Risk on Demand? A Quantitative Content Analysis of the Portrayal of Risky Health Behaviors in Popular on Demand Content

ABSTRACT

Video on Demand (VOD) has become the most popular way for adolescent viewers to consume entertainment media, often without parental supervision. Given the potential for modeling, this study aims to investigate the prevalence and nature with which risky health behaviors are portrayed in popular VOD programs. A quantitative content analysis of trending programs (N = 529) from popular VOD-platforms investigated the prevalence, co-occurrence, tone, social context, and consequences with which alcohol use, tobacco use, drug use, unsafe sexual behavior, reckless behavior, and self-harm behaviors are portrayed in popular VOD programs. In addition, we analyzed the demographic characteristics of the characters who portrayed the risk behaviors Risk behaviors were portrayed frequently, with substance use behaviors (i.e. alcohol, smoking, drugs) being most prevalent and most likely to co-occur. Reckless behavior, self-harm behaviors, and explicitly unsafe sexual behaviors were much less common. Findings show that risk behavior was often portrayed in a normalized manner, with alcohol and smoking, in particular, being portrayed as neutral behaviors that rarely have consequences. Most risk-taking characters were (young) adult white males, mirroring the general overrepresentation of this demographic in popular media. Risk behavior was rarely problematized in popular on demand content. Potential consequences for adolescent viewers are discussed.

Exposure to unhealthy and risky behaviors such as drinking, smoking, drug use, and reckless behavior is positively associated with real-life risk-taking, positive cognitions toward these behaviors, and related positive emotional states (Fischer et al., 2011; Hanewinkel et al., 2014). This is particularly worrisome for adolescent viewers as this group is already drawn to risk-taking and particularly susceptible to social influences (Ciranka & van den Bos, 2021). Within this context, entertainment media portraying risk behavior have been termed a “super peer” (Borzekowski & Strasburger, 2008; Elmore et al., 2017). Similar to actual peers, they have the potential to present adolescents with normative information on risk behavior (e.g., regarding prevalence or acceptability) that they might not otherwise come into contact with. This may result in misperceptions regarding social norms surrounding risk behavior, for instance overestimating the real-life prevalence of such behaviors (Borzekowski & Strasburger, 2008; Elmore et al., 2017; Sadza et al., 2022). Indeed, content analyses of entertainment media popular among adolescents reveal that risk behaviors are prevalent in such content (Bleakley et al., 2014; Ellithorpe et al., 2017) and often portrayed positively with a lack of focus on negative consequences (Mayrhofer & Matthes, 2021; Stern & Morr, 2013).

Previous content analyses of popular media have mainly looked at prime-time broadcasts, audience ratings, or box office sales as indicators of the popularity of entertainment media among young people. However, an increasingly popular subdomain of adolescents’ media environment is that of video-on-demand (VOD) platforms such as YouTube and Netflix, these being the most popular platforms for adolescents to consume media content at the time this study was conducted (Nielsen, 2016; Ofcom, 2018). These platforms allow adolescents to watch whatever, wherever, on any portable device. Notably, adolescents often do so without parental supervision as the accessibility VOD platforms via mobile devices has made it more difficult for parents to regulate what their children see (NICAM, 2021).

When it comes to selecting content on these platforms, recommender systems that present the viewer with lists of content that are “trending” or “popular” play an important role (Frey, 2021; Patch, 2018). The question however remains what kinds of risky content adolescents may see when they view content offered to them through these lists. Only a handful of content analyses have focused on risk behavior on VOD platforms, and these have been specifically focused on alcohol and tobacco (e.g., Alfayad et al., 2022; Giannakodimos et al., 2022) with other risk behaviors remaining underexplored. Furthermore, these studies have focused mostly on prevalence rates while the nature in which these behaviors are portrayed (who engages in them, what are the consequences, etc.), remains more underexplored. Finally, little is known about the portrayal of risk in content that is defined by these platforms as trending or popular, and therefore displayed prominently and actively recommended to (young) viewers. Our aim is therefore to provide a complete overview of the prevalence and nature with which risk behaviors (i.e., alcohol use, drug use, smoking, unsafe sex, deliberate self-harm, and general reckless behavior) are portrayed in popular and/or trending video-on-demand content. We look at these specific behaviors as well as their co-occurrences as these risk behaviors are prevalent among adolescents (Kipping et al., 2012; Rombouts et al., 2020) and because co-occurrence is prevalent both within substance use as well as between substance use and other risk behaviors (Hale & Viner, 2016; Van Nieuwenhuijzen et al., 2009), including sexual risk behavior and self-harm (Portzky et al., 2008; Wright et al., 2020).

Risk behavior prevalence and co-occurrence

In line with cultivation theory (Gerbner et al., 2002), studies have found that exposure to high incidence rates of risk behavior such as substance use, unsafe sexual behavior, or risky driving in entertainment media may contribute to the overestimation of real-life prevalence rates for these behaviors (Borzekowski & Strasburger, 2008; Elmore et al., 2017; Fischer et al., 2011). Several studies have therefore investigated the prevalence with which these behaviors occur in entertainment media, often within specific media formats or genres such as reality tv (e.g., Flynn et al., 2015), music videos (e.g., Cranwell et al., 2017), soap operas (e.g., Barker et al., 2021), teen movies (e.g., Stern & Morr, 2013) and content from video-on-demand platforms (e.g., Alfayad et al., 2022; Giannakodimos et al., 2022). For both on-demand and more traditional entertainment media channels, these studies most often focus on substance use and consistently find high prevalence rates, particularly for alcohol (Barker et al., 2020, 2021). For other risk behaviors, less is known about their prevalence on VOD platforms. Sexual content has mainly been studied in (non-VOD) movies, music, television, and magazines, with studies showing high prevalence rates (Ward, 2016). The portrayal of other risk behaviors such as risky driving and intentional self-harm is studied less frequently. However, studies focusing on the portrayal of these behaviors in movies also indicate that their portrayals are common and often prominent (Beullens et al., 2011; Jamieson & Romer, 2011). In sum, while prior research indicates risk behaviors are prevalent in the majority of entertainment media, relatively little is known about how common these behaviors are in trending content on video-on-demand platforms and how prevalence rates vary between various risk behaviors within this content.

Several of the aforementioned studies focus on more than one risk behavior, most commonly multiple substance use behaviors, such as alcohol and tobacco (e.g., Barker et al., 2020, 2021; Giannakodimos et al., 2022) or alcohol, tobacco, and drugs (e.g., Stern & Morr, 2013). Some look at substance use but also take other risk behaviors into account, most often violence and sex (e.g., Bleakley et al., 2014, 2018; Ellithorpe et al., 2017; Thrasher et al., 2014). However, only a few studies have looked in detail at the co-occurrence of these various risk behaviors, in other words at their combined portrayals and correlations. Those studies that have taken this into account (e.g., Flynn et al., 2015; Bleakley et al., 2014, 2018) show that characters who engage in one type of risk behavior were also more likely to engage in other risk behaviors.

Taken together, these studies show that the (co-)occurrence of multiple risk behaviors is common in popular programming. However, most studies focused on specific characters and/or genres such as reality television (Flynn et al., 2015), violent characters in movies popular among adolescents (Bleakley et al., 2014) and television shows popular with Black adolescents (Bleakley et al., 2018). Specific co-occurrences of various types of risk behavior have not yet been studied for the broader range of entertainment media, and video-on-demand programming in particular. Furthermore, co-occurrences with risk behaviors other than substance use and sex (e.g., reckless behavior, intentional self-harm) remain understudied. Research has outlined concerns about how these behaviors are portrayed due to the idea of mediated modeling but also suggested that these portrayals might be beneficial in reducing stigmas around these risk behaviors and can positively influence help-seeking behavior depending on how they are portrayed (Trewavas et al., 2010). While behaviors such as self-harm are prevalent with up to 20% of adolescents engaging in them (Doyle et al., 2015; Plener et al., 2015), their portrayal in entertainment media remains largely understudied. This study aims to add to this gap in the literature by looking at the prevalence and co-occurrences of a wide array of risk behaviors.

How and by whom is risk behavior portrayed

Besides potentially contributing to overestimations of the prevalence of real-life risk behavior, viewing media portrayals of risk behavior may also contribute to adolescent risk behavior through the process of modeling. Social cognitive theory (Bandura, 2001) suggests behavior may be learned through the process of observational learning. Viewers may observe media models who engage in risk behavior and see what potential consequences this results in, which may or may not result in the modeling of said behavior. Indeed, studies have found exposure to risk behavior in entertainment media to be related to real-life attitudes toward and engagement in these behaviors (Bleakley et al., 2017; Fischer et al., 2011). However, it is important to note that this is not an automatic process. Rather, viewers actively evaluate and interpret media portrayals of risk behaviors and the outcomes of this process determine to what extent these portrayals are internalized and/or modeled (Austin et al., 2000; Elmore et al., 2017).

Several factors relating to how the behavior is portrayed may also impact the chance of behaviors being modeled by viewers. Firstly, social learning theory states that people are more inclined to model behavior when the behavior is rewarded and leads to positive rather than negative consequences, due to a change in outcome expectancies (Austin et al., 2000). Indeed, empirical research (e.g., De Graaf, 2013) has found that seeing alcohol portrayals in which negative consequences were present led to more negative beliefs about alcohol use. Prior studies investigating the consequences with which risk behaviors are portrayed have typically looked at substance use behaviors and found that negative consequences were seldomly portrayed (e.g., Stern & Morr, 2013; Mayrhofer and Matthes, 2021). Again, this was typically focused on specific subgenres and behaviors and has not yet been studied for popular on-demand content. The question remains how positively, neutrally, or negatively, various risk behaviors are portrayed in trending on-demand content. This study will therefore investigate the tone and consequences with which risk behavior is portrayed across a wider range of behaviors and within a wide array of on-demand entertainment media programming.

Secondly, social learning theory as well as other theories regarding social norms (e.g., social norms approach, Berkowitz, 2005; theory of planned behavior; Ajzen, 1991) state that people are more likely to engage in behaviors that they perceive to be “the norm.” This is particularly true for adolescents for whom social norms regarding risk behaviors are one of the most important determinants for their engagement in risk behavior (Ciranka & van den Bos, 2021). We therefore also look at the social context in which the behavior is portrayed.

Finally, both social learning theory as well as theories on media effects that focus specifically on children and/or adolescents (e.g., message interpretation process model, Austin et al., 2000) state that viewers are more likely to model behaviors of characters with whom they identify or feel similar. We therefore also look at the demographic characteristics of those engaging in the behavior. Several studies have previously taken character demographics into account for particular behaviors, demographic groups, or (sub)genres (e.g., Ellithorpe et al., 2017; Flynn et al., 2015; Stern & Morr, 2013). Flynn et al. reported that men and young adults were more likely to engage in risk behavior as compared to women and teenagers and that ethnic minorities were represented as drinking more often than white cast members in MTV reality programming. Ellithorpe et al. looked at popular films and found that Black characters more often engaged in sex and alcohol use, whereas white characters more often in violence. In contrast to these studies, Stern and Morr’s study of popular teen movies found no demographic differences between characters who did or did not use alcohol, tobacco, or drugs. While these studies present relevant insights into how risk behavior is portrayed among specific demographic groups and within specific formats of programming, what remains missing is a more complete overview of how various demographic groups are represented among risk-taking characters across a broad range of risk behaviors and (on-demand) entertainment media programming.

Methods

A quantitative content analysis was conducted based on productions (videos, films, series, television shows – N = 529) that were classified as trending or popular on a newly created (teen) account by the most popular video-on-demand (VOD) services in The Netherlands in 2018. For each platform, we created an account and/or profile for a fictional teenaged (16-year-old) person. This age was used as input for a new e-mail address that was used to create the accounts and filled in if platforms asked for a users’ specific age. Next, we copied the trending and/or popular list of each platform to create our sample, incorporating a maximum of 100 productions per platform. Table 1 shows the distribution of productions per platform. The trending and popular lists were used because of their important role in content selection by (adolescent) viewers (Frey, 2021; Patch, 2018).

Table 1. Sample breakdown per VOD-Platform.

Platform	Netflix	Videoland	RTLXL	Kijk	NPO-gemist	Youtube	Total
Productions	N = 100	N = 98	N = 85	N = 97	N = 100	N = 49	N = 529

Coding procedures and codebook

The productions were coded on the program level and on the level of persons or characters that portrayed risk behaviors. First, prevalence, tone, and context for the various risk behaviors were coded on the program level. Next, engagement in specific risk behaviors and their consequences were coded for up to eight main characters (Weijers, 2014). The coding categories for the behaviors were adapted from previous research on risk behaviors and their consequences (Primack et al., 2015; Stern & Morr, 2013). For each of the included main characters, socio-demographic characteristics were coded. The coding categories for the characteristics were adapted from previous research (Daalmans et al., 2017; Koeman et al., 2007). Main characters who did not engage in risk behavior were not included.

Risk behavior prevalence

Variables were created indicating whether a character that engaged in risk behavior was involved in each of the selected risk behaviors – alcohol consumption, tobacco use, drug use, unsafe sex, reckless behavior, and intentional self-harm – at any point throughout the selected episodes. Each behavior was coded as occurring never, once, or more than once and later collapsed into dichotomous (present or not present) codes. Alcohol, tobacco, and drug use were each defined as either direct use (e.g., on-screen consumption of alcohol, inhaling of smoke, or use of drugs such as snorting or taking pills) or implied use (e.g., handling drinks, lit tobacco products, joints etc.).

When characters engaged in sexual acts (e.g., intercourse, oral sex) this was coded and coders then answered whether any clearly safe or unsafe sexual behaviors occurred, as well as sexual acts for which safety practices were unclear. Coders were instructed to code any sexual act that can potentially lead to STD’s, including any type of intercourse and oral sex, and excluding things like kissing or masturbation. Unsafe sexual acts were defined as any sexual acts (e.g. intercourse, oral sex) for which it was explicitly stated or visualized that no protection (i.e., condoms or other barrier methods) was used in the sexual interaction if the sexual interaction was not specifically geared toward the conception of children. Conversely, safe sexual acts were defined as explicitly stating or visualizing that protection was being used. Cases in which this remained unclear were coded as sexual acts for which safety remains unclear.

Reckless behavior was defined as deliberate reckless behavior (behavior posing risks to one’s own health) such as driving under the influence or performing dangerous stunts (e.g., train surfing). Finally, intentional self-harm behavior was defined as a character deliberately harming themselves, either through hurting (e.g., cutting, burning), risk-related eating behavior (e.g., binging, purging), or attempts to commit suicide.

Tone, social context, and consequences

On the program level, coders coded the tone with which the production portrayed each of the individual risk behaviors: mostly positive, neutral/mixed, or mostly negative. The tone for each risk behavior was judged for the program as a whole and could be judged based on explicit verbal statements as well as storylines. A positive tone was coded when a program clearly showed the risk behavior as something that is solely good or fun. A negative tone was coded when a program clearly showed the behavior as something that is bad or problematic, again this could be in the form of explicitly discussing the problematic nature of the behavior, or by only showing the behavior in a negative context. If the tone of a program was not clearly positive or negative the tone had to be coded as neutral(/mixed).

The social context in which the risk behavior occurred was also coded on the program level. Social context is here defined as whether behaviors were portrayed as being the social norm. Each risk behavior could be coded as “portrayed as something everyone partakes in (i.e., the social norm)” or “portrayed as an individual behavior” (i.e., the person engaging in the behavior was relatively unique in doing so, not the social norm). This was coded on the program level and as such a behavior portrayed as “social norm” would be portrayed by many people in the production; for instance, productions in which drinking is portrayed at parties with many people engaging in it. Conversely, a production in which a behavior was portrayed as an individual behavior rather than the social norm might portray a particular character as “the drug user,” and as being relatively unique in doing so.

On a character level, coders answered whether the risk behaviors main characters engaged in also resulted in consequences for that character. Positive and negative consequences were coded separately from one another and later collapsed into several categories that made clear what types of consequences the character’s risk behavior resulted in: only positive, only negative, both positive and negative or no consequences. Consequences had to be explicitly shown or referred to be scored.

Demographics

For the main characters engaging in risk behavior, we coded gender, age, and ethnicity. The coding categories for gender were “male,” “female,” and “other or unknown.” The coding category for the age of the main character reflected five life cycles (child (0–12), teenager (13–18), young adult (19–35), adult (36–64), senior (65+)), and the coders were instructed to establish the age of the character by determining which age group or life stage the character was supposed to represent, if this could not be determined this was also coded as such. Ethnicity could be coded as Caucasian/White, Black, Asian, Mediterranean Europe, Mediterranean Arabic, Latin American, Other/Mixed, or Unknown/Cannot be coded.

Coder training and reliability

The coding itself was part of a specialized seminar on the methodology of content analysis for Communication Science students in the spring of 2018. Thirty-seven coders (29 female and 8 male) and the first author took part in the data collection. Coder training took place over the course of four weeks for a total of 14 hours of in-class training on a variety of programs and videos that were not part of the final sample. Outside of these sessions, coders coded several more hours of material (also not part of the final sample) which was then discussed at the start of the next session to informally assess intercoder agreement (Lombard et al., 2010) before continuing further training. After these training sessions, coders were given subsets (10–15 programs) of the analytic sample to code individually. Around 15 percent (n = 82) of the programs in the sample were double-coded. Based on this overlap, the levels of inter-coder reliability were calculated using Krippendorff’s alpha (using the macro by Hayes) for all variables used in the analysis. Results for intercoder reliability can be found in Table 2. The reliability scores were good (> .80) for prevalence, tone, context, and consequences for alcohol, tobacco, drugs, self-harm, and reckless behavior. The variables related to unsafe sex tone and context resulted in low intercoder reliability scores (.54 and .36 respectively). With regard to demographics, reliability scores were good (> .80) for gender and age, and acceptable (.78) for ethnicity. The low intercoder reliability scores for the tone and context of unsafe sex are likely in part due to the low occurrence of this behavior. However, given the low-reliability scores and the very rare occurrence of these variables in both the double-coding and full dataset, we decided not to include them in the analyses. As such, reliability for all included variables was acceptable to good (> .78).

Table 2. Intercoder reliability expressed in Krippendorff’s Alpha.

	Occurrence	Tone	Social Context	Any Consequences	Positive consequences	Negative consequences
Alcohol consumption	.89	.94	.96	.93	.95	.95
Drugs use	.87	1	1	.87	1	1
Tobacco use	.86	1	.89	.94	1	1
Unsafe sex*	1	.54	.36	.90	1	1
Reckless behavior	.81	.90	.90	.91	1	1
Intentional self-harm	1	.92	1	1	1	1
Gender	.98	−	−	−	−	−
Age	.88	−	−	−	−	−
Ethnicity	.78	−	−	−	−	−

*To code unsafe sex coders answered whether sexual acts occurred (α = .91), followed by whether any sexual acts occurred that were safe (α = 0), unsafe (α = 1) or where it was unclear whether safe sex practices were used (α = .90). Tone, context, and consequences were only scored for unsafe sexual acts. Due to low intercoder reliability scores on these latter variables we chose not to include them in our analysis. General presence of sexual acts and of sexual acts where safety was unclear are discussed to offer context for unsafe sexual acts.

Analysis

Data were analyzed using statistical analysis software (SPSS). For prevalence, tone, context, and consequences we focused on the frequency of occurrence. To determine the pattern of results for co-occurrence and demographics we used cross-tabulations with Chi-square and Fisher’s exact test and report adjusted standardized residuals to further interpret patterns of over- or underrepresentation, as determined by the expected and observed frequencies.

Results

Prevalence and portrayal of risk behavior on the program level

Table 3 presents an overview of the prevalence and co-occurrence of risk behaviors on a program level. Of the programs analyzed, over half contained at least one risk behavior. Almost half of the programs contained characters engaging in alcohol use and a fifth contained characters engaging in tobacco use. The other risk behaviors were much less common.

Table 3. Prevalence of co-occurring risky health behaviors in programs containing other risk behaviors (%).

Risk behavior	Prevalence within programs containing						Prevalence within all programs
	Alcohol consumption	Drug use	Tobacco use	Unsafe sex	Reckless behavior	Intentional self-harm
Alcohol consumption	−	80.6^a	68.1^b	66.7	56	66.7	45.4
Drugs use	10.4^a	−	19.8^c	33.3	16	16.7	5.9
Tobacco use	32.9^b	74^c	−	33.3	36	33.3	21.9
Unsafe sex	0.8	3.2	0.9	−	4	0	0.6
Reckless behavior	5.8	12.9	7.8	33.3	−	0	4.7
Intentional self-harm	3.3	6.5	3.4	0	0	−	2.3
	N = 240	N = 31	N = 116	N = 3	N = 25	N = 12	N = 529

^aχ2(1) = 16.533, p = < .001, V = .177, Adjusted Residual = 4.1

^bχ2(1) = 30,985, p = < .001, V = .242, Adjusted Residual = 5.6

^cχ2(1) = 52,544, p = < .001, V = .315, Adjusted Residual = 7.2

The portrayal of multiple risk behaviors within one program was also less common than the portrayal of only one risk behavior. Of the programs analyzed, 35.5% contained only one risk behavior and 19.5% contained multiple risk behaviors. Looking at the co-occurrences of specific risk behaviors within programs, co-occurrence was most common among substance use behaviors.

Tone

Table 4 shows the tone with which the risk behaviors were portrayed on the program level. Alcohol consumption and smoking were mostly portrayed as neutral. The tone for drug use was divided more evenly, with half of the portrayals being neutral, one quarter being negative, and one-fifth positive. Reckless behavior was neutral in most cases, followed by negative and then positive portrayals. Finally, intentional self-harm behavior was somewhat evenly divided between neutral and negative portrayals. It was never portrayed in a mostly positive tone.

Table 4. Tone, social context, and consequences.

Portrayal	Alcohol consumption	Drug use	Tobacco use	Unsafe sex	Reckless behavior	Intentional self-harm
Tone
Mostly positive	19.2	19.4	1.7	−	16	0
Neutral	79.2	54.8	92.2	−	56	41.7
Mostly negative	1.7	25.8	6	−	26	58.3
Total	N = 240	N = 31	N = 116	−	N = 25	N = 12
Social Context
Individual	42.9	66.7	71.6	−	68	91.7
Social norm	57.1	33.3	28.4	−	32	8.3
Total	N = 240	N = 30	N = 116	−	N = 25	N = 12
Consequences
None	90	64.9	98.6	100	50	15.4
Only positive	4.4	2.7	0.5	0	17.9	7.7
Positive & negative	1.9	18.9	0	0	0	0
Only negative	3.7	13.5	1	0	32.1	76.9
Total	N = 641	N = 37	N = 209	N = 6	N = 28	N = 13

Tone and context are coded and presented on the program level, consequences are coded and presented on the character level.

Social context

Table 4 also depicts the social context in which risk behaviors were portrayed. As can be seen from the table, alcohol use was portrayed as the norm in more than half of programs containing alcohol use. Two-thirds of programs containing drug use portrayed this as individual behavior while one-third of the programs portrayed it as the norm. Smoking, reckless behavior, and intentional self-harm was presented as individual behavior in the majority of programs in which it occurred.

Prevalence and portrayal of risk behavior on the character level

808 characters were coded as engaging in risky health behavior: 79.3% consumed alcohol, 25.9% smoked cigarettes, 4.6% used drugs, 0.7% had explicitly unsafe sex, 3.5% engaged in reckless behavior, and 1.6% in intentional self-harm. Of these characters, 85.7% engaged in only one risk behavior, and 14.3% engaged in multiple risk behaviors.

As is clear from these numbers, the portrayal of explicitly unsafe sexual acts was extremely uncommon, with only 6 characters coded as engaging in this. A total of 108 characters were coded as engaging in sexual acts, but in 92.5% of cases (n = 100) the safety of these acts remained unclear.

Consequences

Table 4 depicts the consequences risk behaviors had for characters. Most characters engaging in alcohol use and tobacco use did not encounter any consequences for their behavior. Characters engaging in drug use mostly encountered no consequences, followed by mixed consequences and negative consequences, but rarely positive consequences. None of the characters engaging in explicitly unsafe sexual behavior encountered any consequences. Reckless behavior had no consequences for half of the characters engaging in it, only positive consequences for around one-fifth and only negative consequences for one-third of these characters. Intentional self-harm led to negative consequences in most cases.

Demographics

Risk-taking characters were mostly young adults or adult white males. Table 5 shows the demographic characteristics of characters engaging in specific risk behaviors as well as the overall demographic breakdown for all characters engaging in any kind of risk behavior.

Table 5. Demographics of persons and characters engaging in risky health behavior (%).

Demographics	Alcohol consumption	Drug use	Tobacco use	Unsafe sex	Reckless behavior	Intentional self-harm	Multiple risk behavior	Total/Any risk behavior
Gender
Man	54.9	67.6	62	50	64.3	46.2	63.5	56.3
Woman	45.1	32.4	38	50	35.7	53.8	36.5	43.7
Total	N = 641	N = 37	N = 208	N = 6	N = 28	N = 13	N = 115	N = 807
Chi^2/Fisher’s Exact	χ^2(1) = 2.286, p = .131, V = .053	χ^2(1) = 2.016, p = .156, V = .050	χ^2(1) = 3.780, p = .052, V = .068	Fisher’s Exact Test, p = 1, V = .011	χ^2(1) = 0.760, p = .383, V = .031	χ^2(1) = 0.548, p = .459 V = .026	χ^2(1) 2.841, p = .092, V = .059
Age
Teenager	2.8^b	5.4	5.4	0	11.5	38.5 ^a	2.6	4.6
Young Adult	57.1	67.6	61.4	100	61.5	30.8	73.9 ^a	56.3
Adult	35	24.3	29.7	0	19.2	30.8	20.9 ^b	34.4
Senior	5.1	2.7	3.5	0	7.7	0	2.6	4.8
Total	N = 632	N = 37	N = 202	N = 6	N = 26	N = 13	N = 115	N = 790
Chi^2/Fisher’s Exact	χ^2(3) = 21.429, p = < .001, V = .165	Fisher’s Exact Test, p = .474 V = .055	χ^2(3) = 4.397, p = .222, V = .075	Fisher’s Exact Test, p = .206, V = .077	Fisher’s Exact Test, p =. 103, V = .082	Fisher’s Exact Test, p = .001, V = .212	χ^2(3) = 16.946, p = .001, V = .146
Ethnicity
White, Caucasian	88.7	72.2^b	86.2.	100	88.9	92.3	91.1	87.2
Black	5.6	8.3	4.1	0	11.1	0	2.7	5.8
Asian	1.1	8.3^a	0.5	0	0	7.7	0.9	1.4
Mediterranean Arabic	2.3	11.1^a	4.6	0	0	0	3.6	2.8
Mediterranean Europe	0.8	0	1.5	0	0	0	0	1
Latino/Hispanic	1.5	0	3.1	0	0	0	1.8	1.7
Total	N = 620	N = 36	N = 195	N = 6	N = 27	N = 13	N = 112	N = 774
Fisher’s Exact	Fisher’s exact test, p = .067, V = .108	Fisher’s exact test, p = .004, V = .177	Fisher’s exact test, p = .088, V = .110	Fisher’s exact test, p = 1, V = .034	Fisher’s exact test, p = .781, V = .065	Fisher’s exact test, p = .456, V = .082	Fisher’s exact test, p = .568, V = .075

^acell count significantly higher than expected according to adjusted residuals.

^bcell count significantly lower than expected according to adjusted residuals.

For age, the results reveal a pattern in which young adults are overrepresented among most of the risk behaviors. While this overrepresentation was stable across risk behaviors, there were some significant differences. The age distribution of characters engaging in multiple risk behaviors differed significantly from that of characters engaging in only a single risk behavior. Young adults were overrepresented (Adjusted Residual: 4.1) and adults underrepresented (Adjusted Residual: −3.3) among characters engaging in multiple risk behaviors, as compared to characters who only engaged in a single type of risk behavior.

The results further reveal differences in age distribution among characters engaging in alcohol, intentional self-harm, and reckless behavior, particularly for teenagers. Across the sample, teenagers were underrepresented (Adjusted Residual: −4.6) amongst those engaging in alcohol use and overrepresented (Adjusted Residual: 5.9) amongst those engaging in self-harm behaviors.

In terms of ethnicity, most characters engaging in risk behavior were white. While the representation of ethnicities was similar across most risk behaviors, the representation of ethnicities among characters who used drugs differed significantly from that of those who only engaged in other kinds of risk behavior. For drug use, white characters make up the majority of these characters but their prevalence here is below expected frequencies (Adjusted Residual: −2.8). By contrast, characters of Mediterranean Arabic (Adjusted Residual: 3.6) or Asian ethnicity (Adjusted Residual: 3.1) made up a significantly larger part of drug-using characters, as compared to their prevalence among characters engaging in other types of risk behavior.

Discussion

This study aimed to provide an overview of how often, in which combination, in what manner, and by whom alcohol use, drug use, smoking, unsafe sex, intentional self-harm, and general reckless behavior are portrayed in popular video-on-demand content. Of the risk behaviors studied, substance use behaviors were especially prevalent, with alcohol occurring in almost half of all programs, followed by smoking and then drug use. Drug use was much rarer than alcohol and tobacco use but did occur relatively frequently in programs containing alcohol or smoking. Deliberate self-harm and reckless behaviors were rare, but the lowest frequency was explicitly unsafe sexual behavior. This was partly due to our operationalization. In contrast to prior research, we only included (explicitly) unsafe sex as a risk behavior, rather than all sexual conduct. This rarely occurred because in the vast majority of sexual acts safe sex practices were not referenced at all. While unsafe sex was not explicitly portrayed as typical or prevalent behavior, the lack of focus on safe sex practices is still concerning as safe sex practices were also explicitly not portrayed as being typical or prevalent. On the whole, the way in which sexual behavior was portrayed in trending VOD content may be seen as risky. The lack of focus on sexual safety may communicate to viewers that this is not something to concern themselves about, which may have serious implications.

Another concern is posed by the co-occurrence of risk behaviors, which was relatively common for the various substance use behaviors (i.e., alcohol, tobacco, and drugs). The prevalent co-occurrence of multiple risk behaviors is in line with prior research. Though previous studies (Bleakley et al., 2014; Flynn et al., 2015) found co-occurrences between substance use and sexual behavior, we only found significant associations between substance uses. This may be explained by the differences in risk behaviors studied (e.g., unsafe sex vs all sexual conduct) and the comparatively low frequency with which risk behaviors other than substance use occurred within our sample.

The co-occurrence of multiple substance use behaviors reflects societal trends as the use of several substances has been found to cluster (Busch et al., 2013). Polysubstance use is prevalent among young people but can have detrimental consequences for their health (Kecojevic et al., 2017). Despite the prevalence of substance use and polysubstance use in our sample, the portrayal of negative consequences of substance use was rare. However, while it is often said that entertainment media portray risk behavior in a positive and even glorified manner (Fischer et al., 2011) we found that most portrayals were fairly neutral. This is still worrisome when combined with the fact that risk behaviors are rarely portrayed as having negative consequences for those engaging in them. When analyzed through the lens of social learning theory (Bandura, 2001), the lack of a negative tone combined with the lack of negative consequences may normalize these behaviors. These findings are in line with prior studies looking at substance use and its consequences (Stern & Morr, 2013; Mayrhofer & Matthes, 2021; Primack et al., 2015).

The only behavior that was presented as explicitly negative and problematized was intentional self-harm. This behavior also differed from the other risk behaviors in terms of who engaged in it, with teenagers being relatively overrepresented. This reflects real-life trends as self-harm behaviors often start during adolescence and are prevalent within this age group (Doyle et al., 2015; Plener et al., 2015). Despite the real-life prevalence of this behavior among adolescents, the portrayal of intentional self-harm behavior in popular entertainment media was very rare, likely due to the taboo nature of this behavior (Igra & Irwin, 1996). This is further reflected in the fact that self-harm behavior was almost always portrayed as an individual behavior. In contrast to Trewavas et al. (2010) who specifically analyzed movies portraying self-harm behaviors, we found no association with substance use. This may however also be due to a lack of power because of the low prevalence of this behavior within our sample.

With regard to social context, most behaviors were presented as individual behavior rather than the social norm in the majority of cases. Drinking alcohol was the only behavior that was regularly presented as socially normative, likely due to its frequent portrayal in group and party settings. Interestingly, other substance use behaviors were not presented as such. For tobacco and drug use, users were more often portrayed as being the exception, and the behavior was portrayed as something individual. For tobacco, this may reflect changing societal norms (Thrasher et al., 2014). For drug use, similar to self-harm behaviors, this may reflect its more taboo nature as the tone with which this behavior was portrayed was also more negative than that of either alcohol or tobacco use. From the perspective of super peer theory (Borzekowski & Strasburger, 2008) this may be viewed as positive as this type of portrayal may be less likely to contribute to the perception of drug use as acceptable among others (i.e., injunctive norm). Alternatively, however, the portrayal of potentially detrimental behaviors such as drug use and self-harm in this manner may also further contribute to their taboo nature. This may potentially be problematic as it may make these behaviors more difficult to talk about for adolescents which could have negative implications for prevention and awareness efforts.

Finally, the demographic makeup of characters engaging in various risk behaviors was remarkably stable. Characters engaging in risk behavior were mostly (young) adult white males, which is unsurprising as this demographic group is generally overrepresented in media (Koeman et al., 2007) and in media portrayals of alcohol consumption (Flynn et al., 2015). The overrepresentation of young adult males among risk-taking characters is also reflective of real-life society where risk-taking is also particularly characteristic for young men (Tamás et al., 2019).

Flynn et al. (2015) also found ethnic minorities to be relatively overrepresented as engaging in alcohol use, compared to their overall representation. As we only coded characters engaging in some type of risk behavior, rather than all characters, our results do not allow us to investigate this comparatively for our sample. Our analysis did allow us to compare the demographic makeup of characters engaging in various risk behavior, revealing instances of overrepresentation and stereotyping among drug-using characters. Characters shown engaging in drug use were less likely to be white and more likely to be Asian or Arabic, compared to characters who engaged only in other kinds of risk behavior. This stereotyped representation among drug users could have important implications as adolescents from certain non-dominant minority groups may already be at increased risk for engagement in health-risk behavior (Factor et al., 2011). Given the fact that adolescent viewers are more likely to model behaviors portrayed by characters whom they identify with and feel similar to themselves (Austin et al., 2000), these stereotyped portrayals may potentially lead to disparities in health behaviors and outcomes (Bleakley et al., 2017).

Limitations

This study provides insight into the prevalence, tone, social context, and demographic makeup with which various risk behaviors are portrayed in a broad range of popular entertainment media and provides new insights into how the portrayals of various risk behaviors compare to one another. However, there are also some limitations. First, we only coded characters engaging in risk behavior, so we cannot make statements about the percentage with which various types of risk behavior occurred on a character level. In line with this, we only analyzed whether a character engaged in certain risk behavior never or once (or more), rather than counting every instance of risk behavior. Both choices enabled us to analyze a large sample of programs and many risk-taking characters, but also lead to some limitations in terms of the analyses we were able to perform. Future research might expand on this by also considering the frequency with which risk behavior is portrayed, relative to program length. Furthermore, we coded tone and social context with single variables on the program level, while this provides us with information about the overall tone of these productions, it is not possible to draw a full and nuanced picture of the context in which these behaviors are portrayed. Future research might add to this by zooming in on these factors through more in-depth scene-by-scene analyses.

Furthermore, while our results shed light on potential stereotyping with regard to risk-taking within minority groups, we only looked at risk-taking characters and a non-exhaustive selection of demographic characteristics. Given the implications of stereotyped representations of risk behavior discussed above, future research may want to focus more specifically on the representation of risk behavior among various minority groups including gender and sexual minorities.

A final limitation relates to sample selection. The algorithmic nature of these platforms and their lack of transparency regarding viewing figures make it difficult to define what is popular content among adolescents. We decided upon a naturalistic approach by creating accounts for new (teen-aged) users and coding all content that was offered to these user profiles by the algorithm as being trending and popular by the video-on-demand platforms. While these platforms are the most popular way for adolescents to watch media (Nielsen, 2016; Ofcom, 2018), and recommender systems play an important role in content selection (Frey, 2021; Patch, 2018), the lack of viewing figures mean we cannot be sure that these trending programs are also the programs most popular among adolescents. It is possible that our sample missed programs popular among teens and included others that might not be as popular. Future research might investigate how we can reliably determine viewing figures of online and on-demand content for demographic subgroups. In the meantime, it may be useful to delve deeper into the content offered to adolescent users by each platform, by creating multiple accounts per platform using different demographic settings (e.g. age, gender) when possible, to create an even more complete overview of the content offered to users by that specific platform.

Furthermore, sample selection and data collection for this study was conducted in 2018 and it thereby provides a snapshot of how risk behavior is portrayed in popular VOD content that was presented as popular/trending by the platforms at that moment in time. While trends regarding the prevalence of alcohol tend to remain relatively stable over time (Lyons et al., 2011) other behaviors such as tobacco and drug use show more variance (Thrasher et al., 2014) which may also be reflected in VOD content. It may therefore be valuable to conduct content analyses of on-demand platforms at regular intervals.

Conclusion

When considering our conclusions in light of social learning and cultivation theory the portrayal of risk behavior in trending content may not be without consequences. The portrayal of alcohol and tobacco as something that is normal, having few negative consequences, and being particularly prevalent among young adults can affect adolescents’ perceptions of the prevalence and acceptability of such behavior in real life, which may, in turn, affect their own intentions to engage in these behaviors (Nan & Zhao, 2016). As such, this portrayal may be worrisome, especially because the non-linear nature of video-on-demand platforms allows adolescents to watch much of this content without parental supervision. Understanding adolescents’ perceptions of this content is therefore critical. This study has taken the first step in describing how various risk behaviors are portrayed on screen. Future steps may be taken to discuss these portrayals with adolescents, to understand how they perceive them and what steps media literacy interventions may take to enhance critical reflection on these portrayals.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.