Abstract
Introduction
Micro-influencers (1000–100 000 followers) promote e-cigarettes alongside cannabis on social media. This study examined the association of exposure to such marketing with adolescent perceptions of e-cigarettes and micro-influencers.
Methods
In this 2023–2024 survey-based, in-class repeated-measures experiment, California adolescent e-cigarette and cannabis never-users (n = 1402; Mage = 17 [SD = 0.6]; 51.4% female, 41.7% Hispanic) were randomly shown three pairs of Instagram images featuring micro-influencers promoting: (1) e-cigarettes with cannabis (experimental group), (2) e-cigarettes alone (control), or (3) no substances (placebo). After viewing each image pair, participants rated perceived influencer credibility. After viewing all images, susceptibility to use e-cigarettes was assessed. Three susceptibility sub-items and their composite measure were compared across groups (experimental vs. control; placebo vs. control) and between participants who perceived micro-influencers as credible versus noncredible, using binomial generalized linear mixed effects models. The association of self-reported exposure to micro-influencer e-cigarette or cannabis posts with these outcomes was also assessed.
Results
Participants in the experimental group had higher odds of e-cigarette use intentions if offered by a friend among all and among those who perceived micro-influencers as credible, compared to controls. Those who self-reported exposure to micro-influencer e-cigarette or cannabis posts were at higher odds of overall susceptibility, intentions to use if a friend offers, and curiosity about e-cigarette use.
Conclusions
Influencer e-cigarette promotion alongside cannabis, and perceptions of influencers as credible, may contribute to e-cigarette use intentions among adolescent never-users. Improvement of social media community guidelines and greater policy attention to e-cigarette and cannabis marketing are needed.
Implications
The study showed that influencer promotion of e-cigarettes alongside cannabis, and perceptions of influencers as credible, may contribute to intentions to use e-cigarettes among adolescent never-users. The findings suggest the need for improvement of social media community guidelines and greater policy attention to marketing of e-cigarettes and cannabis.
Trial Registration
Introduction
Nicotine e-cigarettes and cannabis products are among the most commonly used substances by U.S. adolescents.1,2 In 2024, 3.5% of middle school and 7.8% of high school students reported past-month e-cigarette use,3 while 8.3% of eighth graders and 29% of 12th graders reported past-year cannabis use in 2023.4 E-cigarette use is linked to progression to cannabis use and vice versa, often concurrently.2,5 Both substances harm brain development, increase addiction risk, and are associated with mental health issues.6,7
Exposure to tobacco-related social media content posted by celebrities or influencers (vs. nonexposure) is associated with higher odds of susceptibility to e-cigarette use among adolescents.8 Research shows that about 30% of e-cigarette marketing on TikTok is influencer marketing.9 However, the associations of exposure to e-cigarette and cannabis influencer marketing on social media (eg, Instagram, TikTok)10,11 with adolescent susceptibility to e-cigarette use has not been extensively studied.
Prior research also showed evidence of promotion of both e-cigarettes and cannabis by the same influencers (often across separate social media posts within the influencer feeds and in some cases even in the same post).12 For example, some micro-influencers—defined as noncelebrity influencers with about 1000–100 000 followers13—who promote e-cigarette products on behalf of tobacco brands on Instagram or TikTok also promote cannabis or hemp products (including CBD [cannabidiol, a cannabis compound] and psychoactive Delta-8 THC [delta-8-tetrahydrocannabinol]).12,14 Promoting cannabis and tobacco products next to each other may increase the risk of use of both substances, as youth may seek more intense psychoactive effects from both substances15 or reduce risk perception of one substance when compared to the other. This could also be partially due to youth misperceptions about which product (e-cigarette or cannabis) is being promoted in posts featuring disposable and cartridge-based nicotine and cannabis vape devices, which look similar to one another and are often featured in influencer marketing.12,16
Considering evidence of cross-product promotion of e-cigarettes and cannabis by influencers,12,14 we used a randomized survey-based experiment to assess the associations of exposure to cannabis products shown alongside e-cigarettes in promotional Instagram images featuring micro-influencers with perceptions of micro-influencers and susceptibility overall, intentions (in the next 6 months and if a friend offers), and curiosity, to use e-cigarettes among adolescents. Consistent with Social Learning Theory (SLT),17,18 consumers may perceive influencers as trusted and credible sources of information and adopt behaviors (eg, e-cigarette use) through observing and modeling influencers.19,20 We hypothesized that adolescents who viewed images of micro-influencers promoting e-cigarettes alongside cannabis (experimental condition) would report higher overall susceptibility, intentions, and curiosity to use e-cigarettes (H1), compared to those who viewed images of micro-influencers promoting e-cigarettes alone (control condition). We also tested an interaction between the experimental condition and perceived influencer credibility, consistent with the Prototype Willingness Model (PWM) and our prior research, hypothesizing that adolescents in the experimental condition (vs. control) would report higher overall susceptibility, intentions, and curiosity to use e-cigarettes, when they perceived influencers as credible (H2).21,22 That is, since PWM suggests that people are more willing to engage in a risky behavior when they have a positive view of the prototypical person performing that behavior.21,22 Finally, we assessed self-reported exposure to content of micro-influencers who post about e-cigarettes and cannabis, hypothesizing that such exposure (vs. no exposure) will be associated with susceptibility overall, intentions, and curiosity, to use e-cigarettes (H3).
Materials and Methods
Participants and Procedures
In total, 3380 12th-grade adolescents (mean age = 17) from 15 socioeconomically and racially diverse schools in the Los Angeles metropolitan area participated in Wave 7, September 2023–January 2024, of an ongoing semi-annual survey about tobacco use behaviors conducted by the University of Southern California (USC) Tobacco Center of Regulatory Science (TCORS).23 They completed a 30-minute survey administered in class via REDCap on classroom Chromebooks. Of these 3380, 1402 participants—never-users of e-cigarettes and cannabis—were randomized to complete the survey-embedded randomized experiment (Figure 1). Written informed assent and parental consent were obtained prior to data collection. The study was approved by the USC Institutional Review Board (UP-23-00810).
Figure 1.
Consortium diagram.
Experimental Design
Each participant viewed two paired images of three micro-influencer profiles (selected from a set of four profiles), totaling six images shown in their assigned condition: (1) a micro-influencer promoting an e-cigarette product (image 1) paired with an image of that micro-influencer promoting cannabis (image 2), (experimental condition); (2) a micro-influencer promoting an e-cigarette product (image 1) paired with a neutral profile (image 2) of that influencer (control); and (3) two neutral images of a micro-influencer without any substances (placebo condition). Each image had a social media username (not the actual name) of the influencer (Figure 2).
Figure 2.
Example of conditions in the randomized survey-based experiment. In the experimental condition, a micro-influencer promotes a disposable device for nicotine vaping, followed by this micro-influencer promoting a cannabis product. In the control condition, the same micro-influencer promotes the same disposable device for nicotine vaping, followed by a neutral profile of that micro-influencer. The placebo condition includes two images of neutral profiles of the same micro-influencer with no substance (nicotine or cannabis) presence. The experiment among California adolescents (n = 1402) featured publicly available images of the actual micro-influencers on Instagram. The individual in Figure 2 was not used in the stimuli and was created by Open AI’s GPT for illustration purposes as an example of the experimental conditions.
Sample for the Experimental Stimuli
The experimental stimuli were created from actual Instagram images of existing micro-influencers with public accounts identified in prior research14 who had disclosed that they were sponsored by e-cigarette companies. These micro-influencers were known to post e-cigarette content, and, in some instances, posted about e-cigarettes alongside cannabis products.
Experimental Stimuli Development
To ensure that the content of the stimuli only differed by the characteristics of interest, the e-cigarette image for each influencer (image 1) remained unchanged for the experimental and control conditions (the placebo condition featured two profile images of influencers with no substance use or display). The experimental and control conditions differed by the second image (image 2) per their respective condition (Figure 2). To create homogenous stimuli, ensuring that no micro-influencer is perceived drastically differently from the others by age and gender characteristics, 30 undergraduate students reviewed influencer profiles, answering the questions about perceived gender (“How likely that the person featured in this promotional image is Man/Woman/Transgender?”) and perceived age of the influencers (“How likely that the person featured in this promotional post is: (a) younger than 21; (b) 21–30; (c) 31–40; and (d) older than 40?”), measured on the scale of 0 (not at all likely) to 10 (highly likely). Profiles with the highest rating agreement and highest rating scores (8 or above) were selected for the experiments. (Two male and two female profiles with the perceived age of “between 21 and 30” were selected out of the overall eight profiles shown to the raters).
Measures
Susceptibility, Intentions, and Curiosity to Use E-cigarettes
Susceptibility, intentions, and curiosity to use e-cigarettes, shown to be successful predictors of future tobacco use,20,24 were measured using the validated three-item scale adapted from the Population Assessment of Tobacco and Health (PATH) Study.25 The outcomes were assessed among never-users of e-cigarettes and cannabis (Survey instrument in Supplementary Material). Three items (based on the images participants viewed): (1) If one of your best friends were to offer you an e-cigarette, would you use it (intention to use if offered by a friend)? (2) Do you think you will try vaping e-cigarettes in the next six months (intention to use in the next 6 months)? (3) Have you ever been curious about vaping e-cigarettes (curiosity)? were assessed separately and as a combined variable (representing susceptibility overall). Three items were examined both individually and as a composite measure to capture distinct dimensions of susceptibility (ie, peer influence, intention, and curiosity) and to allow for comparison with studies that use either approach.20,24 Consistent with prior research,26 separate and combined variables were dichotomized with responses “definitely not” (to all four items for the combined variable) coded as “not susceptible/not curious/not having use intentions” versus responses “probably not,” “probably yes,” or “definitely yes” coded as “susceptible/curious/having use intentions.”
Perceived Influencer Credibility
The perceived influencer credibility for two personality traits (ie, honesty and trustworthiness) was assessed for each of the three micro-influencer profiles after each of the three pairs of images participants viewed using a 0–100 scale validated in prior research.20,27 Supplementary Table 1 shows the scores for each of the two traits and for each of the micro-influencer profiles. The items representing the two traits were combined (by computing the mean of the two trait scores) into one variable because the items were closely related, α = 0.97. The mean “perceived influencer credibility” score (ie, the mean of the two trait scores) was computed for each pair of images.
Self-Reported Exposure to e-Cigarette or Cannabis Content Posted by Micro-Influencers
Self-reported exposure to e-cigarette and/or cannabis content by micro-influencers was measured as exposure (yes/no) assessed among those who reported at least weekly exposure to general e-cigarette and/or cannabis-related social media (Supplementary Material).
Statistical Analysis
Main Effects Models
To test H1, we used binary logistic mixed effects regression models to test the main effects of exposure to images of influencers promoting e-cigarettes alongside cannabis (experimental condition) versus promoting e-cigarettes alone (control) on susceptibility, intentions, and curiosity, to use e-cigarettes (assessed after viewing all images), adjusting for perceptions of influencer credibility (assessed after each pair of three images featuring influencers). Repeated measures in experimental conditions (ie, perceived influencer credibility) were modeled as a random effect (ie, a random intercept). This approach was chosen to reduce potential bias that could occur when stimulus repetitions go unanalyzed as a random effect28 (we also included schools as a random intercept to account for clustering by 15 schools and 1 cluster [n = 117] with missing school ID retained in the dataset). The models were adjusted for fixed effects of sociodemographic and other covariates described below. The associations between the no-substance placebo condition (vs. control) and the outcomes was assessed but was not a focus of hypothesis testing, since the no-substance placebo condition served as an additional control to isolate the associations attributable to substance-related content.
Interaction Effects Models and Stratified Analyses
To test the potential moderating effects of perceived influencer credibility on associations of the experimental condition with susceptibility, intentions, and curiosity, to use e-cigarettes, the interaction term of experimental conditions × perceived influencer credibility (as a fixed effect) was added to the subsequent interaction effects models (while also modeling perceived influencer credibility as a random intercept). To interpret interaction effects of H2, stratified analyses were performed to assess the associations of the experimental condition on susceptibility, intentions, and curiosity to use e-cigarettes, stratified by the level of perceived influencer credibility (ie, high [influencer credibility score at or above the mean] vs. low). In all the main effect, interaction effect, and stratified models, we organized the data in long format, treating influencer credibility as a single variable across three pairs of images featuring micro-influencers. Given the low missingness (2%–5%) for the variables with missing values (indicated in Figure 1 and Table 1), listwise deletion was applied without further comparison. For the susceptibility outcome, where missingness was higher (11%), we compared participants with and without the data and found no statistically significant differences in key demographic characteristics. Therefore, no missing data treatment was used in this analysis.
Table 1.
Participant Characteristics for Adolescent Never-Users of e-cigarettes (n = 1402)
| Experimental conditions | ||||
|---|---|---|---|---|
|
Total (n = 1402) |
E-cigarette + cannabis (n = 452) |
E-cigarette alone (n = 443) |
No substance (n = 507) |
|
| Participant characteristics | % (n)/M (SD) | |||
| Age | 17 (0.6) | 17 (0.6) | 17 (0.6) | 17 (0.6) |
| Sex assigned at birth | ||||
| Female | 51.4 (721) | 51.1 (231) | 52.3 (232) | 50.9 (258) |
| Male | 47.6 (667) | 48.2 (218) | 46.7 (207) | 47.7 (242) |
| Unreported | 1.0 (14) | 0.7 (3) | 0.9 (4) | 1.4 (7) |
| Sexual orientation | ||||
| Heterosexual | 73.5 (1030) | 73.2 (331) | 72.7 (322) | 74.4 (377) |
| Nonheterosexual | 26.5 (372) | 26.8 (121) | 27.3 (121) | 25.6 (130) |
| Race | ||||
| Asian | 46.9 (658) | 47.3 (214) | 48.9 (217) | 44.8 (227) |
| White | 15.5 (218) | 15.7 (71) | 14.2 (63) | 16.6 (84) |
| Othera | 21.7 (304) | 20.6 (93) | 23.0 (102) | 21.5 (109) |
| Multiple race | 15.8 (222) | 16.3 (74) | 13.8 (61) | 17.2 (87) |
| Ethnicity | ||||
| Hispanic/Latinax | 41.7 (585) | 42.0 (190) | 39.7 (176) | 43.2 (219) |
| Non-Hispanic/Non-Latinax | 58.3 (817) | 58.0 (262) | 60.3 (267) | 56.8 (288) |
| Social media use frequency | ||||
| Use of Instagram and/or TikTok daily | 77.3 (1084) | 76.5 (346) | 77.7 (344) | 77.7 (394) |
| Exposure to e-cigarette/cannabis content on social media (%/n yes) | ||||
| Nonsource-specific exposure on Instagram and/or TikTok (weekly or daily) | 16.5 (231) | 18.6 (84) | 13.8 (61) | 17.0 (86) |
| Micro-influencer posts | 6.4 (90) | 7.7 (35) | 3.8* (17) | 7.5 (38) |
| Substance use/susceptibility to e-cigarette use (%/n yes) | ||||
| Past month alcohol or tobacco productb use | 4.8 (67) | 6.0 (27) | 4.3 (19) | 4.1 (21) |
| Unreported | 0.2 (3) | 0.2 (1) | 0.2 (1) | 0.2 (1) |
| Susceptiblec (pre-exposure) | 16.3 (228) | 14.4 (65) | 16.9 (75) | 17.4 (88) |
| Unreported | 2.2 (31) | 2.2 (10) | 2.2 (10) | 2.3 (11) |
| Susceptiblec (postexposure) | 17.9 (251) | 18.1 (82) | 18.3 (81) | 17.4 (88) |
| Unreported | 11.4 (160) | 11.3 (50) | 11.7 (53) | 11.2 (57) |
n in the header stands for the sample size; M (SD) stands for mean and standard deviation.
*** p-value < .001, ** p-value <0.01, * p-value <0.05, ‘p-value <0.1. p-values indicate which combinations of participant characteristics by experimental conditions differ significantly from the expectation under the null-hypothesis of no differences between the observed and the expected values.
a “Other” includes Native Hawaiian or Pacific Islander, American Indian, or Alaska Native, Black, and a category defined as “Other.”
bAlcohol use represents one full drink of any alcoholic beverage used in the past month. Tobacco products include combustible cigarettes, IQOS or other heated tobacco devices, oral nicotine products, big cigars, little cigars and cigarillos, and hookah used in the past month.
cSusceptibility included the three-item scale: Would you try e-cigarette products if one of your best friends offered them to you? Do you think you would use e-cigarette products in the next 6 months? Are you curious about using e-cigarette products? The scale was collapsed into dichotomized responses: “definitely not” to all the items vs. “probably not,” “probably yes”, “definitely yes” to one or more items.
Perceived Influencer Credibility
The mean influencer credibility perception scores were measured for each condition using one-way ANOVA. Tukey’s Honest Significant Difference post hoc test was applied for pairwise comparisons across conditions.
Self-Reported Exposure to E-cigarette or Cannabis Content Posted by Micro-Influencers
Self-reported exposure to e-cigarette and/or cannabis content posted by micro-influencers was included as an additional predictor (H3) in the main effect models assessing susceptibility, intentions, and curiosity to use e-cigarettes.
Covariates
All multivariable models were adjusted for covariates, informed by prior research,20,27 including socio-demographic characteristics (age, race/ethnicity, sex at birth, sexual identity), past month tobacco (other than e-cigarettes) and alcohol use, frequency (daily vs. less frequent or none) of Instagram and/or TikTok use, and pre-existing susceptibility to e-cigarette use, all assessed prior to experimental exposure (Supplementary Material).
Supplementary Analyses
As supplementary analyses, we conducted main effects, interaction effects, and stratified models using the no-substance presence condition as a reference.
Statistical Significance and Software
Adjusted odds ratios (AORs) with 95 % confidence intervals (CIs) were reported with statistical significance set at p < .05 (two-tailed) after applying Benjamini–Hochberg multiple testing corrections to control for the false discovery rate at 0.05. All statistical analyses were conducted using R software (version 4.2.2).
Results
Participant Characteristics
Participant (n = 1402 never-users of e-cigarettes and cannabis) characteristics in each of the three experimental conditions are reported in Table 1. Adolescents were Mage = 17 [SD = 0.6]; 51.4% female, 41.7% Hispanic; 4.8% reported past month use of tobacco (other than e-cigarettes) or alcohol; 16.5% reported daily or weekly exposure to e-cigarette and/or cannabis content on Instagram and/or TikTok; 6.4% reported exposure to e-cigarette or cannabis social media content posted by micro-influencers; 16.3% reported overall susceptibility to e-cigarette use prior to experimental exposure, and 17.9%—postexperimental exposure (Table 1).
Main Effects Models
In partial support of H1, exposure to influencer e-cigarette promotion alongside cannabis (experimental condition) versus influencer e-cigarette promotion alone (control) was associated with higher odds of intentions to use e-cigarettes if a friend offers (AOR = 1.55; 95% CI = 1.15 to 2.09), after adjusting for perceptions of influencer credibility. The association of exposure to the experimental condition with the overall susceptibility (a composite measure), curiosity, and intentions to use e-cigarettes in the next 6 months was not statistically significant. The association of exposure to the placebo condition (vs. control) was not statistically significant with any outcome (Table 2).
Table 2.
Multivariablea Logistic Regression Analyses of Experimental Conditions and Susceptibility, Intentions, and Curiosity to use e-cigarettes
| Independent variables | Outcomes | |||||||
|---|---|---|---|---|---|---|---|---|
| Susceptibility overall b | Intentions to use | Curiosity | ||||||
| If a friend offers | In the next 6 months | |||||||
|
AOR
(95% CI) |
Adj. p-value c | AOR (95% CI) | Adj. p-value c |
AOR
(95% CI) |
Adj. p-value c |
AOR
(95% CI) |
Adj.
p-valuec |
|
| Main effects models | ||||||||
| E-cigarettes and cannabisd | 1.36 (1.03; 1.81) |
.07 | 1.55 (1.15; 2.09) |
.01 | 1.26 (0.89; 1.79) |
.26 | 0.86 (0.65; 1.13) |
.49 |
| No-substance presenced | 0.87 (0.66; 1.16) |
.39 | 1.07 (0.80; 1.44) |
.82 | 1.05 (0.75; 1.48) |
.78 | 0.90 (0.69; 1.18) |
.57 |
| Self-reported exposure to e-cigarette and/or cannabis posts from micro-influencerse | 2.66 (1.84 to 3.84) |
<.001 | 2.19 (1.49 to 3.23) |
<.001 | 1.62 (1.05 to 2.51) |
.09 | 2.43 (1.70–3.49) |
<.001 |
| Perceptions of influencer credibilityf, SD | 0.15 | 0.28 | 0.22 | 0 | ||||
| Interaction effects models | ||||||||
| E-cigarette and cannabis × perceptions of influencer credibility | 1.31 (0.96 to 1.79) |
.20g | 1.51 (1.09 to 2.10) |
.047 | 1.34 (0.91 to 1.98) |
.23g | 1.19 (0.87 to 1.64) |
.50g |
| No-substance presence × perceptions of influencer credibility | 0.80 (0.59; 1.08) |
.21 | 0.84 (0.61; 1.16) |
.51 | 0.87 (0.60; 1.26) |
.55 | 0.77 (0.57; 1.04) |
.19g |
| Stratified analyses | ||||||||
| Effects of the experimental conditions assessed among respondents with higher influencer credibility perception scoreh | ||||||||
| E-cigarette and cannabis | 2.19 (1.48 to 3.24) |
<.001 | 2.57 (1.69 to 3.91) |
<.001 | 1.72 (1.07 to 2.77) |
.05 | 1.20 (0.81 to 1.76) |
.39 |
| No-substance presence | NAi | NAi | NAi | 0.81 (0.58 to 1.15) |
.39 | |||
| Effects of the experimental conditions assessed among respondents with lower influencer credibility perception scoreh | ||||||||
| E-cigarette and cannabis | 0.91 (0.59 to 1.38) |
.74 | 1.05 (0.68 to 1.63) |
.83 | 0.97 (0.57 to 1.64) |
.97 | 0.62 (0.41 to 0.96) |
.06 |
| No-substance presence | NAi | NAi | NAi | 0.90 (0.55 to 1.45) |
.70 | |||
Abbreviations: AOR = adjusted odds ratio; 95% CI = 95% confidence interval; Adj. p-value = adjusted p-value.
Binomial generalized linear mixed effects models to assess the association of the experimental condition (influencer promotion of e-cigarettes alongside cannabis) and placebo condition (no substance) versus influencer e-cigarette promotion alone (control) with susceptibility to e-cigarette use among never users of e-cigarettes and cannabis (n = 1402) were used. All models were adjusted for age, sex assigned at birth, sexual identity, race/ethnicity, past 30 days alcohol and tobacco use (any alcohol, combustible cigarettes, IQOS or other heated tobacco devices, oral nicotine products, big cigars, little cigars and cigarillos, and hookah), exposure to e-cigarette or cannabis social media content posted by micro-influencers, pre-existing susceptibility to e-cigarette use, and frequency of Instagram and/or TikTok use (daily vs. less frequent/no [combined: weekly, monthly or less, no]).
Susceptibility included the three-item scale: Would you try e-cigarette products if one of your best friends offered them to you? Do you think you would use e-cigarette products in the next 6 months? Are you curious about using e-cigarette products? The scale was collapsed into dichotomized responses: “definitely not” to all the items vs. “probably not,” “probably yes,” and “definitely yes” to one or more items.
Benjamini–Hochberg multiple testing corrections were applied to control the false discovery rate at 0.05 (based on two-tailed corrected p-value).
Control condition: Influencer e-cigarette promotion alone.
Self-reported exposure to influencers who posted about e-cigarettes and/or cannabis was assessed among those who reported at least weekly exposure to general (non-source-specific) e-cigarettes and/or cannabis posts. The reference group (no exposure) included participants who explicitly reported no exposure to e-cigarette/cannabis content from specific sources and those who reported less than weekly or no exposure to nonsource-specific e-cigarette/cannabis posts and were therefore not shown the source-specific exposure question but remained in the analysis as part of the unexposed group.
To test the main and interaction effects, the continuous scale (0–100) of the perception of influencer credibility score was standardized in the models (mean = 0; SD = 1).
If unadjusted p-values for interaction effects were marginally significant (<.1) and/or if significant proportion differences were observed as shown in interaction plots in Figure 3, exploratory stratified analyses were conducted. Unadjusted p-values in interaction effects models: experimental condition X influencer credibility perception for the susceptibility overall: p-value = .09; for the intention to use e-cigarettes in the next 6 months: p-value = .1; for curiosity to use e-cigarettes: p-value = .27 (experimental condition) and p-value = .09 in the placebo condition.
To conduct stratified multigroup analyses, the perception of influencer credibility score was dichotomized using the mean split of 40.03 to assess susceptibility to e-cigarette use: the higher score was set at or above 40.03; the lower score was set below 40.03.
Stratified analyses were not conducted since there were no significant interaction effects or significant proportion differences as shown in interaction plots in Figure 3.
Interaction Effects Models and Stratified Analyses
Interaction effects are reported in Table 2 and Figure 3. In partial support of H2, we found a significant interaction between the experimental (e-cigarette and cannabis) condition and perceived influencer credibility on intentions to use e-cigarettes if a friend offers (Table 2). Among participants who perceived micro-influencers as credible (at or above the mean of the influencer credibility perception score), a larger proportion of those in the experimental condition (27%) than those in the e-cigarette-only control condition (19%) reported intentions to use e-cigarettes if a friend offers (p-value = .003, Figure 3). Among participants who perceived micro-influencers as credible, those in the experimental condition had higher odds of intentions to use e-cigarettes if a friend offers (AOR = 2.57; 95% CI = 1.69 to 3.91), compared to those in the control condition (Table 2). Although this was the only interaction effect that remained statistically significant after adjusting for multiple comparisons (Table 2), Figure 3 illustrates similar unadjusted interaction patterns for intentions to use e-cigarettes in the next 6 months and for overall susceptibility. Additionally, among participants who perceived micro-influencers as noncredible (below the mean of the influencer credibility perception score), a smaller proportion in the experimental condition (9%) than in the control condition (13%) reported curiosity about e-cigarette use (p-value = .03, Figure 3). Guided by plot patterns in Figure 3 and marginally significant interaction effects based on the unadjusted p-values (Table 2), our exploratory stratified analyses suggested a potential moderating effect of influencer credibility on the experimental condition for intentions to use e-cigarettes in the next 6 months and overall susceptibility (Table 2).
Figure 3.
Proportion of participants susceptible to use, having intention to use, or curious about, e-cigarettes in the e-cigarette + cannabis experimental condition and the e-cigarette alone control condition. Higher influencer credibility perception score represents the values at or above the mean. Lower influencer credibility perception score represents values below the mean. Differences in proportions between experimental and control conditions among participants who reported a higher or lower influencer credibility perception score were compared using a two-sample test for equality of proportions with continuity correction. Proportions were calculated using long-format data, where each participant contributed multiple rows corresponding to each of the three pairs of images they viewed. Sample sizes reflect the number of participant-level responses across images in each group. Higher influencer credibility perception score, experimental group, n = 440; control group, n = 586; lower influencer credibility perception score, experimental group, n = 741; control group, n = 537. Intention to use e-cigarettes if a friend offers: higher influencer credibility perception score, experimental group, n = 117; control group, n = 109; lower influencer credibility perception score, experimental group, n = 85; control group, n = 71. Intention to use e-cigarettes in the next 6 months: higher influencer credibility perception score, experimental group, n = 73; control group, n = 69; lower influencer credibility perception score, experimental group, n = 48; control group, n = 42. Curiosity about e-cigarette use: higher influencer credibility perception score, experimental group, n = 102; control group, n = 115; lower influencer credibility perception score, experimental group, n = 68; control group, n = 71. Susceptibility, combined measure: higher influencer credibility perception score, experimental group, n = 137; control group, n = 135; lower influencer credibility perception score, experimental group, n = 98; control group, n = 87.
Perceptions of Micro-Influencer Credibility
The perceived influencer credibility score was the highest in the placebo (no substance) condition (mean = 48.6), followed by the e-cigarette only (control) condition (mean = 38.5), and the e-cigarette and cannabis (experimental) condition (mean = 31.9, p-values < .001, Supplementary Figure 1).
Self-Reported Exposure to E-cigarette or Cannabis Content Posted by Micro-Influencers
H3 was supported for all but one susceptibility outcome. Participants were at higher odds of overall susceptibility (AOR = 2.66; 95% CI = 1.84 to 3.84), intentions to use e-cigarettes if a friend offers (AOR = 2.19; 95% CI = 1.49 to 3.23), and curiosity about e-cigarette use (AOR = 2.43; 95% CI = 1.70 to 3.49) if they self-reported seeing e-cigarette or cannabis posts from micro-influencers versus those who did not see this content (Table 2).
Supplementary Analyses
The results of the supplementary analyses where the no-substance presence condition was used as a control condition appear consistent with those where the e-cigarette-only condition served as the control (Supplementary Table 2).
Discussion
Using a randomized survey-based experiment, this study examined associations between exposure to noncelebrity, micro-influencer marketing that promoted e-cigarettes alongside cannabis on Instagram and susceptibility, intentions, and curiosity to use e-cigarettes among California adolescent never-users of e-cigarettes and cannabis. Extending prior research on exposure to celebrity and non-celebrity influencer marketing and its associations with youth e-cigarette perceptions and use,20,24,27 this study found that experimental exposure to micro-influencer posts featuring e-cigarettes and cannabis (vs. e-cigarettes alone) was associated with e-cigarette use intentions if offered by a friend, with influencer credibility moderating these effects.
Participants in the experimental condition (who viewed images of micro-influencers promoting e-cigarette use alongside cannabis) on average rated these influencers as less credible than those in the e-cigarette-only or no-substance use conditions. Nevertheless, after adjusting for perceived influencer credibility, participants in the experimental condition had higher odds of intending to use e-cigarettes if offered by a friend. Moreover, this association was amplified among participants in the experimental condition who, despite lower overall credibility ratings in this condition, perceived the influencers as credible. Thus, e-cigarette images paired with cannabis may have reduced perceived influencer credibility for some adolescents but increased susceptibility to use e-cigarettes if offered by a friend among those who still found the influencers credible. The results align well with the PWM,21,22 which suggests greater willingness to engage in risky behaviors when the prototypical actor (eg, an influencer) is viewed positively. Although not statistically significant in the main and interaction effect models, exploratory stratified analyses—guided by patterns in the interaction plots and marginally significant unadjusted interaction p-values—revealed similar results for other susceptibility measures, suggesting a possible moderating effect of influencer credibility on intentions to use e-cigarettes in the next six months and overall susceptibility.
The proportion of adolescents who self-reported exposure to e-cigarette or cannabis content from micro-influencers was small (about 6%), which could be a result of difficulty recalling the source of the content that participants reported seeing, especially considering that these adolescents were e-cigarette and cannabis never-users. However, our study found that self-reported exposure to such posts was strongly associated with overall susceptibility, intentions to use if a friend offers, and curiosity about e-cigarette use, suggesting that this content may resonate with adolescents, despite overall low recall of its source, and increase the risk of their e-cigarette trials. Self-reported exposure was a stronger predictor than experimental exposure, possibly due to the limited duration of viewing experiment stimuli and potential lack of engagement with the image-based experimental stimulus. Taken together, both the experimental and observational predictors show that e-cigarette and cannabis-related influencer marketing is a non-negligible risk factor that could contribute to adolescent e-cigarette use. The finding is concerning, considering evidence of promotion of e-cigarettes and cannabis or hemp products (including CBD and psychoactive Delta-8 THC) by micro-influencers on social media12 that can contribute to adolescent susceptibility to not only vaping, but also smoking or eating THC, CBD, or hemp products.
This study emphasizes the need for regulation of social media substance-use marketing, especially influencer marketing, and greater research and policy attention to co-use and marketing of e-cigarette and cannabis products. Most platforms prohibit paid tobacco ads,29 with Facebook, Instagram, and TikTok also prohibiting influencer tobacco marketing.29 Some platforms prohibit selling or promoting cannabis,30,31 with, eg, Facebook, Instagram, and YouTube loosening cannabis marketing restrictions.32,33 Cannabis promotion is exacerbated by the lack of legislation (eg, in the United States) targeting cannabis marketing and restrictions for online sales to youth.34 Further improvement in the enforcement of social media community guidelines is necessary.35,36 Platforms could utilize computer vision algorithms to automatically identify substance-related posts, including those featuring youth, and promptly remove them.37,38
While influencers may advertise unhealthy products to pursue commercial interests, consumers of their content (including youth) may perceive these influencers as trusted and authentic sources.39 Authorities could consider penalizing brands and influencers for noncompliance with federal marketing regulations such as restricting access to e-cigarette marketing content to youth and following the Federal Trade Commission (FTC) guidelines for sponsorship disclosures.40,41 At the same time, since influencers can be perceived as trusted sources of information, public health authorities may harness influencers for anti-substance use messaging and in health literacy campaigns about the harmful effects of substance use39,42 or disseminating other positive public health messages.39 The findings can also be disseminated via media literacy curriculum in schools.
Limitations
Since we did not use probability-based sampling or apply sampling weights in this California-based study, the results may not be generalizable to all California or US adolescents. Concerns about generalizability are mitigated by the fact that the ethnic diversity of the samples represents the projected future ethnic diversity of the United States.43 Despite the randomized experimental design, self-reported survey outcomes might be prone to social desirability bias. The measures were assessed only at one point in time (vs. multiple times over time). Participants were only exposed to three paired images of influencer profiles, and such brief exposure may have limited the strength of observed associations for the experimental condition. While the experimental stimuli were created from actual Instagram posts featuring micro-influencers promoting e-cigarette products, the survey did not use an actual Instagram interface to simulate a real-life social media use experience. Our experimental stimuli did not include captions, which are often present in real-world social media posts. We excluded captions to focus on the visual impact of influencer content, but this choice might have reduced ecological validity since text-based marketing elements were omitted. Due to the original study goal to examine e-cigarette use intentions, we did not assess susceptibility to cannabis use as an outcome.
Conclusions
The study demonstrated that an influencer marketing tactic—such as promoting e-cigarettes alongside cannabis products—may contribute to adolescent e-cigarette use intentions, particularly if offered by a friend and among those who perceive influencers as credible. Future research should continue tracking trends in e-cigarette and cannabis influencer marketing, both online and offline, and examine how shifts in tobacco and cannabis policy (eg, marketing authorization of new e-cigarette products, state-level cannabis legalization) and social media community guidelines (eg, loosening cannabis marketing restrictions) affect these influencer marketing practices. Future research should also assess the prevalence of cannabis content in the social media feeds of influencers who post about e-cigarettes. Research should also assess how influencer credibility is perceived in different types of substance-related social media posts (eg, short entertaining clips vs. longer-form narrative-driven videos), and in overtly versus covertly promotional posts: ie, with versus without FTC-required disclosures of tobacco and cannabis sponsorship. Influencers who omit such disclosures may be seen as more genuine and authentic,19,44 rather than motivated by financial gain, which could lower harm perceptions or increase susceptibility to e-cigarette use. Strengthening social media guidelines, especially regulations of youth-appealing promotions of e-cigarettes and cannabis, could help reduce adolescent exposure and mitigate public health risks related to e-cigarette and cannabis use.
Supplementary Material
Contributor Information
Julia Vassey, Department of Population and Public Health Sciences, University of Southern California, Los Angeles, CA.
Julia Chen-Sankey, Institute for Nicotine and Tobacco Studies, Rutgers University, New Brunswick, NJ; School of Public Health, Rutgers University, Piscataway, NJ.
Jennifer B Unger, Department of Population and Public Health Sciences, University of Southern California, Los Angeles, CA.
Author Contributions
Julia Vassey (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Software, Validation, Visualization, Writing—original draft [lead], Resources, Supervision, Writing—review & editing [equal]), Julia Chen-Sankey (Writing—review & editing [equal]), Jennifer B. Unger (Supervision, Writing—review & editing [equal])
Funding
The Tobacco-Related Disease Research Program (TRDRP), T33DT6620. The content is solely the responsibility of the authors, and it does not necessarily represent the views or opinions of the Tobacco-Related Disease Research Program (TRDRP).
Declaration of Interests
None declared.
Data Availability
The data underlying this article will be shared on reasonable request with any qualified researcher.
Ethics Approval
The University of Southern California Institutional Review Board (UP-23-00810) approved all study procedures.
References
- 1. Davis DR, Bold KW, Kong G, Cavallo DA, Jackson A, Krishnan-Sarin S. Cannabis use among youth who vape nicotine E-cigarettes: a qualitative analysis. Drug Alcohol Depend. 2022;234:109413. 10.1016/j.drugalcdep.2022.109413. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Truth initiative: inspiring lives free from smoking, vaping, and nicotine. Accessed February 26, 2024. https://truthinitiative.org/
- 3. Park-Lee E, Jamal A, Cowan H, et al. Notes from the field: E-cigarette and nicotine pouch use among middle and high school students — United States, 2024. MMWR Morb Mortal Wkly Rep. 2024;73(35):774–778. 10.15585/mmwr.mm7335a3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Reported drug use among adolescents continued to hold below pre-pandemic levels in 2023. National Institutes of Health (NIH). 12:2023 Accessed February 26, 2024. https://www.nih.gov/news-events/news-releases/reported-drug-use-among-adolescents-continued-hold-below-pre-pandemic-levels-2023. [Google Scholar]
- 5. Do VV, Ling PM, Chaffee BW, et al. Concurrent use of tobacco and cannabis and internalizing and externalizing problems in US youths. JAMA Netw Open. 2024;7(7):e2419976. 10.1001/jamanetworkopen.2024.19976. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Do EK, Yoon SN, McKay T, et al. Cannabis and vaping nicotine: an exploration of risk factors using a nationally representative sample of youth and young adults. Addict Behav. 2024;157:108094. 10.1016/j.addbeh.2024.108094. [DOI] [PubMed] [Google Scholar]
- 7. Sun R, Mendez D, Warner KE. Use of electronic cigarettes among cannabis-naive adolescents and its association with future cannabis use. JAMA Netw Open. 2022;5(7):e2223277. 10.1001/jamanetworkopen.2022.23277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Lim CCW, Sun T, Hall W, Gartner C, Connor JP. Swipe to inhale: tobacco-related content on social media and susceptibility to tobacco use. Am J Prev Med. 2024;67(0):968–972. 10.1016/j.amepre.2024.07.020. [DOI] [PubMed] [Google Scholar]
- 9. Jung S, Murthy D, Bateineh BS, Loukas A, Wilkinson AV. The normalization of vaping on TikTok using computer vision, natural language processing, and qualitative thematic analysis: mixed methods study. J Med Internet Res. 2024;26(1):e55591. 10.2196/55591. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Sidoti MF, Teens O. Social Media and Technology 2024. Pew Research Center; 2024 Accessed December 16, 2024. https://www.pewresearch.org/internet/2024/12/12/teens-social-media-and-technology-2024/. [Google Scholar]
- 11. Topic: Social media and Generation Z in the United States. Statista; Accessed March 1; 2024: https://www.statista.com/topics/10943/social-media-and-generation-z-in-the-united-states/. [Google Scholar]
- 12. Vassey J, Chen-Sankey J, Unger JB. E-cigarette and cannabis influencer marketing on social media despite regulations. Blog - Tob Control. 2024;3: Accessed June 3, 2024. https://blogs.bmj.com/tc/2024/06/03/e-cigarette-and-cannabis-influencer-marketing-on-social-media-despite-regulations/. [Google Scholar]
- 13. Micro vs. Macro Influencers: The Future of Marketing in 2024. ContentBASE. December 9; 2024: Accessed December 10, 2024. https://contentbase.com/blog/future-influencer-marketing/. [Google Scholar]
- 14. Vassey J, Valente T, Barker J, et al. E-cigarette brands and social media influencers on Instagram: a social network analysis. Tob Control. 2023;32(e2):e184–e191. 10.1136/tobaccocontrol-2021-057053. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Nguyen N, Islam S, Llanes KD, Koester KA, Ling PM. Classification of patterns of tobacco and cannabis co-use based on temporal proximity: a qualitative study among young adults. Addict Behav. 2024;152:107971. 10.1016/j.addbeh.2024.107971. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Vassey J, Kennedy CJ, Chang HCH, Unger JB. Generative AI in a new era of computer model-informed tobacco research: a short report. Tob Control Published online January. 2025;30:tc-2024-058887. 10.1136/tc-2024-058887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Tadayon Nabavi R, Bijandi M. Bandura’s Social Learning Theory & Social Cognitive Learning Theory. 2023.
- 18. Wu J, Harlow AF, Wijaya D, et al. The impact of influencers on cigar promotions: a content analysis of large cigar and swisher sweets videos on TikTok. Int J Environ Res Public Health. 2022;19(12):7064. 10.3390/ijerph19127064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Childers CC, Lemon LL, Hoy MG. #sponsored #ad: agency perspective on influencer marketing campaigns. J Curr Issues Res Advert. 2019;40(3):258–274. 10.1080/10641734.2018.1521113. [DOI] [Google Scholar]
- 20. Vassey J, Vogel EA, Unger JB, et al. Impact of Instagram and TikTok influencer marketing on perceptions of e-cigarettes and perceptions of influencers in young adults: a randomised survey-based experiment. Tob Control Published online February. 2025;13:tc-2024-059021. 10.1136/tc-2024-059021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Gibbons FX, Gerrard M. Predicting young adults’ health risk behavior. J Pers Soc Psychol. 1995;69(3):505–517. 10.1037/0022-3514.69.3.505. [DOI] [PubMed] [Google Scholar]
- 22. Howell JL, Ratliff KA. Investigating the role of implicit prototypes in the prototype willingness model. J Behav Med. 2017;40(3):468–482. 10.1007/s10865-016-9814-y. [DOI] [PubMed] [Google Scholar]
- 23. ADVANCE Study – Epidemiology of Substance Use (EOS). Accessed October 7, 2022. https://eosresearch.usc.edu/advance-study/.
- 24. Donaldson SI, Dormanesh A, Escobedo P, Majmundar A, Kirkpatrick M, Allem JP. The impact of e-cigarette product place in music videos on susceptibility to use e-cigarettes among young adults: an experimental investigation. Addict Behav. 2022;130:107307. 10.1016/j.addbeh.2022.107307. [DOI] [PubMed] [Google Scholar]
- 25. PATH (Population Assessment of Tobacco and Health) Study - Home. Accessed October 1, 2023. https://pathstudyinfo.nih.gov/.
- 26. Cheng HG, Lizhnyak PN, Knight NA, Vansickel AR, Largo EG. Youth susceptibility to tobacco use: is it general or specific? BMC Public Health. 2021;21(1):1913. 10.1186/s12889-021-11956-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Vogel EA, Unger JB, Vassey J, Barrington-Trimis JL. Effects of a nicotine warning label and vaping cessation resources on young adults’ perceptions of pro-vaping instagram influencer posts. Addict Behav. 2024;149:107888. 10.1016/j.addbeh.2023.107888. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Reeves B, Yeykelis L, Cummings JJ. The use of Media in Media Psychology. Media Psychol. 2016;19(1):49–71. 10.1080/15213269.2015.1030083. [DOI] [Google Scholar]
- 29. Kong G, Laestadius L, Vassey J, et al. Tobacco promotion restriction policies on social media. Tob Control. 2022;33(3):398–403. 10.1136/tc-2022-057348. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Restricted Goods and Services| Transparency Center Accessed March 26, 2024. https://transparency.fb.com/policies/community-standards/restricted-goods-services/.
- 31. Branded Content Policy | TikTok. Accessed March 26, 2024. https://www.tiktok.com/legal/page/global/bc-policy/en
- 32. Announcing our new CBD and related products advertising policy. In: Meta for Business. Accessed December 10, 2024. https://www.facebook.com/business/news/announcing-our-new-cbd-and-related-products-advertising-policy.
- 33. Advertiser-friendly content guidelines - YouTube Help. Accessed December 10, 2024. https://support.google.com/youtube/answer/6162278?hl=en#Recreational_drugs&zippy=%2Cpolicy-detail.
- 34. Ouellette RR, Selino S, Kong G. Electronic nicotine delivery systems and E-liquid modifications to vape cannabis depicted in online videos. JAMA Netw Open. 2023;6(11):e2341075. 10.1001/jamanetworkopen.2023.41075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Rutherford BN, Sun T, Lim CCW, et al. Changes in viewer engagement and accessibility of popular vaping videos on TikTok: a 12-month prospective study. Int J Environ Res Public Health. 2022;19(3):1141. 10.3390/ijerph19031141. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Vassey J, Donaldson SI, Dormanesh A, Allem JP. Themes in TikTok videos featuring little cigars and cigarillos: content analysis. J Med Internet Res. 2022;24(11):e42441. 10.2196/42441. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Vassey J, Kennedy CJ, Herbert Chang HC, Smith AS, Unger JB. Scalable surveillance of E-cigarette products on Instagram and TikTok using computer vision. Nicotine Tob Res. 2024;26(5):552–560. 10.1093/ntr/ntad224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Saxena A, Singh P, Narayan Singh S. Gender and age detection using deep learning. In: 2021 11th International Conference on Cloud Computing, Data Science & Engineering (Confluence). 2021:719–724. 10.1109/Confluence51648.2021.9377041 [DOI] [Google Scholar]
- 39. Gell S, Engel E, Heiss R, et al. Social media influencers and adolescents’ health: a scoping review of the research field. Soc Sci Med. 2023;116387. 10.1016/j.socscimed.2023.116387. [DOI] [PubMed] [Google Scholar]
- 40. FDA In brief: FDA requires four E-cigarette brands to provide critical information on social media practices. In: FDA. Published online August 9, 2024. Accessed December 16, 2024. https://www.fda.gov/news-events/fda-brief/fda-brief-fda-requires-four-e-cigarette-brands-provide-critical-information-social-media-practices.
- 41. The FTC releases new guidelines (and its first creator-friendly brochure) for influencers who post sponsored content. Tubefilter 2019; Accessed February 28, 2024. https://www.tubefilter.com/2019/11/06/the-ftc-releases-new-guidelines-and-its-first-creator-friendly-brochure-for-influencers-who-post-sponsored-content/. [Google Scholar]
- 42. truthorange - YouTube. Accessed June 1, 2022. https://www.youtube.com/user/truthorange/featured.
- 43. Projections of the Size and Composition of the U.S. Population: 2014 to 2060. Census.gov. Accessed February 22, 2024. https://www.census.gov/library/publications/2015/demo/p25-1143.html
- 44. Klein EG, Czaplicki L, Berman M, Emery S, Schillo B. Visual attention to the use of #ad versus #sponsored on e-cigarette influencer posts on social media: a randomized experiment. J Health Commun. 2020;25(12):925–930. 10.1080/10810730.2020.1849464. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article will be shared on reasonable request with any qualified researcher.