Article Text

Identifying e-cigarette vape stores: description of an online search methodology
  1. Annice E Kim1,
  2. Brett Loomis1,
  3. Bryan Rhodes2,
  4. Matthew E Eggers3,
  5. Christopher Liedtke1,
  6. Lauren Porter4
  1. 1Public Health Policy Research Program, RTI International, Research Triangle Park, North Carolina, USA
  2. 2Digital Technology and Society Program, Survey Research Division, RTI International, Research Triangle Park, North Carolina, USA
  3. 3Primary Prevention Research and Evaluation Program, RTI International, Research Triangle Park, North Carolina, USA
  4. 4Department of Health, Bureau of Tobacco Free Florida, Tallahassee, Florida, USA
  1. Correspondence to Dr Annice E Kim, Public Health Policy Research Program, RTI International, 3040 E. Cornwallis Road, P.O. Box 12194, Research Triangle Park, NC 27709, USA; akim{at}rti.org

Abstract

Background Although the overall impact of Electronic Nicotine Delivery Systems (ENDS) on public health is unclear, awareness, use, and marketing of the products have increased markedly in recent years. Identifying the increasing number of ‘vape stores’ that specialise in selling ENDS can be challenging given the lack of regulatory policies and licensing. This study assesses the utility of online search methods in identifying ENDS vape stores.

Methods We conducted online searches in Google Maps, Yelp, and YellowPages to identify listings of ENDS vape stores in Florida, and used a crowdsourcing platform to call and verify stores that primarily sold ENDS to consumers. We compared store listings generated from the online search and crowdsourcing methodology to list licensed tobacco and ENDS retailers from the Florida Department of Business and Professional Regulation.

Results The combined results from all three online sources yielded a total of 403 ENDS vape stores. Nearly 32.5% of these stores were on the state tobacco licensure list, while 67.5% were not. Accuracy of online results was highest for Yelp (77.6%), followed by YellowPages (77.1%) and Google (53.0%).

Conclusions Using the online search methodology we identified more ENDS vape stores than were on the state tobacco licensure list. This approach may be a promising strategy to identify and track the growth of ENDS vape stores over time, especially in states without a systematic licensing requirement for such stores.

  • Electronic nicotine delivery devices
  • Non-cigarette tobacco products
  • Surveillance and monitoring

Statistics from Altmetric.com

Introduction

Electronic Nicotine Delivery Systems (ENDS), such as e-cigarettes, are battery-powered devices that heat liquid in a cartridge to deliver to the user an inhaled dose of nicotine and other additives, including a humectant (propylene glycol or glycerol) and flavourings. Although the overall impact of ENDS on public health is unclear,1 ,2 awareness, use, and marketing of the products have increased markedly since these were first introduced into the US marketplace in 2007.3–6 Since 2007, when ENDS first became available in the USA, ENDS sales reportedly have doubled each year, reaching an estimated US$2.5 billion for 2014.7 ,8 Awareness of ENDS among US adults increased from 40.9% to 79.7% between 2010 and 2013, and ever use increased from 3.3% to 8.5%.9 From 2011 to 2013, ever use of ENDS increased from 1.4% to 3.0% among middle-school students and from 4.7% to 11.9% among high-school students.10 ,11 Concurrent with recent increases in awareness and use of ENDS, advertising of ENDS has increased dramatically. From 2011 to 2012, advertising expenditures for ENDS across various media channels tripled from $6.4 million to $18.3 million, particularly in magazines and on television.12 Despite the lack of evidence concerning the safety, abuse potential, or efficacy of ENDS,13 ,14 ENDS are advertised as a safer alternative to cigarettes that are socially acceptable in situations where conventional tobacco smoking is not allowed.15 ,16

ENDS are sold in traditional retail outlets that are licensed to sell tobacco products such as convenience stores and grocery stores. A recent national study found that ENDS were available in more than 30% of licensed tobacco retailers sampled nationwide17 and in 71% of licensed tobacco retailers in Florida.18 At licensed tobacco retail stores, a handful of top ENDS brands are sold alongside traditional tobacco products and other consumer goods. Availability and sales of ENDS at these outlets are being monitored through existing retail audit systems and retail scanner sales data; however, these do not capture data from the increasing number of vape stores that primarily sell ENDS.19 These vape stores can sell a wider selection of ENDS devices than traditional retail outlets, including higher end tank devices, accessories and liquids/juices that until now were primarily available online.19 A study of online consumer reviews of vape stores suggests that consumers considered the availability of flavours, devices, and accessories as particularly important in their evaluation of such stores.20 Some vape stores (eg, Henley Emporium (http://www.thehenley.com)) also offer tasting bars and ‘lounges’ designed to encourage socialising and entertainment, thereby, reinforcing positive social norms around ENDS use.19 Understanding how ENDS are advertised and sold in these specialty vape stores is important given the influence of retail tobacco marketing on behaviours; an extensive body of research shows that retail advertising of tobacco products increases youth susceptibility to smoking and underage retail tobacco purchases, and increases craving and unplanned tobacco purchases among adults.21 Studies have also shown that retail stores selling tobacco products are disproportionately located in ethnic minority and low-income communities.22 The extent to which ENDS vape stores are geographically concentrated in certain communities, and whether exposure to these stores influences youth and adult behaviours is largely unknown. Furthermore, while 40 states have passed laws prohibiting sales of e-cigarette products to youth,23 retailers’ compliance with these laws is unknown, especially among the rapidly growing number of ENDS vape stores.

To document the marketing practices of ENDS vape stores and to assess retailers’ compliance with youth access laws, we need to first identify and audit stores in business. However, identifying ENDS vape stores can be challenging given the lack of policies regulating these businesses, including licensing. In most states, an ENDS vape shop can be opened with a simple business license. These licenses are not readily indexed or searchable for generating a sampling frame of ENDS vape stores in a given geographic area. In the absence of such a list, online directories and search engines may provide an alternative data source, but the quality and representativeness of these data sources are largely unknown. However, in some states, such as Florida, ENDS vape store owners are required to obtain a tobacco license, similarly to retailers who sell tobacco products in traditional outlets like convenience stores and grocery stores. Selling ENDS without the appropriate state license is a non-criminal violation, punishable by a fine of not more than $500. Tobacco licenses are more systematically codified and indexed because state and local agencies use these lists to conduct audits to ensure retailer compliance with tobacco age of sale restrictions and advertising regulations. In states that license tobacco retailers, the licensure list is the best way to identify tobacco retailers. We wanted to investigate whether this was also true for identifying ENDS vape stores.

The purpose of this study is to assess how well online search methods identify ENDS vape stores compared to tobacco/ENDS licensure lists. In Florida, retailers that sell ENDS are required to obtain a license to sell tobacco and therefore, should be represented on the tobacco business license list that is publicly available. We searched for ENDS vape stores in the state of Florida using Google, Yelp, and YellowPages, and confirmed the businesses that primarily sold ENDS in a storefront for consumers; we then compared the results to Florida state's tobacco licensure list. Results could help inform other states and researchers in their efforts to track ENDS vape stores over time, especially in states without a systematic licensing requirement for ENDS vape stores, and to generate a sampling frame of stores that could be used for retail audits.

Methods

Online searches to generate list of potential ENDS vape stores

To generate an initial list of potential vape stores throughout Florida, we used automated web scraping software to search and gather retailer contact information from several public business directories. Web scraping software allows us to conduct large numbers of searches on a particular website automatically and to gather the results of those searches into a single usable data file. Once the scraping software is set up, it can gather data on thousands of retailers within a few hours. This allows us to repeat this part of the process to gather the most up-to-date information at any given time. Although a number of web-based retailer directories are available, we chose three leading online directories that have extensive and complimentary listings: Google Maps, Yelp, and YellowPages. Google Maps contains a list of over 95 million businesses and places of interest pulled from Google's web-crawl and searchable using Google's search function. Yelp is a web-based directory of businesses, built primarily by crowdsourcing from users’ information and reviews on local businesses. Yelp contains information on more than 50 million businesses, including contact information, type of business and reviews. Since Yelp uses the public to generate much of its information (rather than crawling the web) and focuses primarily on local brick-and-mortar businesses (rather than all businesses and points of interest), it complements Google Maps. YellowPages is an online directory of businesses modelled on the traditional Yellow Pages listings in which business information is provided by phone companies or the businesses themselves, thereby complementing the other two sources. YellowPages contains over 50 million business listings.

We conducted searches in Google Maps, Yelp, and YellowPages using an ENDS or vape-related term (‘ecig,’ ‘e-cigarette,’ ‘vape,’ ‘vapor,’ ‘vaper,’ ‘vapin’) for each of the 409 cities/towns in Florida.i The searches were conducted between May 1 and May 6, 2014. The name of the business, address, and telephone number were extracted from all search results and merged into one data file. Data cleaning involved multiple steps. First, we removed duplicate listings with the identical name, address and telephone number. Second, we visually reviewed the list for duplicates that were not initially identified due to variations in how business names and addresses are formatted by the directories. Third, we reviewed the names of business for obvious false positives (eg, hotels) and then searched for these terms and systematically removed these businesses from the list. At this stage, we were fairly conservative in our data cleaning to maximise identifying potential ENDS vape stores.

Phone verification of ENDS vape stores via crowdsourcing

After generating the list of potential ENDS vape stores from our online searches, we wanted to determine which businesses were actual ENDS vape stores. Traditionally, this might be done by reviewing administrative records or by having professional data collectors call or visit the establishments. However, given the relatively recent rise of ENDS vape stores, administrative records for ENDS vape stores are limited. Furthermore, given the large number of search results, having professional data collectors visit these establishments would be time consuming and potentially cost prohibitive. To overcome these challenges, we used a crowdsourcing platform, Amazon Mechanical Turk (MTurk), to verify the information we had gathered. MTurk allows users to post discrete jobs or tasks called Human Intelligence Tasks (HITs) to be completed by an online network of MTurk workers. These tasks are typically small jobs that can be quickly completed by a human worker but are often difficult for a computer, such as gathering location data on retailers24 ,25 and sentiment analysis of social media posts.26

To verify which establishments in our list were actual ENDS vape stores, we submitted a series of HITs to MTurk along with the list of store names and phone numbers identified from the online searches. In the first HIT, we asked MTurk workers to call the establishment and determine (1) whether the business sells ENDS (yes/no), (2) whether the business primarily sells ENDS and/ or ENDS-related juices/fluids or accessories (yes/no), and (3) whether the business sells other tobacco products (eg, cigarettes, cigars) (yes/no). These questions were asked to decipher the ENDS vape stores of interest from businesses that either did not sell ENDS or sold ENDS along with other tobacco products (eg, tobacco shop, convenience store). In the second HIT, we submitted the list of businesses that were positively confirmed as primarily selling ENDS and asked MTurk workers to call the establishment and determine (1) whether the business sells ENDS primarily to individual customers or other businesses like a distributors (customer/distributor), and (2) whether the business has a storefront that is open to the public (yes/no). These questions were asked to decipher the ENDS vape stores of interest from businesses that were distributors/resellers or web-only retailers with no physical brick-and-mortar storefront for selling directly to individual consumers. The full list of questions posted on mTurk is provided as online supplementary material. Each HIT was completed in less than 8 h.

To introduce quality control, we posted three identical HITS for each establishment so that each store on our list was called by three different workers. For each question, majority agreement (answer posted by two of three workers or all three workers) was determined as the ‘true’ answer. If there was no majority agreement, the HIT was submitted again to three different workers; this occurred for 17.4% of HITs. For the measures, the level of majority agreement (ie, at least two of three Turkers agreed) ranged from 84.1% to 90.9%.

List of licensed tobacco/ENDS retailers from Florida state

In Florida, retailers that sell tobacco/ENDS can apply for a tobacco license at any time and permits must be renewed annually on or before January 15 every year (http://www.leg.state.fl.us/Statutes/index.cfm?App_mode=Display_Statute&URL=0500-0599/0569/0569.html). The list of licensed tobacco retailers was obtained from the Florida Department of Business and Professional Regulation in April 2014. The list contains information on street and mailing location, store and owner name, primary and secondary license status code, license number, original licensure issue date, effective license date, and expiration license date. The list was cleaned to remove stores with incomplete address information or stores that were closed and no longer in business.

Analysis

The list of confirmed ENDS vape stores from the online search and crowdsourcing methodology was then compared to the full list of stores on the state tobacco licensure list. We prioritised the street address as the main data point of comparison as store names could be spelled differently or listed differently (eg, doing-business-as name vs storefront name) and phone numbers can also vary (eg, landline vs owner's mobile phone line). Stores from the two lists were considered the same if they matched on the entire address (street number, street name, city, zip code). If stores had slightly different addresses (eg, street numbers were different), they were considered as different stores.

Results

Figure 1 summarises the results of the online search methodology and comparison with the tobacco licensure list.

Figure 1

Online search results and comparison to tobacco licensure list. ENDS, Electronic Nicotine Delivery Systems; MTurk, Mechanical Turk.

Online search results: Using the series of ENDS and vape search terms, Google provided the most results (1054), followed by YellowPages (882) and Yelp (473). Of the combined 2409 search results, nearly half (49.1%, N=1184) were duplicates and one-third (33.3%, N=803) were irrelevant listings (eg, hotels).

Phone confirmation results: A total of 1459 stores identified from online searches were uploaded to MTurk so that crowdsourced workers could call and confirm whether stores identified were brick-and-mortar ENDS vape stores that primarily sell ENDS devices and accessories to individual consumers. Of the 1459 stores, more than one-quarter fit these criteria (27.6%, N=403). Only six stores were ENDS distributors or resellers, while another 33 stores (2.3%) sold ENDS along with other tobacco products (eg, tobacco shop). More than half of the stores on our initial list did not sell ENDS (N=819, 56.1%), while 13.6% (N=198) were closed, had no valid phone number, had incomplete data (eg, wrong number) or were duplicates.

Comparison of the two lists: When we compared our final list of vape stores from the online search methodology (N=403) to the full tobacco licensure list (N=29 039), we found that 131 of the 403 stores were on the licensure list (32.5%), while 272 stores were not (67.5%) (figure 1).

Table 1 summarises the accuracy and overlap in stores identified by the three online sources. YellowPages accurately identified the highest number of ENDS vape stores (N=680), followed by Google (N=559) and Yelp (N=367). Accuracy of results was highest for Yelp (77.6%), followed by YellowPages (77.1%) and Google (53.0%), with an average of 66.6% accuracy among all three online sources.

Table 1

Comparison of search results by online search engine/directory

Discussion

In summary, we identified and confirmed a total of 403 ENDS vape stores in the state of Florida using online search engines and directories. Only 32.5% of these vape stores were licensed to sell vapour products according to the Florida state tobacco licensure list. It is possible that the additional stores we identified opened for business after the tobacco licensure list was obtained. Since online directories and search engines can be updated more readily, the data extracted from these online sources may have captured the most current data on the number of ENDS vape stores open for business. Alternatively, the additional stores identified via online searches may have been in business for some time but did not appear on the tobacco licensure list because they never obtained a tobacco license. As there was little regulatory enforcement of ENDS licensing requirements by the state, it is unclear if ENDS retailers who do not have a license were simply unaware of the requirement or were purposely evading the licensing process. Although tobacco licensure lists are often the gold standard for identifying stores that sell conventional tobacco products, this may not be the case for an emerging area like ENDS vape stores. In Florida (and likely in other states), the licensing agency does not require a separate license category for ENDS vape stores, therefore making it challenging to distinguish between tobacco retailers and ENDS vape stores. Future studies need to assess how changes in policies to license ENDS vape stores are communicated to existing and future owners to ensure that tobacco licenses are obtained. In an emerging area like ENDS vape stores, multiple data sources (online searches, tobacco licensure lists) may be needed to generate a comprehensive list of ENDS vape stores in business.

Although online search methods may be a promising approach, our results suggest that multiple online sources should be considered, and a fair amount of data cleaning is needed. We found that Google and YellowPages identified substantially more ENDS vape stores than Yelp, but these also generated considerably more inaccurate listings; this is likely because these search existing databases while Yelp is curated content crowdsourced by consumers. Compared with Google, both Yelp and YellowPages had substantially more accurate results. There was a fair amount of noise in the initial search results data; we manually reviewed the list of store names and removed obvious false positives but, nearly half of the list of potential ENDS vape stores were dropped during the MTurk phase because these were confirmed as not selling ENDS. Without more information about online search algorithms, which are often proprietary, it is difficult to assess whether there are systematic biases in how search engines and online directories identify and catalogue businesses.

This study has several strengths and limitations. First, comparing the results of online searches to the official tobacco licensure list (which we were able to do in Florida) could be one way to identify vape shop owners who are operating businesses without a license or who may have recently opened their business and are, therefore, not reflected in the annual license list This information could help inform state's education and compliance efforts on vape shop licensure requirement. Second, by using multiple search engines and directories, we were able to examine the overlap as well as unique stores identified by the different sources. Third, by using an online crowdsourcing platform like MTurk, we were able to confirm whether stores did in fact primarily sell ENDS to customers in a brick-and-mortar retail storefront in considerably less time (8 h after job was posted online) than would be feasible by sending our data collectors out in the field to collect similar data. Additionally, by posting the same job three times and using data with majority agreement, we were able to introduce more quality control in the crowdsourced data collection process. Despite these strengths, there are some key limitations to the study. First, the ENDS and vape-related terms may have systematically missed some relevant vendors in the online search. As such, we cannot say how representative the 403 vape stores we identified are of the total universe of vape stores in Florida. We compared our list to the state tobacco licensure list because in theory this should be the most comprehensive list given Florida's licensing requirements. However, our results suggest that the Florida licensing list is not entirely comprehensive and multiple sources should be used when attempting to identify the universe of ENDS vape stores. Second, although we had each store called by three different workers, without a gold standard verification, we cannot determine whether the majority rule was the valid answer.

In conclusion, we find that online search methodology coupled with crowdsourced data collection may be a promising strategy for identifying ENDS vape stores. Through this process, we were able to identify over 400 unique vape shops in the state of Florida, of which nearly 70% were not on Florida's tobacco licensure list. The scraping software and text analysis enable us to automate the process of gathering data on thousands of retailers within a few hours. Therefore, this approach can be easily repeated to capture new vape stores over time. In the absence of a gold standard list for ENDS vape stores, a combination of online and tobacco licensure sources should be used to generate a sampling frame for retail audits and other studies, such as examining the geographic distribution and density of ENDS vape stores. To the extent that websites like Yelp also include consumer reviews and photos, researchers can also collect data on the store's product availability and marketing practices as well as consumer perceptions about ENDS vape stores. This online approach can be easily focused on specific areas, even down to specific neighbourhoods and streets, and could be readily deployed to provide local-level information quickly to community partners and policymakers, as needed. Future efforts should include developing methods to maximise sensitivity and specificity of online keyword searches and continually comparing this approach to other data sources, like the tobacco licensure list, to see how this approach compares to existing methods. Additionally, states should consider distinguishing ENDS vape stores from traditional tobacco retailers during the licensing phase so ENDS vape stores can be tracked more effectively.

What this paper adds

  • The lack of regulatory policies and licensing of Electronic Nicotine Delivery Systems (ENDS) vape stores make it difficult to track the growth of such stores and their geographic concentration in certain communities. To date, no studies have examined the utility of online search methods to identify ENDS vape stores.

  • This study finds that an online search methodology, coupled with crowdsourced data collection, may be a promising strategy for identifying ENDS vape stores. Through this process we were able to identify a substantial number of vape stores that were not on a state tobacco licensure list.

  • Findings could help inform other states and researchers in their efforts to track the proliferation of ENDS vape stores, especially in states without a systematic licensing requirement for such stores.

Acknowledgments

The authors thank Christine Hebert for assistance with data collection and Brian Bradfield for assistance with data analysis.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors AEK conceptualised the study and led the protocol development, and drafted the manuscript. BL contributed to the study design, protocol development and analyses. BR contributed to protocol development, analyses and drafting of manuscript. CL contributed to the analyses and manuscript drafts. ME and LP contributed to manuscript drafts.

  • Funding The Florida Department of Health provided funding for the study. LP from the Florida Department of Health is a co-author of the study.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Please direct data sharing requests to AK.

  • i This list was pulled from the Florida Department of State website: http://dlis.dos.state.fl.us/fgils/cities.html.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.