Associations of Recreational Cannabis Dispensaries’ Availability, Storefront Signage, and Health Benefit Signs with Cannabis Use: Findings from a Representative Adult Sample in California, United States

Abstract

Background and Aims.

There are concerns that availability of recreational cannabis dispensaries (RCDs) and point-of-sale marketing may lead to increased cannabis use in jurisdictions where cannabis retail sale is legal. This paper examined whether RCD’s availability and presence of storefront signage indicative of cannabis and signs promoting health benefits in RCDs were associated with cannabis use and risk perceptions.

Design.

Cross-sectional, secondary data analysis.

Setting.

California, USA.

Participants.

A representative sample of 3,385 adults in California who participated in the 2020 probability-based Marijuana Use and Environment Survey.

Measurements.

Binary outcomes included past-month overall cannabis use, perceiving cannabis smoking as harmful, and past-month cannabis use by purpose (medical only, recreational only, and dual). The objectively assessed predictors included proximity and density of RCDs and presence of storefront signage indicative of cannabis and signs promoting health benefits in RCDs.

Findings.

In terms of proximity, presence of storefront signage in the nearest RCD was associated with smaller odds of perceiving cannabis smoking as harmful (odds ratio [OR]=0.63, 95% confidence interval [CI]: 0.39, 0.99) if the RCD was located within 2 miles of homes. Presence of health benefit signs in the nearest RCD was associated with greater odds of overall cannabis use (OR=2.45, 95%CI: 1.17, 5.16) and recreational use (OR=3.50, 95%CI: 1.027, 11.91) if the RCD was located within 2–4 miles of homes. In terms of density, count of RCDs, count of RCDs with storefront signage, and count of RCDs with health benefit signs within 2 miles of homes were each separately associated with greater odds of overall cannabis use and cannabis use for dual purposes.

Conclusions.

The availability of recreational cannabis dispensaries within 2 miles of one’s home and the presence of storefront signage indicating the availability of cannabis and signs promoting health benefits of cannabis appear to be associated with increased cannabis use and reduced risk perceptions among adults in California, USA.

INTRODUCTION

Since 2012, 18 states and Washington D.C. in the United States (U.S.) have legalized cannabis for adult use. Among these jurisdictions, most also approved retail sale of cannabis in recreational cannabis dispensaries (RCDs). There are concerns that the surge in the number of RCDs following cannabis commercialization may lead to greater prevalence of cannabis use and problem use. Some emerging evidence provided support to these concerns, showing that statewide cannabis commercialization was linked to increased cannabis exposure and cannabis use disorders among adults. (1, 2)

Studies on statewide policies may provide limited insights regarding the influences of local variations in retail environments. Within a state with recreational cannabis legalization and commercialization, local jurisdictions may opt to prohibit RCDs or restrict the density and location of RCDs. RCDs may also self-select their locations for considerations such as cost, demand, and competition. These local variations may result in considerable differences in proximity and density of RCDs across neighborhoods. (3, 4)

There are several pathways through which RCDs in neighborhoods may influence individual cannabis use. First, a closer proximity and a greater density of RCDs will reduce travel costs and make cannabis more available and accessible. Second, the large variety of products in RCDs, particularly novel and youth-appealing products such as vaping devices and edibles, may promote interest in and initiation of cannabis use and lower success rate of quit attempts. (5, 6) Third, the simple presence of RCDs in neighborhoods may increase visibility, awareness, and social acceptance of cannabis. Fourth, exposure to point-of-sale marketing activities in RCDs may shape favorable perceptions, attitudes, and behaviors.

Research investigating the influences of RCDs in neighborhoods remains limited. A study in Oregon with data during 2010–2018 reported positive association between density of RCDs in counties and cannabis use among adolescents. (7) Another study in Oregon with 2017 data reported that presence of RCDs within 1-mile from school was positively associated with cannabis use and negatively associated with perceived harm among adolescents. (8) A study in California with 2018 data reported that a closer proximity and a greater density of RCDs were associated with a higher likelihood of cannabis use among pregnant women. (9) A study in Los Angeles County with data during 2017–2019 reported positive associations of RCD’s density within a 4-mile radius of individual homes with cannabis use, heavy use, and cannabis use disorders among young adults. (10) A study in Washington with data during 2015–2019 reported positive association between RCD’s presence within 1- and 2-kilometer of individual homes and cannabis use among young adults. (11) These findings were generally consistent with most studies on medical cannabis dispensaries, which found positive associations between proximity or density of dispensaries and cannabis use. (12–21)

This study aimed to examine the associations of RCD’s availability and point-of-sale marketing activities with cannabis use outcomes in a representative sample of adults in California, U.S. We expected to contribute to literature on the influences of neighborhood cannabis retail environments in the following ways (the citations below include studies on both medical cannabis dispensaries and RCDs). First, most studies assessed data at aggregate level (zip code, city, county, etc.) because detailed address of individuals was often unavailable. (7, 12–19, 22–25) Such studies were unable to control for heterogeneities across individuals or account for the fact that dispensaries located in close proximity have larger impacts than dispensaries located further away. We obtained detailed address of individual homes, derived precise distance between dispensaries and homes, and conducted individual-level analysis.

Second, almost all the individual-level studies used convenience samples except for some school-based ones. (9–11, 21, 23) We recruited a probability-based sample that was representative of adults in California, the state with the longest history of medical cannabis legalization and currently the largest cannabis retail market in the U.S. The diverse demographic representations may make the findings more generalizable.

Third, most studies relied upon either license directories or crowdsourcing websites to identify dispensaries, among which only a few conducted validation. (10, 21, 25) License directories do not capture unlicensed dispensaries whereas crowdsourcing websites include inactive businesses. (26, 27) We used a validated approach to construct and verify a comprehensive list of both licensed and unlicensed RCDs. (26)

Fourth, for the first time we examined the associations between RCDs’ point-of-sale marketing and cannabis use outcomes. Dispensary storefront became the most commonly reported source of marketing after recreational cannabis commercialization, (28, 29) but the research on its impacts is almost nonexistent. Point-of-sale marketing has been widely examined in tobacco and alcohol research and suggested to be associated with harm perceptions and behaviors. (30–32) Among various marketing activities, our study specifically focused on storefront signage indicative of cannabis and signs promoting health benefits of cannabis. Storefront signage is a type of advertisement noticeable from outside (advertises that the business sells cannabis). Storefront advertising conveys normative messages, shapes positive beliefs, and links products to desirable personal traits. (33–35) A high density of stores in neighborhoods not only makes substances easily accessible but also promotes a local culture through individuals’ constant exposure to storefronts. (36) Storefront advertising was associated with increased tobacco and alcohol consumption and substance use disorders. (37–39) Regarding advertising through health claims, exposure to generalized positive statements may frame users’ favorable responses and use expectations. (40) Tobacco research indicated that health claims were associated with decreased risk beliefs, increased consumption, delayed quitting, and reduced impacts of warning messages. (41–43)

Fifth, cannabis marketing in previous research was measured by self-reported exposures (8, 23, 40, 44–47) except for one (21). We used a validated auditing instrument to objectively audit all the RCDs in California, which addressed limitations in self-reporting such as recall and selection bias. (48, 49)

Lastly, unlike most previous studies measuring overall cannabis use, we differentiated medical use from recreational use, which was considerably more concerning from a public health perspective.

METHODS

Data Collection and Study Sample

Our cross-sectional study used three datasets, all of which were collected by our team: 1) individual data on cannabis use outcomes and covariates, which came from the national Marijuana Use and Environment Survey (MUES), 2) RCD listing and location data in California, which were collected and verified using a validated approach, and 3) RCD’s point-of-sale marketing data in California, which were audited using the Standardized Marijuana Dispensary Assessment – Children Focused (SMDA-CF). We briefly describe data collection below and published details elsewhere. (26, 48–50) Data collection was approved by the Human Research Protections Program at the University of California San Diego.

Marijuana Use and Environment Survey (MUES)

MUES is the largest probability-based survey in the U.S. focusing on cannabis-related behaviors and environments. (50) It was conducted online from December 2019 to February 2020 and offered in both English and Spanish. MUES recruited respondents from Ipsos’ KnowledgePanel^®, which represents 97% of adult population in the U.S. MUES covered all the 50 states and Washington DC with California purposely oversampled. A total of 21,903 adults aged 18+ participated in the survey. Of all the 3,411 respondents in California, 23 were excluded because of invalid or missing home address and 3 were excluded because of missing information on cannabis use purposes. The final analysis included 3,385 respondents with complete information. The geodemographic benchmarks in 2018 American Community Survey were used to weight the sample.

Collection and Validation of Dispensary Listings

In May 2019, we used California state license directories and major online crowdsourcing websites (Weedmaps, Leafly, and Yelp) to identify both licensed and unlicensed RCDs with storefront in California. These RCDs included recreational-only (accessible to adults only) and recreational and medical (accessible to both general adults and patients) cannabis dispensaries. Trained researchers called dispensaries to verify information. This validated approach showed that a single data source had strengths and limitations yet a combination of data sources along with call verification could provide an accurate and comprehensive list of RCDs. (26)

Standardized Marijuana Dispensary Assessment

From June to September 2019, we audited all the RCDs with storefront that were identified and verified in California (N=700). (48) The validated SMDA-CF instrument was used, which included a comprehensive list of point-of-sale marketing activities and had moderate to high reliability overall. (49) Trained undergraduate researchers aged 21–23 conducted audits undercover in teams of two to improve data reliability.

Measures

Individual Outcomes: Cannabis Use and Perceptions

The outcomes of interest included five binary indicators for: 1) past-month overall cannabis use, 2) perceiving cannabis smoking harmful (selecting the first two options out of “extremely harmful, somewhat harmful, mostly safe, and totally safe” in the question “how harmful do you think smoking marijuana is if it is consumed at least 3 times a week?”), 3) past-month cannabis use for medical purposes only, 4) past-month cannabis use for recreational purposes only, and 5) past-month cannabis use for dual purposes. Medical-only purposes were defined as consuming cannabis only “to treat or decrease symptoms of a health condition”, recreational-only purposes were defined as consuming cannabis only “to get pleasure or satisfaction”, and dual purposes were defined as consuming cannabis for both medical and recreational purposes. The three measures on past-month use by purpose were mutually exclusive. Similar measures have been used in national surveys. (51, 52)

Primary Predictors: RCD’s Proximity, Density, and Point-of-sale Marketing

We geocoded the street address of RCDs and individual homes and computed two metrics for RCD’s availability around individual homes: proximity and density. In the main analysis, proximity to the RCD closest to individual home was categorized into 0–2 miles, 2–4 miles and 4+ miles and the density of RCDs was measured by count of RCDs within 2-mile radius of individual home. Euclidean distance was used to compute distance between RCD and individual home. The 2- and 4-mile cutoffs were selected to reflect distances that are reachable by walking and cycling/driving/public transportation, respectively. (10, 21, 49, 53)

We evaluated two point-of-sale marketing activities: storefront signage indicative of cannabis and signs promoting health benefits of cannabis. Specifically, binary indicators were created to indicate presence of storefront signage and presence of health benefit signs in the RCD closest to individual home, respectively. We also counted the number of RCDs with storefront signage and health benefit signs within 2-mile radius of individual home.

Individual Covariates: Sociodemographic Factors

Individual sociodemographic factors included age, sex, educational attainments, race/ethnicity, employment status, household income, marital status, and metropolitan core area. Rural-Urban Commuting Area (RUCA) codes were used to identify metropolitan core area for individual homes. (54)

Statistical Analysis

Locations of RCDs and individual homes were mapped in ArcGIS. We described summary statistics of individual and RCD measures. A series of logistic regressions were conducted to assess associations, with individual sociodemographic factors controlled for in all regressions. The first series of regressions assessed proximity measures. We first estimated the associations between proximity of the nearest RCD (0–2 miles, 2–4 miles, and 4+ miles) and individual outcomes. Controlling for RCDs’ proximity, we then added presence of storefront signage (or health benefit signs) in the nearest RCD and the interaction between proximity and presence of storefront signage (or health benefit signs) in regressions. The second series of regressions assessed density measures. Variance inflation factor tests suggest that the three density measures (density of RCDs, density of RCDs with storefront signage, and density of RCDs with health benefit signs) had multicollinearity problem when they entered the same regression simultaneously along with individual covariates (Table S1). Following previous research, (21) we entered each density measure into a separate regression. In sensitivity analysis, we used alternative proximity measures (0–4 miles, 4–8 miles, and 8+ miles) and density measures (1-, 4-, 6-, and 8-mile radius) to account for shorter distance that is easily walkable and longer distances that are also reachable by driving/public transportation. (8, 10, 21) The analysis was weighted to represent California adult population. The analysis was not pre-registered and the results should be considered exploratory.

RESULTS

Descriptive Statistics

Table 1 reports descriptive statistics of the 3,385 respondents included in the study sample. In the sample, 22.22% respondents were past-month cannabis users, 68.44% perceived smoking harmful, 7.40% were medical-only users, 8.054% were recreational-only users, and 6.74% were dual-purpose users. About 37.90% and 18.18% respondents had the nearest RCD located within 2 miles and 2–4 miles, respectively. About 54.19% and 44.98% nearest RCDs had storefront signage and health benefit signs, respectively. Counts of RCDs, RCDs with storefront signage, and RCDs with health benefit signs within 2-mile radius were on average 1.81, 1.034, and 0.58, respectively.

Table 1.

Descriptive Statistics of the Study Sample (n=3,385)

Variable	% or Mean
Individual Outcomes
Past-month overall cannabis use (%)	22.22
Perceiving cannabis smoking harmful (%)	68.44
Past-month cannabis use for medical purposes only (%)	7.40
Past-month cannabis use for recreational purposes only (%)	8.054
Past-month cannabis use for dual purposes (%)	6.74
Predictors of Interest
Proximity to the nearest RCD (%)
≤ 2 mi	37.90
2–4 mi	18.18
≥ 4 mi	43.92
Presence of storefront signage in the nearest RCD (%)	54.19
Presence of health benefit signs in the nearest RCD (%)	44.98
Count of RCDs within 2-mile radius (mean)	1.81
Count of RCDs with storefront signage within 2-mile radius (mean)	1.034
Count of RCDs with health benefit signs within 2-mile radius (mean)	0.58
Individual Sociodemographic Characteristics
Age (%)
18–29	19.18
30–44	27.40
45–59	25.75
60+	27.67
Male (%)	47.51
Educational attainment (%)
High school or less	35.54
Some college	31.14
Bachelor or higher	33.32
Race/ethnicity (%)
Non-Hispanic White	41.17
Non-Hispanic Black	5.46
Non-Hispanic Other	17.60
Hispanic	35.76
Employed (%)	66.50
Household income (%)
≤ 25k	11.75
25–50k	16.25
50–100k	29.35
100k+	42.64
Married/cohabiting (%)	63.84
Living in metropolitan core area (%)	89.55

Figure S1 maps RCDs and individual home locations in California. Figure S2 presents the distribution of RCD density measures within 2 miles of individual homes.

Logistic Regression Results

Main analysis results are reported in Tables 2–5 (full results in Tables S2–S7).

Table 2.

Associations between Proximity of the Nearest RCD and Cannabis Use Outcomes

	Odds Ratio (95% Confidence Interval)
			By Use Purpose
	Past-month Overall Cannabis Use	Perceiving Cannabis Smoking Harmful	Past-month Cannabis Use for Medical Purposes Only	Past-month Cannabis Use for Recreational Purposes Only	Past-month Cannabis Use for Dual Purposes
Proximity of the nearest RCD
≤ 2 mi	1.078 (0.81, 1.43)	0.87 (0.68, 1.11)	0.72 (0.47, 1.11)	1.18 (0.76, 1.83)	1.42 (0.89, 2.26)
2–4 mi	0.88 (0.61, 1.27)	0.94 (0.68, 1.30)	0.73 (0.41, 1.28)	1.081 (0.60, 1.94)	0.87 (0.49, 1.53)
≥ 4 mi	---	---	---	---	---

Notes. Predictors entered regressions jointly. Individual sociodemographic factors were controlled for but not reported. Detailed regression results are reported in Table S2.

Table 5.

Associations of Count of RCDs, Count of RCDs with Storefront Signage, and Count of RCDs with Health Benefit Signs within 2-mile Radius with Cannabis Use Outcomes

	Odds Ratio (95% Confidence Interval)
			By Use Purpose
	Past-month Overall Cannabis Use	Perceiving Cannabis Smoking Harmful	Past-month Cannabis Use for Medical Purposes Only	Past-month Cannabis Use for Recreational Purposes Only	Past-month Cannabis Use for Dual Purposes
Count of RCDs within 2-mile radius	1.033 (0.99, 1.067)	0.98 (0.96, 1.011)	0.97 (0.93, 1.023)	1.040 (0.99, 1.094)	1.054* (1.010, 1.099)
Count of RCDs with storefront signage within 2-mile radius	1.063* (1.013, 1.11)	0.96 (0.92, 1.0027)	0.98 (0.91, 1.042)	1.062 (0.98, 1.15)	1.097** (1.029, 1.17)
Count of RCDs with health benefit signs within 2-mile radius	1.11* (1.0067, 1.22)	0.95 (0.88, 1.036)	0.96 (0.84, 1.11)	1.14 (0.99, 1.32)	1.12* (1.0054, 1.25)

^*p<.05;

^**p<.01

Notes. Predictors entered regressions separately. Individual sociodemographic factors were controlled for but not reported. Detailed regression results are reported in Tables S5–S7.

In terms of proximity measures, no evidence was found for the associations between RCD’s proximity and cannabis use outcomes (Table 2). After controlling for RCD’s proximity, presence of storefront signage was associated with smaller odds of perceiving cannabis smoking harmful (interaction OR=0.63, 95%CI: 0.39, 0.99) if the nearest RCD was located within 2 miles (Table 3); presence of health benefit signs was associated with greater odds of overall cannabis use (interaction OR=2.45, 95%CI: 1.17, 5.16) and recreational use (interaction OR=3.50, 95%CI: 1.027, 11.91) if the nearest RCD was located within 2–4 miles (Table 4).

Table 3.

Association between Presence of Storefront Signage in the Nearest RCD and Cannabis Use Outcomes

	Odds Ratio (95% Confidence Interval)
			By Use Purpose
	Past-month Overall Cannabis Use	Perceiving Cannabis Smoking Harmful	Past-month Cannabis Use for Medical Purposes Only	Past-month Cannabis Use for Recreational Purposes Only	Past-month Cannabis Use for Dual Purposes
Proximity of the nearest RCD
≤ 2 mi	1.17 (0.76, 1.79)	1.13 (0.79, 1.64)	0.71 (0.38, 1.35)	1.12 (0.56, 2.23)	1.86 (0.91, 3.80)
2–4 mi	0.83 (0.49,1.41)	1.061 (0.67, 1.67)	0.57 (0.28, 1.15)	0.78 (0.30, 2.024)	1.35 (0.61, 2.99)
≥ 4 mi	---	---	---	---	---
Presence of storefront signage in the nearest RCD	1.37 (0.94, 1.99)	1.069 (0.77, 1.48)	1.22 (0.71, 2.11)	1.30 (0.70, 2.38)	1.38 (0.73, 2.68)
Interaction between proximity and storefront signage
≤ 2 mi & storefront signage	0.88 (0.51, 1.51)	0.63* (0.39, 0.99)	1.022 (0.46, 2.25)	1.086 (0.46, 2.59)	0.63 (0.25, 1.60)
2–4 mi & storefront signage	1.11 (0.54, 2.31)	0.80 (0.42, 1.51)	1.47 (0.54, 4.0045)	1.62 (0.47, 5.59)	0.42 (0.14, 1.30)
≥ 4 mi & storefront signage	---	---	---	---	---

^*p<.05

Notes. Predictors entered regressions jointly. Individual sociodemographic factors were controlled for but not reported. Detailed regression results are reported in Table S3.

Table 4.

Association between Presence of Health Benefit Signs in the Nearest RCD and Cannabis Use Outcomes

	Odds Ratio (95% Confidence Interval)
				By Use Purpose
	Past-month Overall Cannabis Use	Perceiving Cannabis Smoking Harmful	Past-month Cannabis Use for Medical Purposes Only	Past-month Cannabis Use for Recreational Purposes Only	Past-month Cannabis Use for Dual Purposes
Proximity of the nearest RCD
≤ 2 mi	1.053 (0.72, 1.54)	0.82 (0.59, 1.14)	0.69 (0.39, 1.20)	1.38 (0.76, 2.51)	1.27 (0.69, 2.33)
2–4 mi	0.59* (0.35, 0.98)	0.99 (0.65, 1.51)	0.45 (0.20, 1.028)	0.54 (0.21, 1.41)	0.99 (0.49, 2.031)
≥ 4 mi	---	---	---	---	---
Presence of health benefit signs in the nearest RCD	1.073 (0.74, 1.56)	0.83 (0.60, 1.14)	0.96 (0.55, 1.69)	1.53 (0.82, 2.83)	0.83 (0.45, 1.52)
Interaction between proximity and health benefit signs
≤ 2 mi & health benefit signs	1.075 (0.62, 1.85)	1.093 (0.68, 1.75)	1.12 (0.50, 2.48)	0.78 (0.32, 1.89)	1.29 (0.55, 3.022)
2–4 mi & health benefit signs	2.45* (1.17, 5.16)	0.82 (0.42, 1.59)	2.74 (0.84, 9.0017)	3.50* (1.027, 11.91)	0.61 (0.20, 1.84)
≥ 4 mi & health benefit signs	---	---	---	---	---

^*p<.05

Notes. Predictors entered regressions jointly. Individual sociodemographic factors were controlled for but not reported. Detailed regression results are reported in Table S4.

In terms of density measures (Table 5), count of RCDs with storefront signage and count of RCDs with health benefit signs within 2 miles were separately associated with greater odds of overall cannabis use (OR=1.063, 95%CI: 1.013, 1.11 and OR=1.11, 95%CI: 1.0067, 1.22, respectively) and dual-purpose use (OR=1.097, 95%CI: 1.029, 1.17 and OR=1.12, 95%CI: 1.0054, 1.25, respectively). Count of RCDs was also associated with greater odds of dual-purpose use (OR=1.054, 95%CI: 1.010, 1.099).

Sensitivity Analysis

Sensitivity analysis results with alternative proximity cutoffs are reported in Tables S8–S10. The evidence found in main analysis using 2- and 4-mile cutoffs were no longer observed when 4- and 8-mile cutoffs were used.

Sensitivity analysis results with alternative density measures are reported in Tables S11–S14. The associations found in main analysis using 2-mile cutoff generally held when 1-, 4-, and 6-, and 8-mile cutoffs were used. Particularly, the three density measures were consistently associated with greater odds of cannabis dual-purpose use in almost all the regressions. Some associations not revealed in main analysis were found. For example, some regressions show that density measures were negatively associated with perceiving cannabis smoking harmful and positively associated with recreational-only use.

DISCUSSION

This study suggests that density of RCDs in a 2-mile radius of individual homes was positively associated with past-month cannabis use for dual purposes among adults in California, U.S. Such positive association was also observed in many previous studies on medical cannabis dispensaries (13, 14, 18–21) and RCDs (7, 9–11), but these studies did not distinguish cannabis use by purpose. Dual-purpose users accounted for approximately one third of cannabis user population in the U.S., (50) but little is known about their distinct characteristics and behavioral patterns. Our previous research found that this group was unique: cannabis smoking as the primary administration method was more prevalent in dual-purpose users than medical-only and recreational-only users. Accordingly, non-combustion methods such as vaping and ingestion were less likely to be used by dual-purpose users as primary methods. Dual-purpose users were also more likely to use multiple methods to consume cannabis. (50) Cannabis smoking is related to excessive risks of adverse health consequences and using multiple methods is related to addictive harms. (50) Dual-purpose users as a high-risk population deserve more research to explore their distinct risk profiles and risk factors.

For the first time, this study reports associations between presence/density of storefront signage indicative of cannabis and cannabis use and perception. This finding is supported by a previous study on medical cannabis dispensaries in Log Angeles County in California, which found positive associations between density of dispensaries displaying signage indicative of cannabis and young adults’ cannabis use and positive perceptions. (21) Storefront signage may be particularly influential to never users and occasional users, who were not familiar with the retail environments in neighborhoods. Tobacco research suggested that users’ consumption is caused by affective impulse rather than rational consideration when responding to marketing activities. (55) In California, about 60% RCDs had storefront signage. (48) The wide presence of storefront signage may provide an affective impulse for nonusers to try cannabis and users to purchase cannabis. The RCDs with storefront signage may be also more likely to promote positive image and social acceptance of cannabis and shape favorable perceptions and behaviors.

This study is also the first to suggest positive association between presence/density of signs promoting health benefit of cannabis and cannabis use. Interestingly, we found that such associations were only observed among recreational-only or dual-purpose users but not medical-only users. Our speculation is that medical-only users were already familiar with health benefits of cannabis use and not responsive to the signs. Future research is encouraged to explore the heterogeneities among cannabis users with different purposes and the mechanisms for the heterogeneities.

The findings have policy implications. Although some local jurisdictions regulate the location and number of licensed RCDs, the ordinances are usually applied to protect children. For example, California requires 1,000 feet to be the minimum distance of RCDs from school. Much fewer local jurisdictions implement specific regulations to protect residential areas. For example, the city of San Diego in California requires RCDs to be located at least 100 feet from residential areas. The city of Port Orchard in Washington prohibits RCDs within any residential areas. Given that the density of RCDs within 2 miles of individual homes was associated with cannabis use, controlling the proximity and density of RCDs in residential areas may deserve consideration from policymakers. Currently California has no regulations specific to storefront signage indicative of cannabis. It is worth investigating whether regulating the size and content of the signage has impacts on cannabis use. Unsubstantiated health claims may promote positive perceptions and risky cannabis use behaviors such as overdose. The Food and Drug Administration and several states in the U.S. (e.g. Oregon, Mississippi, and Massachusetts) prohibit marketing cannabis and related products with health or therapeutic claims. (56) Our dispensary auditing data demonstrated that about 40% dispensaries in California displayed health benefit signs. (48) Actions are called for to monitor and enforce the regulation of such activities and prevent exposure of misleading information.

The study has limitations. First, this cross-sectional study estimated associations instead of causality. We were not able to control for unobserved factors that may confound the association estimation. For example, a neighborhood with pre-existing high prevalence of cannabis use may have more supportive social norms, greater demand, and less stringent regulation and enforcement, which may attract both cannabis users to live and RCD owners to select the storefront location. Second, we measured presence of storefront signage and health benefit signs with binary indicators. Detailed information on size, content, and quantity was not collected during dispensary auditing due to feasibility and cost concerns. Future research could examine whether and how these factors affect outcomes.

Third, dispensary data collection was completed approximately three months before individual data collection started. If RCDs’ location and marketing activities had substantial variations over these months, dispensary measures may not perfectly match the status when individual made the recall. Such mismatch of timing, however, should not be a major concern given the short gap between data collections. Fourth, we were not able to evaluate all the marketing activities in RCDs.

Fifth, we assessed RCDs around individual homes. RCDs around individual workplace or schools may also influence cannabis use outcomes. We did not audit or control for medical-only cannabis dispensaries that only sell cannabis to patients. It was less concerning considering that such dispensaries accounted for roughly 10% of the total dispensaries in California and the access was limited to patients. We did not assess delivery-only services, which were allowed in California. Delivery-only services do not influence individual cannabis use through presence of storefront or point-of-sale marketing activities, but they likely have considerable influences on cannabis availability and accessibility particularly during the COVID-19 pandemic. We also did not account for other ways of obtaining cannabis, such as through cultivation, family or friends, or illicit market. Some error variance may be introduced to the regressions as a result. Further, we did not make distinctions between licensed and unlicensed RCDs. The license status of RCDs during dispensary audit may not accurately reflect the status when individuals made the recall in surveys. We hope future research could investigate whether license status has impacts on the associations.

Lastly, like all previous studies on RCDs, the study sample was restricted to population in a jurisdiction instead of the entire nation. The findings may not generalize to other jurisdictions in the U.S. or other countries. For instance, our sample in California had a greater prevalence of past-month cannabis use than the US adult population. (50) Nonetheless, the probability-based sampling strategy made the findings at least generalizable to adult population in California, the most populous state in the U.S.

CONCLUSION

In this cross-sectional study with a representative sample of adults in California, U.S., there was some evidence suggesting the associations of RCDs’ availability and presence of storefront signage and health benefit signs with cannabis use and risk perceptions. Given the continuous expansion of recreational cannabis legalization and commercialization in the U.S., it is imperative to establish the causal mechanisms of observed associations. Continuous surveillance on RCDs’ retail environments and local cannabis use patterns is also recommended.