From July 2009 through July 2010 we used the Legacy Tobacco Document Library (http://legacy.library.ucsf.edu
) to systematically search archived internal tobacco industry documents. Tobacco industry document research presents unique methodological challenges,30
and we followed standard techniques to conduct document research that combined traditional qualitative methods31
with iterative search strategies tailored for this data set.32
Document research is typically not a hypothesis-driven experiment. Thus, we initially posed exploratory, open-ended research questions, such as “What did tobacco companies know about people's preferences and patterns of alcohol use?” or “What did tobacco companies understand about the relationship between alcohol use and smoking?” Previous studies have documented increasingly frequent tobacco promotional activities in bars during the last few decades,27,28,33
so exploratory research questions related to marketing activities included such questions as “Was it a coincidence that tobacco and alcohol companies seemed to sponsor similar special events, or was it an intentional marketing strategy?” and “What marketing activities did tobacco companies conduct featuring alcohol?”
Initial keyword searches combined terms related to:
- marketing activities, such as joint sponsorship, cosponsor, joint promotion, cross promotion, tie in, plan, strategy, proposal, presentation, benefit, policy, guideline;
- promotional events, such as spring break, resort program, sports, soccer, softball, ski, golf, volleyball, auto racing, ethnic sports, concert, festival, bar night, nightclub;
- marketing research, such as focus group, qualitative, quantitative, research, study, consumer profile;
- concurrent use of tobacco and alcohol, such as co-use, cross usage, drinking, life styles, “drinking and smoking”;
- types of alcoholic beverage, such as beer, liquor, wine, cocktail, vodka, whiskey, champagne;
- alcohol brands, such as Miller, Budweiser, Olympia, Jack Daniel's, Canadian Mist, Coors, Schlitz, Southern Comfort;
- alcohol companies, such as Miller, MBC, Anheuser-Busch, Brown-Forman, Olympia;
- tobacco companies, such as Philip Morris, R.J. Reynolds, Brown and Williamson, British American Tobacco, American Tobacco, Lorillard; and
- cigarette brands, such as Marlboro, Virginia Slims, Benson & Hedges, Camel, Winston, Salem, Kool, Misty, Capri, Lucky Strike, GPC, Viceroy.
Initial searches yielded thousands of documents. For example, a search of all tobacco industry document collections on the Legacy Tobacco Documents Library for the keyword “cosponsor” yielded 10172 documents. Additional keywords were added to narrow searches. For example, a search of “cosponsor” and “Miller” yielded 1734 documents, and “cosponsor” and “Miller” and “soccer” yielded 164 documents. Searches were narrowed until they generally yielded fewer than 400 documents.
After we narrowed the searches, we read all the documents retrieved to determine their relevance to the exploratory research questions. We removed duplicate documents and documents that were clearly unrelated to our research questions (such as tobacco companies' research on behaviors other than alcohol consumption and tobacco companies' joint sponsorship with companies that did not sell alcohol). Documents containing the same text but different marginalia were counted as separate documents instead of duplicates.
Both authors then reviewed the documents in detail to discern their themes and contexts, such as the names of marketing studies, marketing promotional project titles, the time lines of these activities, and the relations between marketing studies and promotional activities. We wrote summary memoranda while reading the documents and listed more specific search terms from the documents retrieved, such as study titles (e.g., MMI Lifestyle Research, PSC Survey, GF Survey), marketing promotional project titles (e.g., Marlboro Soccer Cup, Winston-Budweiser Heritage Cup World Softball Tournament, Canadian Mist and Viceroy Rich Lights Designer Fashion Showcase), and other important information (e.g., key individuals responsible for these activities, third parties involved, dates of meetings, locations and dates of events).
We then asked a series of subsequent questions to guide further searches to help us better understand the context and implications of each document, such as “What was the purpose of the GF study?” “Did the results of this study affect marketing and promotional activities?” “What criteria did tobacco companies use to select alcohol companies and types of alcoholic beverage for joint sponsorship and joint promotion?” “What benefits did tobacco companies expect to gain from alcohol co-marketing activities, and did they obtain these benefits?” and “What were the job titles of the individuals mentioned in the documents, and what were the relationships among those people?”
We followed reviews of relevant documents with “snowball” searches to find related documents using additional search terms such as consecutive reference (Bates) numbers, file locations, dates, and individuals' associates. Through iterative searches, we attempted to gain a more comprehensive understanding of the research and marketing activities. We wrote additional memos incorporating the newly retrieved documents, noting the theme and context of each document, and we organized the documents in clusters by topic (e.g., research study, promotional project, guideline for marketing activities).
We repeated the iterative search process until we reached saturation of both keywords and documents. We based the present analysis on a final collection of 128 documents relevant to tobacco companies' research on alcohol consumption, concurrent use of tobacco and alcohol, and joint promotion of cigarettes and alcohol, and 358 documents related to marketing activities that linked cigarettes with alcohol. The documents we located ranged in date from 1974 to 1999. We limited our analysis to documents describing marketing activities in the United States.
To triangulate findings in the documents, we used additional data that we obtained from general online search engines (e.g., Google), news coverage (e.g., Google News or Lexis-Nexis), and advertising archives (e.g., the Trinkets and Trash archive at the School of Public Health of the University of Medicine and Dentistry of New Jersey, or the Pollay advertising collection at http://tobaccodocuments.org
). We used these adjunctive sources to validate and contextualize the marketing activities discussed in tobacco industry documents. Triangulation is a standard method that qualitative researchers use to check for validity and consistency of information among various independent sources; specific triangulation strategies using tobacco industry documents and advertising archives have been described previously.34
Conversely, we also retrieved advertisements with alcohol-related content from advertising archives; then, to find plans for these campaigns, we searched tobacco industry documents for these advertisements' taglines, slogans, or other descriptors. To build a comprehensive picture of how the different activities were related to each other both historically and conceptually, we made a summary of all research memos containing all major clusters of documents, summaries of all relevant documents, and direct quotes from the documents.
The following criteria guided iterative analyses to maximize our confidence in the credibility, dependability, and confirmability of the findings: we emphasized activities that were conducted repeatedly over time, activities that were pursued by multiple tobacco companies, plans that were created by or sent to people in positions of power within the tobacco companies (and thus were more likely to be carried out or to be representative of company policy), and activities that were supported both by internal documents and outside data sources.
We selected quotes that were consistent with the general themes across the multiple tobacco companies, that were consistent with their specific context in a document cluster, and that represented themes that were discussed repeatedly. We conducted this analysis to gain a greater understanding of tobacco industry research and marketing activities related to tobacco and alcohol, rather than to generate theory or to test specific hypotheses about these activities.