How Americans encounter guns: Mixed methods content analysis of YouTube and internet search data

1. Introduction

Firearm-related injury and death are a major public health problem, with more than 40,000 deaths and an estimated 85,000-130,000 injuries a year (CDC, 2021). Firearm injury is now the leading cause of death for American youth age 1-19, the most common mechanism of death from intimate partner violence, and the most common mechanism of suicide death (Andrews et al., 2022; CDC, 2021). Young Black men are disproportionately more likely to be victims of firearm homicide, while older White men are at highest risk of firearm suicide (Kaufman & Richmond, 2020; Riddell et al., 2018). Rising rates of firearm injury and death since 2020 have been accompanied by a reported increase in the proportion of Americans who own a firearm (Gun Violence Archive, n.d.-b; Kaufman et al., 2021). Prior to COVID-19, 40% of Americans were firearm owners, and preliminary data suggests that another 2.9% became first-time owners during the first year of the pandemic (Anestis et al., 2021; Miller et al., 2022).

Shifting norms and behaviors regarding firearm “safety” and “harm” may be among the reasons for these increases in American firearm ownership, injury, and death. Subjective beliefs about the risk of firearm harm, and perceived ability to control that harm, are informed by factors ranging from in-person social networks to the media. Indeed, extensive evidence supports that one’s likelihood of firearm injury depends on patterns in one’s in-person social network (Papachristos & Wildeman, 2014). Qualitative studies of small groups of firearm owners suggest that so-called “gun culture” influences firearm ownership choices, firearm safety behaviors and firearm injury risk (Thomas et al., 2022; Yamane, 2017; Yamane et al., 2020, 2021) - although the determinants of this culture remain poorly defined. Unfortunately, data from firearm retailers suggest that the majority of new firearm owners did not receive basic safety training, despite ample evidence that consistent safe storage and safe handling are key to prevention (NSSF, 2022; Rowhani-Rahbar et al., 2016). Other preliminary work suggests that misleading and selective reporting by the media about firearm injury may lead to the American public making inaccurate judgements about the most common firearm injury risks (Jetter & Walker, 2018; Kaufman et al., 2020) and the availability of prevention strategies (Pallin et al., 2021).

As more Americans rely on online and social networking sites for news and information, it is increasingly likely that these spaces - and the words and narratives included in them - are an important driver of Americans’ current understanding of firearm injury, safety, and injury prevention (Betz et al., 2021; Walker & Matsha, 2021). For example, Twitter conversations in Latin America have been demonstrated to over-represent the likelihood of violent crime, and to represent societal fears more than actual prevalence (Prieto Curiel et al., 2020). Specific types of content of Twitter conversations in Detroit and Chicago have been shown to drive community conceptions and patterns of firearm injury (Patton et al., 2017; Stuart et al., 2020). Similarly, qualitative analysis of comments about online firearm-injury-prevention news articles have been used to identify areas of agreement and tension between health professionals and firearm owners (Knoepke et al., 2017). To our knowledge, no one has used large online media datasets to empirically examine the spaces and ways in which a diverse sample of the American public encounters narratives about guns.

As a first step to understanding the nature of firearm-related content to which consumers are exposed in social media, we aimed to define the nature, prevalence, and risk-related firearm content on YouTube. We chose YouTube as it is the most commonly consumed online social media platform across demographic groups in the United States (Auxier & Anderson, 2021), and as proprietary, anonymized databases of YouTube watching data are available.

Specifically, the goals of this analysis were to, first, characterize distinct online firearm-related narrative typologies on YouTube; second, describe the demographic characteristics of subpopulations that consume content within each typology on YouTube; and third, understand how these distinct audiences interact with firearm injury-related narratives in internet searches.

2. Methods

3. Results

Between December 1, 2020 and August 31, 2021, the opt-in panel included 2,111,309 person-days (measuring behavior of a total of 72,205 distinct American adults) of identifiable videos consumed on YouTube. The analysis considered descriptions for 16,803,075 person-video encounters with 7,274,093 distinct videos. Of these, 282,419 videos were identified as related to firearms using our search terms. On an average day, 7% of people in the panel who watched any YouTube content consumed some content relevant to firearms.

Partitioning the network of user and channel interactions yielded 7 ecosystems that had a reach greater than 1 per 1000 YouTube users per day (see Table 1 & Figure 1). An ecosystem might contain channels that have high-reaching content unrelated to guns (i.e., Movies or Music), but a lower reach of gun-specific content. These ecosystems were qualitatively described as:

1. Hunting & Fishing (characterized by celebration of firearms and sport, treating firearms as both a hobby and a pedagogy). The stories in this ecosystem tend to feature air rifles specifically and are about shooting for sport or shooting animals, but not shooting people.

2. Gaming (online or video-gaming). a highly social and exclusively virtual ecosystem. While it includes discussions of technical specifications of firearms, this gaming community places more emphasis on firearms being used to connect with others through shared enjoyment.

3. Movies (representative of mainstream imagination and big-screen narratives). Much of the content in this ecosystem comes from Movieclips; it is all creative and imaginative imagery.

4. Guns & Gear (featuring guns themselves as the subjects). All of the videos in this ecosystem are explicitly about firearms, and the ecosystem’s language is characterized by notions of “self-defense,” safety, and responsibility. The ecosystem’s videos’ goals are explicitly about shooting (lethally or non-lethally) other people.

5. Guns 4 Fun (recreational and simulation real-life “battle” games, such as Nerf gun wars). This ecosystem mostly relates to mock guns, including military simulation airsoft games and Nerf games simulating Fortnite gaming battles. The tone of this ecosystem is playful and fun. These creators celebrate their weaponry but clearly know the weapons are pretend.

6. Music (primarily metaphorical use of gun imagery and words). While many of the gun references in popular music are highly metaphorical or referential, much of the top content in this ecosystem centers real gun violence either from the perspective of the victim of violence.

7. News & Hot Takes (discussions of current events). The tone of gun-related content in News & Hot Takes is one of danger paired with combative arguments between partisan extremists. Videos move from recent reports of national shootings on local news to angry political rants for and against gun control. Language about safety, responsibility, or enjoyment is largely missing.

• Download figure
• Open in new tab

Figure 1:

The average daily reach across the seven naturally occurring firearm narrative ecosystems on YouTube.

Demographic differences were observed between the users of the ecosystems. For example, with the exception of the Music ecosystem, consumers of all “gun” ecosystems were disproportionately male compared with the average YouTube user. The “Hunting and Fishing” and “Guns and Gear” ecosystems were more often White Non-Hispanic than the average YouTube user (70% and 81%, respectively); these ecosystems were also older than the average YouTube user. In comparison, the “Guns 4 Fun” and “Music” ecosystems were more often Black, and the “Guns 4 Fun” and “Gaming” ecosystems were more often age less than 30 (42% and 62%). (See Table 2.)

Examining the internet searches used by all internet users and by people in the seven content ecosystems, we identified 103,727 internet searches conducted by panel members during this period.

Using the methods described above to identify firearm-related searches, after initial coding, nine of 250 potential firearm-related clusters were deemed irrelevant to firearm harm or injury (e.g., things like “revolver news”). Forty-five were clearly and exclusively related to firearm harm and injury; 193 were about firearms, but unrelated to harms; and three were ambiguously related to firearm harms (including one that contained searches about both threatening people with guns and also using firearms in biathlon events). For instance, the six examples examined for one cluster were:

Since both the “good” and “bad” examples of this category were relevant to firearm harm and injury, we were confident that the entire cluster was relevant.

On average, 0.5% of panelists searched for something related to firearms on any given day. Of those, ∼4% searched for anything related to firearm policies (e.g., “gun control” or “laws”) and ∼3% searched for anything related to firearm safety (e.g., “safes” “training”). The most common categories of firearm related searches were information about firearm-related injuries or harms (e.g “active shooter”), of which 96% were about mass shooting events or police-involved shootings, and virtually none were about more common kinds of firearm harm such as suicide or unintentional injuries.

Different search terms mapped differentially onto the ecosystems. See Figure 2 for specific words commonly contained in firearm-related searches associated with each ecosystem: for example, searches for “lyric” and “song” were contained in larger searches for specific songs about guns and violence, and “grain” was used in searches that refer to gunpowder specifications.

• Download figure
• Open in new tab

Figure 2:

Spider diagram of common words contained in firearm-related searches associated with each firearm ecosystem. Words closer to ecosystem names were searched more often by same-day viewers of those ecosystems. These words are selected based on both frequency and distinctiveness with respect to the topics.

4. Discussion

By examining national YouTube video consumption data using a mixed-methods social network approach, we surfaced seven distinct naturally-occurring online ecosystems in which consumers are exposed to a variety of narratives about firearm use and safety. To our knowledge, this mixed-methods analysis is the first to provide data-driven insights into types of firearm-related narratives presented on YouTube, the demographic characteristics of groups who consume these narratives, and related internet search patterns. Our findings highlight the dynamic and pervasive nature of online social network discussions of and narratives about firearms. It also demonstrates gaps in, and the potential value of engaging the social media universe as part of efforts to improve on, Americans’ knowledge about and prevention of firearm harm.

Our study highlights that online social narratives - including YouTube - should be part of the discussion about who and how Americans learn about and conceive of firearms and firearm safety. According to national data, YouTube is used by 54% of its American adult users each day (Auxier & Anderson, 2021). In our analysis, nearly 7% of panelists encountered some firearm-related YouTube content each day. In other words, a substantial portion of American adults are likely exposed to online social media firearm narratives on a daily basis. The salience of narrative in influencing beliefs about, use of, and response to firearms and firearm injury have been amply demonstrated by others (McGinty et al., 2013; Yamane et al., 2020)(Gerbner & Gross, 1976; Iyengar, 1991). The importance of trusted messengers in conveying accurate information about firearm safety, harm, and injury prevention is also frequently discussed (Anestis et al., 2021; Betz et al., 2021). Our findings suggest that YouTube is a potentially important source of trusted messengers.

This study also provides a first glimpse into the wide diversity of narratives about firearms on YouTube. Quantitative and qualitative data on pre-COVID firearm ownership suggest there is more than one “gun culture” among firearm owners (Betz et al., 2021; Boine et al., 2020; Yamane, 2017). Our qualitative and quantitative examination of YouTube data and corresponding internet searches demonstrates that online cultural narratives around firearms, in general, are even more diverse than those suggested through studies of firearm owners - whether in reach, narrative content, and user demographics. Each of the seven narrative ecosystems had its own stories, users, and corresponding searches on firearm safety and harm.

For example, the small and self-contained “Hunting & Fishing” and “Guns & Gear” ecosystems - which were most focused on the actual use of firearms, largely for self-protection or as a hobby - were primarily watched by older, white men. Internet searches for firearm safety information were most commonly seen among panelists affiliated with these ecosystems; but, even here, discussions of the prevalence of, or risk factors for, the most common kinds of firearm harm were sparse. The demographics of these two ecosystems (and related searches) do not, of course, reflect the most quickly growing groups of gun owners - women and minorities (Miller et al., 2022; NSSF, 2022). In other words, although some firearm safety information is present, the spaces in which it shows up are limited and less diverse.

Indeed, despite popular attention to the issue of “misinformation” (Aspen Institute, 2021; Office of the Surgeon General, 2021), our analysis (while not focused specifically on identifying misinformation) suggests that the bigger issue in firearm injury prevention is absence of information in online narratives about firearms (Golebiewski & boyd, 2019). Throughout all seven ecosystems, we identified few if any spontaneous stories, much less corresponding internet searches, about either the ways in which guns harm people most commonly (i.e., suicide, unintentional injury, and partner violence) or how to maintain safety for one’s self and one’s family (outside of the self-protection narrative). Indeed, in the firearm narrative ecosystems frequented primarily by women, non-White, or younger YouTube users (i.e., “Games”, “Music,” “Guns 4 Fun”), searches and stories about safety were virtually non-existent. Future work should further describe the different kinds of firearm safety narratives and the sentiment and trust of those who provide those stories. Americans are simply not exposed to the personal, localized burden of, or risks for, firearm injury that dominate actual injury patterns. Our work therefore suggests that adding to existing narratives - rather than seeking to replace or out-shout them - may be a fruitful avenue for future work; it also reinforces the importance of tailoring to different ecosystems’ audiences.

For the majority of the YouTube ecosystems, firearms are incorporated into daily definitions of fun, social connection, and leisure, and not into politics. Although others’ work suggests that news and creative media have created a polarizing narrative about firearms (Bonnevie et al., 2020; Dada et al., 2022; Quinn & Andrasik, 2021), in our analysis, polarizing narratives were limited to only a few ecosystems, particularly “News & Hot Takes”. Misleading and selective reporting about firearm injury certainly begets more firearm injury (Beale, 2006; Garland, 2000; Jetter & Walker, 2018; Kaufman et al., 2020). Given the influence of the “New & Hot Takes” ecosystem, it’s likely that it drives the common misconception of firearm-related harms as random, unpreventable attacks by strangers (Anestis et al., 2021; Getman et al., 2018; Ye et al., 2022). This analysis emphasizes, however, that a focus on the extremes ignores the ways in which most Americans encounter, play with, and discuss real or imagined firearms (Downs, 2002).

Finally and most importantly, ample work on other public health topics - ranging from suicide prevention and bullying prevention (Uhls et al., 2021), to tobacco and e-cigarette use (Majmundar et al., 2021; Sargent et al., 2003), to vaccine uptake (Getman et al., 2018), to sexual harrassment (Huseth-Zosel et al., 2021), to firearm injury itself (Ojo et al., 2021) - demonstrates that cultural cognitions and behavior can be changed by combining data and narrative in the online world. In work on other public health problems, we see the importance of moving beyond the trope that one narrative is right, and another is wrong. Rather, awareness of the various narratives allows incorporation of facts - around actual patterns of injury and ways to protect from injury - into these online communities. With expanded qualitative and quantitative understanding of the narratives that specific sub-groups interact with, we can begin to develop interventions to shift norms, beliefs, and self-efficacy, in ways that are tailored to both ecosystem content and user group.

Limitations of this study include potential biases in the YouTube and internet search data, reflecting the short time-window, the single source of social media data, and the limited numbers of internet search providers and internet searches. Although the proprietary panel is supposed to reflect a diversity of American adults, it is possible that the user participation is biased in unknown ways. These unmeasured biases reinforce that this is a preliminary study, and should be reproduced with a larger number of social media sources and a wider array of internet searches. A further limitation is that although the study combined rigorous quantitative and qualitative methods, more nuanced inquiry into meaning and sentiment is needed; for example, advanced machine learning sentiment analysis could be conducted to better quantify the types and meaning of internet searches within the various ecosystems. Finally, and most importantly, this study was purely observational, with no measurement of actual behaviors or beliefs, no claim of causality between exposure and outcomes, and no intervention delivered.

5. Conclusion

In this study, we combined content analysis of online social networks with the search patterns of a representative sample of users to not only identify distinct narratives about firearms, but to also understand which segments of the population engaged with these narratives. Our results highlight the degree to which firearms are present in, and influence, Americans’ public and private life, in areas ranging from hobbies to entertainment to news and social events. Our results also highlight the diversity of people and experiences that are part of this multiplicity of narratives. As Americans experience firearms on YouTube, online searches and associated results, and then circle back again, a complex understanding of “firearms” and “firearm injury” is created. Future work should further define the types of narrative and information to which people are exposed, the demographic differences in narrative exposure, and how narratives of harm and safety can be incorporated into these pervasive online narrative ecosystems.