The impact of cannabis packaging characteristics on perceptions and intentions

Abstract

Introduction:

As cannabis increasingly becomes a consumer product in the United States, its product packaging has become critically important to regulators. This study examined the influence of recreational cannabis packaging characteristics.

Methods:

Five online between-subjects experiments were conducted in April 2021 and data were analyzed May-July 2021. Experiments randomized participants to view different 1) types of cannabis, 2) visual displays of THC content, 3) cannabis packages designed around brand personality research, 4) health warnings, and 5) health claims. Outcomes included cognitive, affective, and behavioral responses.

Results:

A total of 841 adults from the United States (49% male; 50% young adults; 44% white; 17% Hispanic) were included in the study. Edible gummies were perceived as healthier (β=.32, 95% CI .03, .62), less grown-up (β=−.58, CI −.86, −.28), and more socially acceptable to consume (β=.30, CI .01, .59) than cannabis concentrate in a medical dropper. Participants also had more interest in trying edible gummies (β=1.33, CI 1.04, 1.62) and trying a free sample (β=1.30, CI 1.01, 1.60) than cannabis concentrate. Cannabis packages with a “helps you relax” health claim elicited more happy (β=.34, CI .04, .64) and good feelings (β=.37, CI .07, .67) than cannabis packages without this claim. Minimal effects were found for visual displays of THC content and health warnings.

Conclusions:

Edibles are a unique type of cannabis that should be given special consideration under state laws, and lawmakers should consider limiting and governing the use of both implicit and explicit health claims on recreational cannabis packages when implementing laws.

1. INTRODUCTION

A growing number of states in the United States (U.S.) have legalized the recreational use of cannabis, despite its federal schedule I status. As state-level legalization of recreational cannabis has expanded across the country, so has its use.¹ Cannabis is currently the most commonly used drug after tobacco and alcohol, and in 2019, 18% of people in the U.S. over the age of 12 reported using cannabis in the past year and 12% reported using cannabis in the past month.² Use is even higher among young adults (35% for past-year use and 23% for past-month use).²

Despite evolving research that has studied the potential benefits and harm of cannabis use, no products are without risk. Despite some known benefits such as relief from nausea and chronic pain,³ research shows that between 10-30% of cannabis users may develop some degree of cannabis use disorder characterized by dependence^{4, 5} and that disorders are most likely for people who begin using cannabis before the age of 18.⁶ Cannabis use is also associated with: impaired brain development, respiratory problems when cannabis is smoked, increased risk of motor vehicle crashes, lower birth weight when cannabis is used during pregnancy, unintentional ingestion by children, increased risk for development of schizophrenia or other psychoses, and adverse events leading to hospitalization.^{3, 7-9}

The cultivation, harvesting, sale, possession, and use of cannabis remains illegal under federal law, except for very narrow research purposes; therefore, each state has been faced with making their own cannabis regulations. For example, only some states require a symbol visually communicating that packages contains cannabis;¹⁰ only some states have specified criteria for determining what constitutes false or misleading packaging claims;¹¹ and states have different requirements for the content, length, and formatting of text warnings.¹⁰ In addition, while state laws require that the total amount of tetrahydrocannabinol (THC, the main psychoactive compound in cannabis responsible for the 'high' or intoxication) and number of servings be listed on a package, there is no requirement for visually depicting the strength of THC content in packages.^{10, 12}

Research on other legal substances, like tobacco, suggests that 1) lighter color packs denote “lighter” tasting smoke and less harm,^{13, 14} 2) package shape, size, color, and name communicate brand imagery, which is used to foster connections to the product and consumption,^15-17 3) health descriptors (e.g., low tar) increase favorable perceptions of the product and can mislead consumers on product safety,¹³ 4) flavor descriptors increase product appeal and use,¹⁸ and 5) large pictorial warnings can motivate people to quit and encourage cessation.¹⁹ Research on alcohol confirms many of these findings.^20-22

As recreational cannabis increasingly becomes a consumer product like alcohol or tobacco, its product packaging has become critically important to regulators. Yet, there is a lack of information regarding how people perceive cannabis packaging, which hinders states’ abilities to develop effective regulations. A few studies have found that: 1) edible packages are perceived as more appealing than other types of cannabis packages,²³ 2) a visualization of THC content on packages can help consumers correctly identify THC levels,²⁴ 3) cannabis packages denoting associations with different lifestyles (e.g., showing music references or celebrities) can increase product appeal,²⁵ 4) warnings on cannabis packages can educate consumers on product harms and decrease product appeal,^{23, 26} especially if they are pictorial²⁷ and 5) packages with health claims can decrease product ham perceptions.²⁵ However, most of these studies have been conducted outside the U.S. To address gaps in previous research, an exploratory study was conducted to examine the influence of recreational cannabis packaging characteristics on product appeal in the U.S. Because this study was exploratory, a-priori hypotheses were not set.

2. METHODS AND MATERIALS

2.1. Participants

This experimental study was conducted using an online survey panel provider, Qualtrics Research Services. To be eligible for participation, adults had to be 18 years or older, speak English, and live in the U.S. Participants could reside in any U.S. state, regardless of whether that state has legalized medical or recreational cannabis, and participate regardless of whether they used cannabis. The inclusion criteria were intentionally broad because many people see cannabis packaging, regardless of where they live in and whether they use cannabis. Accordingly, the goal of this study was to assess perceptions among a wide sample of people. Qualtrics recruited a final sample of 841 participants in April 2021. The original goal was to recruit 500 participants based on a power calculation demonstrating that this would allow us to detect small to medium effect sizes (Cohen’s d = 0.28). The sample size was increased due to additional funds.

2.2. Procedures

Five randomized, between-subjects experiments were conducted to manipulate different elements of cannabis packaging (see Supplementary Table A for all stimuli). Each participant therefore saw five different packages and answered questions about each package. A professional graphic designer designed the cannabis packages. To avoid associations with real cannabis brands, a fake brand “Carrboro Farms” was created. To create the packages, the study team reviewed existing cannabis packages sold from different retail stores and online. The study team also reviewed examples of tobacco packaging, especially for the experiments manipulating the visualization of THC content and health warnings.

The five experiments were:

1. Product type: Participants were randomized to view one of three cannabis packages: cannabis concentrate in a medical dropper, dried cannabis in a glass container, or edible gummies.

2. Visualization of THC content: Participants were randomized to see one of four cannabis edible packages. Three packages included visual displays of THC content level (e.g., a thermometer dial to indicate how much THC was included in the package) and one package was a control condition with no visual display of THC content.

3. Types of imagery based on brand personality theory: Based on different brand personalities from marketing theory,²⁸ participants were randomized to see one of three cannabis edible packages modified to make the product look medical (portrayed with a green cross), natural (with outdoor scenery), or fun/adventurous (with interesting fonts and bright colors).

4. Health warning: Participants were randomized to view one of three cannabis edible packages with different health warnings. One warning was about addiction, one warning was about risks to driving under the influence, and one warning was about risks to driving under the influence and included an image of a car crash. The text for the warnings was selected because of its real-world use in Alaska.

5. Health claim: Participants were randomized to view one of three cannabis edible packages modified with a health claim (“helps you relax” or “relax”) or a control condition package with no claim. These two health claims were selected because they commonly appear on cannabis packages marketed in the U.S.

All participants saw the product type experiment first, the visualization of THC content experiment second, and so on. Participants could view the package for as long as they wanted. Incentives were paid directly by Qualtrics. The University of North Carolina at Chapel Hill Institutional Review Board approved the study.

2.3. Measures

Following each experiment, the survey assessed outcomes guided by the Context of Consumption Framework, which provides a theoretical orientation to consumer responses to product visual design and posits cognitive, affective, and behavioral responses to packaging (see Supplemental Table B for all measures and references).²⁹ The framework has previously been utilized to inform tobacco packaging research,³⁰ including with the design and evaluation of packaging experiments³¹ and with the interpretation of focus group data examining packaging perceptions.³²

For cognitive responses, four items assessed semantic responses (i.e., factual content conveyed by the product’s design), two items assessed aesthetic responses (i.e., the visual appeal of the design), and three items to assess symbolic responses (i.e., the design’s connection with individual identity and style) domains. Two items assessed affective responses (e.g., how good people felt while viewing packages). Two items assessed behavioral responses (e.g., interest in trying cannabis featured in the package). For the health warning experiment, one item assessed warning effectiveness and three items from a validated scale assessed reactance.³³ Finally, for the health claim experiment, one item assessed the impact of the health claim by asking how much people thought the cannabis featured in the package would help them relax. All survey items were assessed with 4-, 5-, or 6-point response scales.

2.4. Data analysis

Several steps were taken to ensure data quality. First, participants were asked if they were committed to providing thoughtful and honest answers to questions in the survey and removed participants who reported that they would not provide their best answers. Second, participants were prevented from giving multiple submissions. Third, the survey panel provider removed straight-liners and outliers on how long it took to complete the survey.

All analyses were conducted in Stata (v. 17.0) and used ordinal regressions. Models controlled for the conditions of prior experiments. For example, for the analysis of the second experiment on visualization of THC content, the model controlled for the condition in the first experiment that participants were assigned to. Missing data were less than <5% in each of all model variables; therefore, multiple imputation would be unlikely to affect current parameter estimates,³⁴ so all missing data were treated as random and list-wise deleted. Instead of presenting p-values, which do not provide information on direction or effect size,³⁵ results include ordered logit regression coefficients and 95% confidence intervals. Non-overlapping confidence intervals were interpreted as p <.05.³⁶ The analysis was not pre-registered, and the results should be considered exploratory. Because this study was exploratory, models did not adjust for having multiple dependent variables or multiple comparisons.^37-39

3.0. RESULTS

3.1. Participant characteristics

The sample was diverse in terms of gender (49% male), age (50% young adults), race (44% white, 34% Black or African American), ethnicity (17% Hispanic), and education (41% reported having a high school degree/GED or less) (Supplemental Table C). In terms of cannabis use, 47% reported ever use, 33% reported past year use, and 27% reported past 30-days use.

3.2. Product type

Descriptive statistics for each experimental condition can be seen in Supplemental Tables D-H. Compared with cannabis concentrate in the medical dropper, dried cannabis was perceived as more harmful (β=.31, 95% CI .10, .61) and edible gummies were perceived as healthier (β=.32, 95% CI .03, .62) (Table 1). The edible gummies were perceived as less grown-up (β=−.58, 95% CI −.86, −.28), but more socially acceptable to consume (β=.30, 95% CI .01, .59) than cannabis in the medical dropper. Further, edible gummies evoked higher feelings of interest in trying the product (β=1.33, 95% CI 1.04, 1.62) and higher interest in trying a free sample of the product (β=1.30, 95% CI 1.01, 1.60), compared to cannabis concentrate.

Table 1.

Differences by product type, N=841

	Experiment 1 Conditions: Product Type
Outcomes	Cannabis concentrate in a medical dropper	Dried cannabis β [95% CI]	Edible gummies β [95% CI]
Cognitive responses: Semantic domain
Perceived relative harm	Ref	.31 [.10, .61]	−.01 [−.28, .31]
Perceived relative healthfulness	Ref	−.06 [−.36, .24]	.32 [.03, .62]
Perceived quality	Ref	.10 [−.20, .39]	.26 [−.03, .56]
Perceived harm	Ref	−.03 [−.33, .27]	−.06 [−.35, .24]
Cognitive responses: Aesthetic domain
Looks expensive	Ref	.24 [−.05, .53]	−.41 [−.70, −.12]
Looks cool	Ref	.02 [−.26, .31]	.10 [−.20, .39]
Cognitive responses: Symbolic domain
Meant for someone like me	Ref	.06 [−.23, .35]	.17 [−.12, .45]
Looks grownup	Ref	.07 [−.21, .36]	−.58 [−.86, −.28]
Socially acceptable to consume	Ref	.21 [−.08, .50]	.30 [.01, .59]
Affective responses
Felt good while viewing the package	Ref	.03 [−.25, .32]	.17 [−.12, .46]
Felt happy while viewing the package	Ref	.20 [−.09, .49]	.26 [−.03, .55]
Behavioral responses
Interest in trying	Ref	.13 [−.17, .43]	1.33 [1.04, 1.62]
Interest in free sample	Ref	.21 [−.08, .51]	1.30 [1.01, 1.60]

Boldface indicates statistical significance p<0.05

CI = Confidence Interval

3.3. Visualization of THC content

No significant findings were found regarding the manipulations of THC content visuals (Table 2).

Table 2.

Differences by visualization of THC level, N=841

	Experiment 2 Conditions: Visualization of THC content
Outcomes	No visual display of THC content	Flowers demonstrating THC content β [95% CI]	Thermometer demonstrating THC content β [95% CI]	Dial demonstrating THC content β [95% CI]
Cognitive responses: Semantic domain
Perceived relative harm	Ref	−0.18 [−.52, .15]	−.05 [−.30, .39]	.16 [−.51, .18]
Perceived relative healthfulness	Ref	.16 [−.17, .50]	−.06 [−.41, .27]	.06 [−.28, .40]
Perceived quality	Ref	−.02 [−.35, .32]	.03 [−.31, .37]	.08 [−.26, .42]
Perceived harm	Ref	.01 [−.33, .35]	−.14 [−.49, .20]	.07 [−.28, .41]
Cognitive responses: Aesthetic domain
Looks expensive	Ref	−.08 [−.41, .25]	−.11 [−.44, .23]	.04 [−.29, .38]
Looks cool	Ref	.05 [−.28, .38]	.10 [−.24, .44]	−.11 [−.45, .22]
Cognitive responses: Symbolic domain
Meant for someone like me	Ref	.19 [−.14, .52]	.03 [−.31, .36]	.16 [−.17, .50]
Looks grownup	Ref	−.10 [−.43, .23]	.15 [−.19, .49]	.03 [−.31, .36]
Socially acceptable to consume	Ref	.02 [−.31, .35]	−.14 [−.48, .20]	−.03 [−.37, .30]
Affective responses
Felt good while viewing the package	Ref	.05 [−.27, .38]	−.08 [−.42, .25]	−.10 [−.44, .24]
Felt happy while viewing the package	Ref	.10 [−.23, .43]	−.002 [−.34, .33]	−.03 [−.37, .30]
Behavioral responses
Interest in trying	Ref	−.004 [−.34, .33]	−.01 [−.35, .34]	.09 [−.25, .43]
Interest in free sample	Ref	−.13 [−.46, .20]	−.14 [−.48, .21]	−.09 [−.44, .24]

Note: Boldface indicates statistical significance p<0.05. Models reported for Experiment 2 control for condition of Experiment 1.

CI = Confidence Interval

3.4. Types of imagery based on brand personality theory

Cannabis packaged to look “natural” was perceived as more harmful than cannabis packaged to look “medical” (β=.46, 95% CI .07, .66) (Table 3). Cannabis packaged to look “fun/adventurous” was perceived as less expensive (β=−.29, 95% CI −.58, .003) and less grown up (β=−.40, 95% CI −.70, −.10) than cannabis packaged to look “medical.”

Table 3.

Differences by types of imagery based on brand personality theory, N=841

	Experiment 3 Conditions: Brand Personalities
Outcomes	Medical	Natural β [95% CI]	Fun / Adventurous β [95% CI]
Cognitive responses: Semantic domain
Perceived relative harm	Ref	.46 [.07, .66]	−.01 [−.29, .31]
Perceived relative healthfulness	Ref	−.15 [−.44, .14]	−.08 [−.21, .38]
Perceived quality	Ref	−.12 [−.41, .17]	−.11 [−.41, .19]
Perceived harm	Ref	−.10 [−.40, .20]	−.24 [−.55, .06]
Cognitive responses: Aesthetic domain
Looks expensive	Ref	−.15 [−.45, .14]	−.29 [−.58, .003]
Looks cool	Ref	.02 [−.27, .31]	−.004 [−.30, .29]
Cognitive responses: Symbolic domain
Meant for someone like me	Ref	−.003 [−.29, .29]	−.22 [−.51, .07]
Looks grownup	Ref	.02 [−.27, .30]	−.40 [−.70, −.10]
Socially acceptable to consume	Ref	.01 [−.28, .30]	−.10 [−.39, .19]
Affective responses
Felt good while viewing the package	Ref	−.11 [−.40, .18]	−.22 [−.52, .07]
Felt happy while viewing the package	Ref	−.05 [−.33, .24]	−.12 [−.41, .18]
Behavioral responses
Interest in trying	Ref	.07 [−.22, .36]	−.07 [−.36, .23]
Interest in free sample	Ref	.05 [−.24, .34]	−.05 [−.35, .24]

Note: Boldface indicates statistical significance p<0.05. Models reported for Experiment 3 control for condition of Experiments 1 and 2.

CI = Confidence Interval

3.5. Health warning

Cannabis packages with the driving under the influence warning (β=.41, 95% CI .10, .71) and the driving under the influence plus image warning (β=.31, 95% CI .01, .60) were perceived as more grown-up than the cannabis package with the addiction warning (Table 4). No other significant differences were detected.

Table 4.

Differences by health warnings, N=841

	Experiment 4 Conditions: Health Warnings
Outcomes	Addiction	Driving under the influence β [95% CI]	Driving under the influence warning + image β [95% CI]
Cognitive responses: Semantic domain
Perceived relative harm	Ref	−.15 [−.15, .45]	−.15 [−.15, .45]
Perceived relative healthfulness	Ref	.15 [−.14, .45]	−.004 [−.30, .29]
Perceived quality	Ref	.11 [−.19, .41]	−.13 [−.43, .16]
Perceived harm	Ref	.08 [−.22, .39]	.22 [−.08, .52]
Warning effectiveness	Ref	−.05 [−.35, .25]	.07 [−.22, .37]
Reactance	Ref	−.02 [−.44, .09]	.33 [−.06, .47]
Cognitive responses: Aesthetic domain
Looks expensive	Ref	.01 [−.29, .31]	−.10 [−.40, .19]
Looks cool	Ref	−.003 [−.30, .29]	−.07 [−.36, .22]
Cognitive responses: Symbolic domain
Meant for someone like me	Ref	−.05 [−.35, .24]	−.18 [−.47, .11]
Looks grownup	Ref	.41 [.10, .71]	.31 [.01, .60]
Socially acceptable to consume	Ref	.05 [−.25, .35]	−.14 [−.43, .15]
Affective responses
Felt good while viewing the package	Ref	−.01 [−.30, .29]	.02 [−.27, .32]
Felt happy while viewing the package	Ref	−.02 [−.32, .27]	.11 [−.41, .18]
Behavioral responses
Interest in trying	Ref	.20 [−.11, .51]	.04 [−.26, .34]
Interest in free sample	Ref	.13 [−.18, .43]	−.01 [−.31, .29]

Note: Boldface indicates statistical significance p<0.05. Models reported for Experiment 4 control for condition of Experiments 1, 2, and 3.

CI = Confidence Interval

3.6. Health claim

The cannabis package with the “helps you relax” health claim elicited happier (β=.34, 95% CI .04, .64) and more good feelings (β=.37, 95% CI .07, .67) than the cannabis package without a health claim (Supplemental Table I).

4.0. DISCUSSION

This study found strong effects of type of cannabis product on cognitive and behavioral responses. Specifically, participants reported edible gummies to be healthier and more socially acceptable to consume than cannabis concentrate and were more interested in trying them. These findings align with research showing that: people often perceive edibles to be more appealing than other types of cannabis,²³ consumers like edibles because they are convenient, discreet, and tasty,⁴⁰ and edibles are an increasingly popular product in states that have legalized cannabis.⁴¹ While edibles may not pose the same risks to respiratory health as smoking cannabis,⁴² there are two concerns with these products. First, it can be difficult for consumers to estimate the proper serving size to consume if edibles are sold in packages containing more than one serving. Since the resulting intoxication may not be felt right away, a consumer may ingest more of the product than is recommend. This overconsumption may result in the consumer experiencing more adverse side effects, such as panic and paranoia.⁴³ Second, and perhaps more importantly, edibles can appeal to children and lead to accidental ingestion if they do not realize that the item is not a “regular” food product.⁹ While the current study found that participants viewed edibles as more youth oriented (i.e., less grown-up) than cannabis concentrate, future research should be conducted with minors, including experiments and longitudinal research, to examine the appeal of edible cannabis packaging and advertising. This research is especially urgent given increasing cannabis legalization and the need for evidence-based laws.

Although minimal effects of the health claim manipulation were observed on cognitive and behavioral outcomes, consistent effects were found on affective responses. Many states have prohibited cannabis establishments from false or misleading advertising, however, rarely are clear criteria specified for what this entails. Unsurprisingly then, many cannabis products explicitly contain or strongly suggest health claims that are not supported by evidence. In one study, 66% of online cannabis retailers in Colorado and 44% in Washington made health claims about the benefits of cannabis⁴⁴ and common health claims noted the benefits of cannabis for reducing anxiety and treating depression, insomnia, and pain. Cannabis use can have certain therapeutic effects when used for medical purposes, however, the claims made by many cannabis manufacturers that are selling or distributing products for non-medical, recreational purposes are not supported by research.³ The current study suggests that more research is needed on the impact of health claims on adult-use recreational packages, given that even a small, one-time exposure to a health claim altered participants’ affective perceptions.

This study also found some effects of the manipulations of brand personality on outcomes.²⁸ Packages were designed to either look medical, natural, or fun/adventurous and results suggest that the medical package looked less harmful than the natural package and more expensive and grown-up than the fun/adventurous package. Inherent in the medical packaging is an implicit claim that the product carries medical benefits, which could explain why people perceived the cannabis in this packaging to look less harmful. This finding is important because it suggests that even in the absence of an explicit claim, images used on cannabis packaging can imply health benefits. For products meant to be sold to medical cannabis patients, this might be acceptable if these medical claims are supported by clear scientific evidence. However, state lawmakers should consider whether they wish to allow any implied or explicit medical claims or potential health benefits to be included on products designated for adult-use recreational consumption. Research from tobacco suggests that implicit messages can influence intentions;⁴⁵ therefore, findings from the current study suggest that regulators should consider both explicit and implicit ways in which cannabis manufacturers make claims about product benefits.

Prior research shows that cannabis packages with a warning can educate consumers about product harms and decrease product appeal,^{23, 26} especially if they are pictorial.²⁷ Contrary to prior research, the current study found minimal effects of different warnings on study outcomes, which could be due to several reasons. First, the driving under the influence warning was lengthy (2 sentences with a total of 16 words) and written at a college reading level, which may have made it too difficult for most participants to quickly read and comprehend. Research from tobacco warning labels indicates that shorter warnings may be more effective than longer ones and that language should be at a sixth-grade level or lower.¹¹ Similarly, the image of the car crash may have been too small for people to see. The current study also did not include a true control condition of a cannabis package without any warning, which contrasts with previous research.^{23, 26} Future research could examine if and how people notice warnings on cannabis packages using objective measures of attention (e.g., eye tracking) in addition to self-report measures.

Finally, no effects were found for the manipulation of THC content visuals. Research shows that consumers have little understanding of THC labels.⁴⁶ Only one previous study to the authors’ knowledge has manipulated THC levels in cannabis packages and found that a traffic light system for labeling THC content (i.e., green for low potency and red for high potency) helped people correctly identify THC levels.²⁴ In contrast to this previous study, the current experiment only assessed the impact of THC visualization on aesthetics and appeal and did not assess whether participants understood the labeling. Given the nascence of cannabis legalization, more research is needed on developing appropriate standards for communicating THC potency in edible packages.

4.1. Limitations

Strengths of this study include use of a randomized experimental design, outcome measures that were guided by the Context of Consumption Framework,²⁹ and use of high-quality stimuli developed by a professional graphic designer. Limitations include 1) the use of a one-time experiment that precluded measurement of future behavior and 2) the use of a convenience sample, although well-designed experimental studies using convenience samples can yield results that are consistent with those found in representative samples.^{47, 48} In addition, since this was an exploratory study, models did not adjust for multiple testing, which is in line with recommended best practices^37-39 and subgroup analyses were not conducted (e.g., cannabis users vs. non-users) since a-priori hypotheses were not set for how different subgroups would compare to one another.

5.0. CONCLUSIONS

As states and countries increasingly legalize recreational cannabis, this study provides useful preliminary data to lawmakers on how to design packaging regulations.