To be included in this study, polls had to meet the following criteria:
 

• Occur after November 1996. This was the month in which California voters passed Proposition 215. This single event launched medical cannabis into the national spotlight and laid the groundwork for today’s political battle.

• Utilize the principles of scientific polling. Polls qualified as "scientific" are usually taken from a randomly selected subgroup of the population being surveyed. Because the selection is random, the subgroup is thought to be representative of the population as a whole. Indicators of a scientific poll include references to a "margin of error" and "confidence levels."

• Fielded in the United States. Polling concerning medicinal marijuana has been active in Canada as well as the United States. In February 2002, The Ottawa Citizen reported that study, conducted by Price Waterhouse-Coopers for Health Canada, showed that over 85% of Canadians feel marijuana should be available medically. Another statistic reported in this survey – that 4% of marijuana use is medicinal – was pulled from this study to compute number of medical marijuana user in the United States and the ten-year arrest figure for medical marijuana offenses. Outside of this computation, all polls taken outside the United States have been excluded from this analysis.

• Asked a question pertaining to the medical use of marijuana. Some polls in this report are stand-alone surveys in which a polling firm is hired by a sponsor to conduct a study of medical marijuana only. The Ten State Poll (#56-65) by the Lucas Organization for the Marijuana Policy Project represents one example of a stand-alone poll. In other cases, such as The Virginia Quality of Life Survey by the Virginia Tech Center for Survey Research (#46, 32, 28, 14, and 4), a multiple question survey may contain one question about medical marijuana among others that are unrelated. For the purposes of this report, both kinds of studies are termed "polls."

Many organizations invite visitors to their Web sites by creating online polls, and a frequent topic is medical marijuana. For example, a 1998 CNN Internet poll, that gathered nearly 25,000 respondents, claimed that 96% "support(ed) the use of marijuana for medical purposes." While such a high percentage is heartening to those who share similar views, these results hold little more than publicity value. The underlying reason goes to the heart of scientific polling. For results to be valid, a sample should represent its population, and this doesn’t happen when a single respondent can answer multiple times, when a respondents’ relative location can’t be deduced, or when controls can’t be placed on demographic factors like age or race. Scientific Internet polling is only in its infancy. Thus, almost all Web-based polls are respondent-driven and therefore, not scientific.

To locate the polls that comprise this study’s 66-member list, OPN members began with a search of the Media Awareness Project (MAP), which archives newspaper, magazine, and Web articles on drug policy going back to 1992. After this initial review, which garnered numerous poll references, searches were conducted on the Google, Lycos, and other search engines. Web sites for such drug policy reform organizations as the National Organization for the Reform of Marijuana Laws (NORML), the Marijuana Policy Project (MPP), the Drug Reform Coordination Network (DRCNet), and the Drug Policy Alliance (formerly, the Lindesmith Center and Drug Policy Foundation) were checked as were the Web pages of such medical marijuana opponents as the Family Research Council (FRC) and the Office of National Drug Control Policy (ONDCP). Sites for the major polling organizations and their trade groups were also referenced. Essentially, the World Wide Web is the source for the polling data listed in this report.

To create this report, a database containing the polls was created from which several tables have been generated. In addition, cross tabulations by party from MPP’s Ten State Poll and subsequent survey of New Hampshire voters (#66) have been formatted into the tables that accompany this analysis.

While OPN feels that it conducted a thorough search of the Web for medicinal cannabis polling information, the data do have a few missing pieces. These holes are marked by "??" in the database and resulting tables. OPN feels that these missing data do not substantially detract from the overall findings. In fact, patching these holes only gives the data more power, not less. OPN welcomes any comments concerning these numbers and invites interested parties to provide information that fills the gaps.

Appendix III, which lists polls in descending order by date, numbers them from 1 to 66 beginning in 1996. It also contains only one percentage for polls in which two or more questions were asked. When creating this table and inputting the favor/oppose percentages for medical marijuana, OPN looked for the question that came closest to asking whether respondents favored or opposed the medical use of marijuana generally speaking. This also meant that polls, in which only one medical marijuana-related question was asked, defaulted to the only reported percentage.

One analysis technique used in this report aggregates all poll numbers into summary totals. OPN understands that sampling techniques and questioning vary from survey to survey and that summarizing disparate data can call into question assumptions gleaned from making cross-the-board calculations. However, such aggregations do provide a benchmark for summarizing the state of medical marijuana in the U.S. With an issue that rarely polls favorably at less than 50%, the very few polls whose favorability rate dips below half (Poll #52 and 51 – both related to voting) actually weight the data downward, meaning that the aggregate approval for medical marijuana may actually be higher.

One group of polls that can be analyzed aggregately and retain their scientific underpinning is the Ten State Poll (#56-65) conducted in March 2002. Approximately 1,000 people in each of ten Western states were asked the same five questions during the same time frame. For this report, each of these polls was counted separately because each had a separate margin of error and confidence level. However, their similarity allows data analysis of the same questions not only at the state level, but also aggregately and cumulatively.

Another data caveat that should be noted is grouping. Many surveys did not evoke simple yes/no responses. Some questions asked survey participants to categorize their responses into "Strongly support (or agree), somewhat support, somewhat oppose, strongly oppose (or disagree), and not sure/no opinion." With the exception of Appendix IV, this report groups all of the "support/agree" into one total and all "oppose/disagree" into another even if qualified by "somewhat." All "No answer/Not sure/No opinion" answers were placed in the "other" response column. The reason for this grouping goes back to California Proposition 215 and the other medical marijuana initiatives passed since then. Initiatives ask voters to only to signal approval or disapproval. Votes aren’t qualified. Aggregating all of the polling percentages along with placing a clear dividing line between support and opposition may provide some insight as to what the outcome might be if a medical cannabis initiative were to be voted on nationwide.

While telephone surveys by far remain the most common method of scientific public opinion polling, the advent of such technologies as Caller ID that individuals use to screen calls can bring into question the reliability of these kinds of studies especially when their results stand on narrow margins. Scientific polling rests on the principle that random samples are representative of a larger population, and when respondents choose whether or to not to simply answer the phone, the omission of their opinions obviously renders the sample less representative. The problem of selective answering that has emerged with Caller ID and other new technologies applies to all telephone polling, not just those dealing with medical marijuana.

Also, findings from the Erney, Busher & Associates poll of Columbus, Ohio, voters (#13) indicated an additional problem with telephone surveys, this one focused directly on the issue. Interviewers who conducted the survey reported a higher than average hang-up and refusal-to-answer rate and an obvious reluctance by respondents to answer all questions. Respondents who fail to answer completely or clearly, again, impact the reliability of the study.

Even with these problems and caveats, because so many studies have arrived at the same conclusion, the preponderance of favorable responses to the issue of medical cannabis isn’t diminished by anomalies in methodology.