Study design
We applied multi-sitecyber-ethnography to social media profiles maintained by retail cannabis businesses. This study was determined to be exempt by the University of Wisconsin Institutional Review Board as it involved evaluation of publicly available information.
Profile identification
We identified retail cannabis businesses with Facebook and Instagram presence from Alaska, Oregon, Colorado, and Washington, in order to have a representative sample of states where non-medical cannabis is legalized. We prioritized states in which legalization had been present for a longer period of time, such that enough time had elapsed for states to open retail businesses and those businesses to establish social media profiles with at least a year of content.
Data for this study were obtained from two social media sites, Facebook and Instagram, which are among the most popular social media sites for businesses using social media, as well as youth at the time of this study [41]. Based on pilot testing with various keyword options, potential retail cannabis businesses were identified through a Facebook search using the terms [retail cannabis + state name] under Pages and Local Business or Place. The “About” section or business name from each profile was evaluated to confirm that business’s focus on retail, non-medical cannabis sales. We then determined whether the business had profiles on both Facebook and Instagram. If so, we evaluated further inclusion criteria: business location in the appropriate state, content in English, posting twice within the past two months to ensure the profile was still active, and having maintained both social media profiles since June 2017 so that a full year of content was available for evaluation. Of the businesses that met inclusion criteria, the top 5 most popular profiles were selected for each state based on total followers on Facebook and Instagram. Sixteen businesses initially met search criteria; however, two businesses had profiles removed before data collection began and were removed from our sample. The final sample included 28 social media profiles from 14 businesses. One year of content, from June 2017 through May 2018, was evaluated from each profile.
Role of investigator
In this study, investigators acted as complete observers, where the observer remains detached from the participants who are unaware they are being observed [42]. This allows for an unbiased snapshot of typical practices of those being observed. We chose this type of participant observation because we were interested in acting as observers of the retail profiles, similar to how we hypothesized that adolescents may experience online cannabis marketing. Carspecken’s five stages of qualitative research was used as a framework for our study procedures [43].
Data collection
Stage 1 – compiling the primary record
This step is comprised of fieldwork by observers and recording of field notes [43]. Ideally, ethnography should begin with a conscious attitude of almost complete ignorance, and the researcher then builds a “thick description” by unobtrusively observing social practices. The goal is to avoid biases or preconceived notions of the culture being observed. For this paper, we defined ‘field notes’ as recorded observations by an individual investigator based on online content they evaluated.
Individual investigators were each assigned to evaluate six months of content from one business’ social media profile, either Facebook or Instagram. For each social media profile, an investigator was immersed in the profile’s content for approximately 3 weeks to collect 6 months of data. Investigators used a template to collect field notes throughout the 3-week immersion period. The template prompted for insights specifically on major themes, tone, and potential target audience for each business’s content. An example of this template is included in Additional file 1. Raw data from the sites was also collected and available for investigators to refer to within field notes and in later synthesis of the notes. A set of field notes for each business profile was created at the end of each immersion period.
Stage 2 – preliminary reconstructive analysis
Stage 2 of the Carspecken framework involves analysis and review of the compiled data [43]. This step also involves reflection on the cultural context and site of observation, including relations between participants, or in the case of this study, businesses. During this stage, initial impressions of themes are documented and areas are identified that may need further exploration in proceeding stages.
Two investigators created field notes for each social media business profile, each observing a distinct 6 months of content. Therefore, four different investigators in total completed field notes for each businesses’ content on Facebook and Instagram to produce a final set of compiled field notes covering all content from that business over a one-year period. This approach was meant to bolster observations through triangulation of data, as well as to reduce biases from any individual observers [44]. After data collection, all investigators met for a debriefing meeting to discuss experiences and initial thoughts on larger themes. This step allowed investigators to consider and document big picture ideas and themes before delving into detailed data analysis. A list was then complied of initial themes.
Stage 3 – dialogical data generation
Stage 3 involves a dialogical approach to gain an insider’s position with respect to culture [43]. This is unique from Stages 1 and 2, which solely utilize an outsider perspective. This step also involves making connections between data points and checking the data against initial themes identified in Stage 2.
This was a multi-site study involving two social media sites, therefore data were collected for individual cannabis businesses’ Facebook and Instagram profiles. Analysis was focused on the individual business, not on distinct social media platforms, as our goal was to capture a more holistic picture of the businesses’ interaction with social media across multiple sites. Two investigators read through all compiled field notes and constructed lists of themes individually, informed by initial themes identified by the group in Stage 2 and responses to the data. Three investigators met and came to consensus on a combined list of proposed themes based on the individually constructed lists of themes.
Stage 4 – describing system relations
During this stage, themes are extracted from the data, providing evidence to support proposed themes from the primary data [43]. A full description of the documented relationships and typical events is developed. Reflexivity is also involved in this stage, where investigators reflect upon their own biases and assumptions to prevent them from influencing the study.
Dedoose (Dedoose, Los Angeles, California), a qualitative analysis software, was used to upload field notes and then extract themes from the text. Dedoose allows the user to create a list of variable and highlights excerpts of uploaded text to apply theme to the text directly. Multiple users can apply themes blinded to other users’ analysis, and these can be unblinded after analysis is complete for comparison. Three investigators were involved in thematic analysis. The investigators utilized a constant comparative approach [44]. Inductive reasoning based in grounded theory guided theme identification [45].
In the first cycle of analysis, two investigators independently applied themes to combined field notes from one business, blinded to the other’s analysis. After applying categories to this transcript, the investigators met and the selected text was reviewed with themes un-blinded. The investigators discussed and reached consensus on theme and subtheme nomenclature for the combined list of themes during this stage, as well as any additions or revisions to the themes list. The categorization process was then applied to combined field notes for a second business. The purpose of this second review was to evaluate reliability and validity of the proposed themes. After review, discussion and achieving consensus on the themes, the themes were applied to all remaining data using this constant comparative approach. Throughout this process, the third investigator was available to help reach consensus if the two primary investigators were not in agreement.
Stage 5 – system relations as explained by findings
An important part of ethnographic research and the Carspecken framework is to create a narrative that describes the specific culture that was observed [43]. Our goal for this study was to identify ways cannabis businesses cultivate an online presence and exert influence that may reach youth; therefore, findings were analyzed through the lens of this goal. This analytic step aims to develop a thick description through the process of qualitative analysis. Qualitative analysis data and immersive experience were reflected upon to determine the most important themes. Narratives for major themes were constructed using our experience as observers, qualitative analysis data and process, and all available collected data from investigators. In this way, themes became interconnected to create a cohesive picture.