, was drawn using PROC SURVEYSAMPLE in SAS software.19
The sample size was 1,601 subjects, chosen to provide 95 % confidence intervals of ±3 % when equal (50 %) proportions of respondents provided alternate responses to questions with binary responses, and the response rate was 60 %.20
The final sample included 16.5 % of licensees in the sampling frame.
No licensees from seven states (California, Connecticut, Hawaii, New Jersey, Nevada, Pennsylvania, and Virginia) were identified in the FBI tabulation as having NICS check requests. The non-represented states are all full Point of Contact (POC) states, in which firearms licensees contact a state agency, not the FBI, for NICS checks on all firearm transactions.21
These states may include the license number of the requesting licensee when submitting the NICS check request to the FBI but are not required to and frequently do not. Because the data are incomplete, the FBI generally does not include POC state transactions in their tabulations of NICS checks by licensee (personal communication, Andrew F. Clay, December 9, 2011). Retail licensees from these states accounted for 12.6 % of all retail licensees on the ATF roster.
Questionnaire design followed recommendations made by Dillman and colleagues.20,22
The questionnaire was designed to promote readability and was limited to 38 questions on 12 pages, with a final text box for respondents to offer additional comments. A nationally recognized expert in survey research, including research on topics related to firearms, served as a consultant throughout the development of the questionnaire.
We were concerned that there might be adverse effects on the implementation of the study should its existence be disclosed prematurely and did not pre-test the questionnaire on a sample of licensed retailers. Instead, we conducted in-depth, multi-session cognitive interviews with two independent experts who had extensive knowledge of the firearms industry and its practices, and three policy development experts reviewed a draft of the questionnaire.
Some firearm licensees are named individuals, and others are corporations. Along with questions about their demographic characteristics, subjects were asked whether they themselves were the named individual to whom the questionnaire was addressed, if such an individual existed. They were asked an open-ended question about the nature of their job at the retail establishment where they worked.
Questions about the number and types of firearms sold requested data for calendar year 2010. Subjects were asked how many firearms and handguns they sold and the percentage of their overall sales that were to law enforcement personnel, women, individuals who purchased multiple firearms within five business days, over the Internet, or at gun shows. They were asked for the percentage of their handgun sales that were of handguns “priced at more than $250” (inexpensive “Saturday night specials” have new retail prices
$200, and handguns from other manufacturers are priced above $250) and for the percentage of their rifle sales that were of “tactical rifles or modern sporting rifles, such as ARs, AKs, and SKSs.”
Questions regarding denied sales and firearm trace requests, which are infrequent events, requested annual averages over the 5 years preceding the survey—as a percentage of all firearms transactions, for denials, and as occurrences per year for trace requests.
A copy of the questionnaire is available on request.
The survey was implemented followed procedures developed by Dillman and colleagues20
and used by the author in two prior surveys of individuals.23,24
Modifications recommended by Dillman and colleagues20
for establishment surveys were incorporated. The survey was conducted by mail beginning June 16, 2011. NICS check requests are least frequent during the summer,25
and the survey was conducted during this time of relatively slow licensee business activity to improve the response rate.
The survey design required up to three mailings of the questionnaire, with a reminder postcard sent to all subjects between the first and second questionnaire mailings. Taking the mailing date of the first questionnaire as day 0, the postcard was sent on day 7, and subsequent questionnaires were sent to nonrespondents on day 21 and day 42. A cash incentive—three uncirculated $1 bills—was included in the first mailing. Respondents were also offered the opportunity to request a copy of publications arising from the survey.
Each questionnaire mailing included a cover letter explaining that the purposes of the survey were “to understand better the unique perspective of firearms licensees on important social issues and the firearms business itself” and to collect “the first nationwide information on the day-to-day business experience of firearms licensees.”
Letters were addressed to the licensee when the licensee was a named person or to the “firearms manager” when the licensee was a corporation and were hand-signed by the author (as were the postcards). The letters included telephone numbers for the study team (a dedicated telephone line for queries and comments was established) and the university’s Institutional Review Board. All mailings were sent to premises addresses so as to be received by on-site managers rather than by staff at a centralized corporate headquarters. The first two questionnaires and the postcard were sent by first class mail, using commemorative stamps selected as appropriate to the project. The third questionnaire was sent by Priority Mail.
Messages left at the project’s telephone number were transcribed daily, and a log was kept. Return calls, when needed, were made by the author. Returned questionnaires were reviewed by the author within 1–2 days to identify unanticipated problems that could be corrected during survey execution. We did not attempt to contact respondents who refused to participate or indicated that they were prohibited from doing so.
Our monitoring established that the response rate for subjects affiliated with corporations with that held licenses for multiple business locations (chain stores) was lower than that for others. We sent personalized letters to the chief executive officer or chief regulatory officer of each of the 25 corporations with more than one licensee in our study sample, requesting that they authorize their store managers to participate. The author made multiple attempts to contact each corporate officer by telephone to follow up.
We also established procedures to detect and monitor discussion of the survey on the Internet. Our primary interest was in any attempt to discourage subjects from participating or to encourage collective or strategic responding. We performed searches daily on an array of relevant keywords and phrases, beginning 1 week before the first questionnaire mailing to establish a baseline level of activity.
Data Management and Statistical Analysis
We performed data entry when questionnaires were received, using dual-entry procedures and automated and manual comparisons.
We determined response and refusal rates and assessed questionnaire completeness using guidelines developed by the American Association for Public Opinion Research.26
The response rate was calculated as the percentage of subjects in the sample who returned filled-out questionnaires. Complete questionnaires provided answers to >80 % of questions, partial questionnaires to 50 %–80 %, and break-off questionnaires to <50 %.
Respondents were classified by the number of questionnaires they were likely to have received prior to responding: early, response received by us no more than 2 days after the second questionnaire mailing; intermediate, response received more than 2 days after the second questionnaire mailing and no more than 2 days after the third; late, response received more than 2 days after the third questionnaire mailing.
All licensees in the sample were categorized by general business structure: The licensee was an individual named person; the licensee was a corporation, and only one establishment owned by that corporation appeared in the sample (corporate/single site); the licensee was a corporation, and multiple establishments owned by that corporation appeared in the sample (corporate/multi-site). Establishment locations were categorized by Census region.
Respondents who were known not to be owners, managers, or other senior executives (n
21) or whose status could not be determined (n
27) were excluded from analyses of respondents’ personal characteristics and attitudes toward firearms and the firearms business.
Continuous variables were stratified into quartiles for most analyses, to minimize effects due to outliers and clustering of responses at numbers ending in zero. The strata were chosen to produce quartiles of approximately equal size for all respondents considered together. Responses on the percentage of handguns sold that were priced at more than $250 were recoded for analysis to the percentage priced at $250 or less. The reported average annual number of trace requests was converted to a percentage of firearm sales in 2010. Strata for the two questions addressing general attitudes about firearms and crime were collapsed from five to three.
Most analyses relied on simple descriptive measures, with the significance of differences assessed using the Pearson Chi-squared test, or the Mantel–Haenszel Chi-squared test when a test for a linear association was desired. P<
0.05 was taken as the threshold for statistical significance. All analyses were performed using SAS version 9.1.3 for Windows.19
The UC Davis institutional review board approved this project.