Missouri Firearm Attitudes and Behavior Study (MFABS)

Summary:  View help for Summary This study examines the prevalence, contextual factors, beliefs, and perceptions surrounding firearm behaviors (e.g., ownership by gun type, safe storage, training) among a randomly selected sample of households in Missouri.In addition, the survey identifies perceptions and attitudes about both firearm behaviors and healthcare-based firearm safety, as well as topics covering perceptions of mental health,suicide, violence, and more. The study includes both gun owners and non-owners, as well as people whodo not own guns but live in homes where others do. Results will inform firearm safety programs and future studies examining firearm safety.

Subject Terms:  View help for Subject Terms Firearm attitudes; Firearm behaviors; Firearm Injury Risk

Geographic Coverage:  View help for Geographic Coverage Missouri, USA

Time Period(s):  View help for Time Period(s) 5/10/2023 – 7/12/2023

Collection Date(s):  View help for Collection Date(s) 5/10/2023 – 7/12/2023

Collection Notes:  View help for Collection Notes Participantswho responded online entered the web survey platform using a unique survey access code sent in the mail invitation. On average, online respondents took 29 minutes to complete the survey. Gunowners who were asked more survey items on average took longer (37.5 minutes) than respondents who did not report having a gun (24minutes). Respondents who lived alone and did not own a gun had a shorter survey experience covering topics such as their opinions on gun legislation and safety, general behavior and attitude items, and demographics. Respondentswho owned different types of guns (longguns and handguns), who lived in households where other people owned guns, and had kids,were asked to complete more survey items. At minimum, the survey contained about 130 items that all participants were asked to complete. Participants who had situations that led to more survey items received up to 120 additional questions, for a maximum of about 250 survey items.

Response Rate:  View help for Response Rate Response rates were calculated following the guidelines established by the American Association for PublicOpinion Research 10th edition Standard Definitions.

Initial Sample:18,000
Removed (Bad addresses, RTS):1,636
Total invites successfully sent:16,364
Survey returns (before cleaning):2,439
Response Rate: (2,439/16,364):14.9% (AAPOR response rate 2)

Cases cleaned out (see: Cleaning):590
Final cases in the data file:1,849
Completion rate: (1,849/16,364):11.3% (AAPOR response rate 1)

Sampling:  View help for Sampling All participants were recruited by Address Based Sampling (ABS) methodology. All residential addresses in Missouri were available to be sampled and selected at random. Gallup selects households at random, and because all households in Missouri have an equal and known probability for selection, the ABS methodology produces a representative sample of Missouri households. Selection from within the household was done at time of taking either the web survey online or the paper survey mailed to sampled homes by within-household self-selection of the adult in the home who had the next birthday. For this study, an initial sample of 18,000 households in Missouri was drawn.

Data Source:  View help for Data Source The original list of survey items was developed by the University of Michigan research team. Many of the survey items were selected from existing prevention and safety studies, including the previous University of Michigan/Gallup FACTS study. To fill gaps in the topics covered by the survey, some custom items were developed in partnership between Gallup survey scientist and the University of Michigan team. In total, the survey contained up to 250 survey items; 180 ofthese items are also contained in the 2023 National Firearm Attitudes and Behaviors Study (NFABS). This study contains all items from the 2023 NFABS withan addition ofup to 70 survey items, depending on the respondent’ssituation. Drawing survey items from multiple sources can lead to unintended and harmful inconsistencies in item measurement. The final survey standardized items across different measurement sources to ensure a seamless and less burdensome experience for respondents and limit measurement error. Order and context effects were also considered when ordering the survey to limit bias and error

Collection Mode(s):  View help for Collection Mode(s) mail questionnaire; web-based survey

Weights:  View help for Weights Base weights: Gallup constructed base sampling weight by considering the selection probability of the study. Post-stratification weights: Due to reasons such as nonresponse, the demographic distributions of respondents in the unweighted dataset can be different from their corresponding distributions among the adult population (aged 18 or older) in Missouri. For example, about 5.82% of the adult population in Missouri is Black only and Female, according to the American Community Survey five-year estimates 2017-2021, while there are 3.57% respondents identifying themselves as Black only and Female in the sample. (See the comparison table below for details.) To improve the relationship between the sample and the target population, Gallup further calibrated weights to match the sample to known population targets for age by race, gender by race, education by race, ethnicity, parents, and gun ownership using raking. Raking is an iterative process that adjusts the base weights of the cases in the sample so the marginal totals of the adjusted weights, on specified characteristics, agree with the corresponding totals for the population. Considering disparities in age, gender, and education levels among different racial groups, Gallup used a dual-target approach, considering age by race, gender by race, and education by race factors in our weighting process.This approach was designed to ensure that not only the overall age, gender and education distribution in the sample aligns with the target population among adults aged 18 and older, but also that it corresponds closely within each specific racial category, including white only, black only, and others.The post-stratification weights accounted for age by race, gender by race, education by race, ethnicity, parental status,and gun ownership. The targets for house hold gun ownership were based on RAND Corporation 2020 estimates, while the target for the others were all based on American Community Survey five-year estimates 2017-20216. The weights obtained through raking may however exhibit considerable variability, with some sampling units having extremely low or high weights relative to most of the other sampling units. It can lead to inflated sampling variance of the survey estimates. To resolve the issue, the weights obtained through the raking step were trimmed within each race group to avoid extreme weights. Trimming points were selected based on the distribution of post-stratification weights within each race group. The trimmed weights were then normalized to make sure they sum up to the number of completed interviews. With the above weighting approach applied, the differences in demographic distributions between the sample and the target population decrease, making the sample better representative of the adult population in Missouri. Weighting Variables in University of Michigan National Community Safety Study Dataset:
•Age (group) by Race(white only, black only, and the rest)
•Sex by Race (white only, black only, and the rest)
•Education by Race (white only, black only, and the rest)
•Ethnicity (Hispanic vs. non-Hispanic)
•Parent %
•Household gun ownership

The weight variable should be used when analyzing data and will generate state-level representative estimates. This weight should also be used to generate estimates when analyzing sub-groups, such as individuals who are Black, parents, or gun owners. Although the weighting procedures used were designed to improve the estimates of key subgroups (such as parents or gun owners), it is important to note that weighting cannot perfectly eliminate bias for all potential subgroups. This is primarily because reliable population targets are not available for all potential sub-groups that can be considered in analysis (such as gun-owning parents).

Unit(s) of Observation:  View help for Unit(s) of Observation Individual Respondents