Context: As of November 2018, medical cannabis was legal in 33 US states and recreational cannabis in 10, mostly enacted via ballot initiative.
Methods: We identified 32 cannabis legalization initiatives from 2004 to 2016 and obtained campaign contribution and state political and demographic data. After exploratory analyses of 15 potential independent variables, we quantified effects of 4 factors (initiative year, voter turnout, population born before 1946, advocate funding advantage) on voter support and likelihood of passage.
Findings: A small number of campaign contributors dominated both sides of the issue, with little involvement by health advocates. Time and turnout, not money, were the factors most associated with electoral outcomes, consistent with increases in public opinion favoring cannabis legalization over time. Year, turnout, and population age were associated with voter support, while year, turnout, and advocate funding advantage were associated with likelihood of passage. When adjusting for an anomalous result, initiative year was the only variable that remained significantly associated with odds of passage, with a 1-year increase in initiative date associated with 2.02 times higher odds of passage (p < .01).
Conclusion: These results underscore the importance of health advocate participation in developing cannabis legalization frameworks.
Cannabis laws are changing. As of November 2018, over 1 in 5 Americans lived in a jurisdiction with a recreational cannabis law, and over 2 in 3 lived in a jurisdiction with a medical cannabis law. Underlying these changes, public support for cannabis legalization has grown from 12% in 1969, to 31% in 2000, to a 62% majority in 2018 (Hartig and Geiger 2018).
Following enactment of the federal Controlled Substances Act in 1970, several states took a different approach and began adopting limited decriminalization measures reducing penalties, deprioritizing enforcement, or loosening state-law restrictions on cannabis cultivation, possession, and consumption (Pacula, Chriqui, and King 2003). The movement toward more permissive state cannabis laws stalled during the 1980s, but in 1996, California became the first state to legalize cannabis for medical use (NCSL 2018a). By July 2018, 14 states had decriminalized possession of small amounts of cannabis, 31 states and the District of Columbia (DC) had legalized medical cannabis, and 9 states and DC had legalized recreational cannabis (MPP n.d.-b; NCSL 2018a). Most of these new laws (9 of 10 recreational and 16 of 31 medical) were adopted via ballot initiative rather than legislative process (NCSL 2018b). The ballot initiative process, a lawmaking device available in 23 US states, allows citizens to vote directly on a proposed statute or state constitutional amendment, bypassing the state legislature.
Both driving and capitalizing on changing public opinion, well-organized campaigns have advanced legalization initiatives on state ballots across the country and led sophisticated efforts to pass them, often successfully. We examine the funding supporting and opposing campaigns to pass cannabis decriminalization, medical legalization, and recreational legalization ballot initiatives from 2004 to 2016 to understand the role of funding within the context of other political and demographic factors in explaining the success of legalization campaigns and to assess the future trajectory of the legalization movement.
We identified cannabis-related ballot measures using Ballotpedia.org (BP) (Ballotpedia n.d.) and FollowTheMoney.org (FTM) (National Institute on Money in State Politics 2018), cross-referenced with other compilations and analyses (MPP n.d.-b; NCSL 2018a; NORML n.d.; Rusche 2017). We excluded legislatively adopted laws, laws in nonstate US territories, referenda, initiatives that failed to qualify for the ballot, initiatives prior to 2004 (the first year for which FTM data were available), and initiatives addressing only changes in tax or fee levels. Our final dataset included 32 state-level cannabis ballot initiatives in 16 states from 2004 to 2016. (Additional methodological details are provided in the appendix.)
We classified each initiative as Decriminalization, Medical, or Recreational based on initiative text and descriptions obtained from BP and the National Conference of State Legislatures. We obtained vote totals from BP, voter turnout data from the US Census Bureau (US Census Bureau n.d.), population demographics from the Centers for Disease Control and Prevention (CDC n.d.), and state legislature and governorship data from BP and the National Conference of State Legislatures (NCSL 2018c).
We included all contribution types (e.g., direct, in-kind) for each registered ballot campaign committee to account for total available resources and excluded contributions from one committee to another to avoid double counting. Though requirements vary by state, organizations that accept resources to support or oppose a ballot initiative are required to register with a state agency (often the secretary of state) and disclose campaign-related contributions and expenditures. Contributions are typically used by committees in relation to gathering petition signatures to place an initiative on the ballot, obtaining office space and legal assistance, employing staff and consultants, developing campaign materials, purchasing advertising time or space (e.g., print, television, radio, social media, billboards), voter canvassing, and similar campaign activities, though specific uses depend on political environment and strategy.
For committees supporting/opposing more than one measure in the same election, it was not possible to distinguish between contributions associated with each measure; therefore, we attributed all contributions to the cannabis-related initiative to account for available resources. We adjusted contributions to 2016 dollars using the Consumer Price Index for January of the election year (BLS n.d.).
We assessed cannabis industry affiliations of the 10 largest contributors to each campaign using FTM data, membership in the National Cannabis Industry Association (NCIA) using the NCIA “Member Search” tool (NCIA n.d.-a) as of May 2018, and basic Google search techniques. The NCIA directory does not provide date of membership, so membership status at the time of contribution is unknown. We limited analysis of industry affiliation to the 10 largest contributors because they represented the overwhelming share of pro-legalization funding in most cases (appendix table A1). For political action committees (PACs), we obtained supplemental contribution records from OpenSecrets.org (CRP 2018) and approximated industry contribution share based on contributions to the PAC during the applicable year, as detailed in the appendix.
We calculated a “funding advantage score” using campaign contributions as ((Total PRO÷(Total PRO + Total CON)) × 100 − 50)÷10, with each point representing a 10% funding advantage. Positive values indicate a funding advantage for legalization proponents; negative values indicate a funding advantage for opponents.
We used two dependent variables: voter support and initiative success. We measured voter support as a continuous variable equal to percentage of votes cast in favor of legalization. We measured initiative success as a binary variable (1 = pass; 0 = fail).
As detailed in the appendix, we collected data on 15 independent variables (initiative year, election cycle, voter turnout, population born before 1946, governor's political party, legislative political party control, total legalization advocate/opponent contributions, advocate/opponent contributions per vote cast, advocate/opponent contributions per eligible voter, advocate funding advantage, initiative type, and cannabis industry percentage of top 10 donor contributions).
We used stepwise regression to select four independent variables for our final model: initiative year, voter turnout, percent of the population born before 1946 (the so-called “Silent Generation,” which has the lowest support for legalization [Hartig and Geiger 2018]), and advocate funding advantage. We also report univariate analyses of contribution data over time to provide context for discussion of donor characteristics and differences between initiative types. We tested for goodness of fit by examining leverage and residuals. Analysis was done using Stata version 14.2.
Initiative Types and Outcomes
Of the 32 initiatives, 3 were Decriminalization, 15 were Medical, and 14 were Recreational; 18 succeeded and 14 failed. Of 16 states represented, 12 had multiple initiatives from 2004 to 2016.
From 2004 to 2011, prior to the first successful recreational initiatives in Colorado and Washington, voters approved 1 of 3 decriminalization initiatives, 4 of 8 medical initiatives, and 0 of 3 recreational initiatives (figs. 1A and 1B). Of 6 states where initiatives failed, 5 (Alaska, California, Colorado, Nevada, Oregon) subsequently passed recreational initiatives. From 2012 to 2016, there were no decriminalization initiatives, and voters approved 5 of 7 medical initiatives and 8 of 11 recreational initiatives. Of 5 states where initiatives failed, 4 subsequently enacted either medical legalization (Arkansas and Florida by initiative, Ohio legislatively) or recreational legalization (Oregon), and 1 had previously passed medical legalization (Arizona).
Legalization advocates raised a total of $139.1 million (in 2016 dollars) compared to $37.3 million for opponents (table 1). Single-election contributions to advocates ranged from $34,000 to $29.3 million, with a mean of $4.3 million and median of $1.7 million. Single-election contributions to opponents ranged from $0 to $8.7 million, with a mean of $1.2 million and median of $256,000. Pro-legalization advocates had a funding advantage in 29 of 32 contests, with a mean advantage score of 3.3 (equating to 81% of funding going to advocates) and median advantage score of 4.1 (equating to 92% of funding to advocates). (Appendix table A2 presents state-level data.)
Without controlling for other variables, contributions generally increased over time. Total advocate contributions increased by $587,000 ± $258,000 (SE) per year (p < .05), but there was no significant change in advocate contributions per voter over time (p = 0.447). Total opponent contributions increased by $198,000 ± $83,000 (SE) per year (p < .05), and opponent contributions per voter increased by $0.08 ± $0.03 (SE) per year (p < .05). There was no statistically significant relationship between advocate funding advantage and time (p = 0.206) in univariate analysis. (Results are summarized in appendix table A3.)
There was a statistically significant difference by initiative type for total advocate contributions, per-voter advocate contributions, total opponent contributions, and per-voter opponent contributions, but not for advocate funding advantage (Table 2). Post hoc testing revealed significant differences between Recreational and Medical initiatives for the same four measures, but no other differences were significant when adjusted for multiple comparisons.
Major Contributors and Funding Concentration
Two major advocacy organizations were dominant contributors across several elections: the Marijuana Policy Project (MPP) and the Drug Policy Alliance (DPA, affiliated with the Center for Policy Reform and with Drug Policy Action). The MPP, the DPA, or both were involved as major contributors or ballot committees in 28 of 32 initiatives. Among the 20 largest pro-legalization contributors (appendix table A4), 9 contributed in multiple states: MPP (14), DPA (13), members of the Lewis Family (7), New Approach PAC (5), Henry Van Ameringen (5), Dr. Bronner's Magic Soaps (5), Thomas Cody Swift (3), George Soros and the affiliated Open Society Policy Center (2), and Weedmaps owner Ghost Management Group (2). A group of 10 Ohio business entities, representing the intended recipients of exclusive and oligopolistic cultivation rights under the state's proposed 2015 initiative (Graham 2015), contributed only in Ohio but were collectively the largest funder overall, surpassing even the MPP's combined contributions.
As noted by others (Rusche 2017), the MPP received funding from Peter Lewis (the late CEO of Progressive Insurance), who served on the MPP's board. Lewis, his brother Daniel, and son Jonathan, also contributed directly to campaigns in seven states. Similarly, the DPA received funding from major progressive political donor George Soros and his Open Society Policy Center, who also contributed directly to campaigns in California and Massachusetts. Because both the MPP and the DPA also receive contributions from other donors and the total funds provided by specific donors were not available, the MPP's and the DPA's contributions are considered separately from major individual donors in this analysis.
Among the 20 largest antilegalization contributors (appendix table A5), casino magnate and conservative political donor Sheldon G. Adelson was by far the largest contributor, providing over 3 times the resources of the next largest contributor, the Arizona Chamber of Commerce. Among the 20 largest opposition contributors, only 4 contributed in more than one state: Melvin F. Sembler and Betty Sembler and their advocacy group Save Our Society from Drugs (7), Julie Schauer (5), Adelson (4), and Smart Approaches to Marijuana (SAM) / SAM Action, Inc. (3).
Advocate contributions were approximately similar (fig. 2a) for individuals and entities contributing over $10 million (35%) and those contributing between $5 and 10 million (31%), with smaller shares coming from those contributing less than $1 million (25%) and between $1 and 5 million ($9%). If the unusual Ohio election is excluded (fig. 2b), the distribution shifts away from the largest donors over $10 million (23%), and toward those between $5 and 10 million (37%) and less than $1 million (30%). Opponent contributions were approximately evenly divided (fig. 2c), with approximately one-third coming from contributors over $10 million (33%), those between $1 and 5 million (32%), and those under $1 million (35%). Notably, there were no contributors between $5 and 10 million, in stark contrast to advocates. Excluding Ohio (fig. 2d) did not substantially change the relative contribution shares. (Donor totals within each tier are presented in table 3 and table 4.)
Cannabis Industry Contributions
Total contributions from industry-affiliated top 10 donors to legalization campaigns (appendix table A3) did not change significantly over time (p = 0.242). However, the percentage of industry-affiliated top 10 contributions increased by 2.3% ± 1.0% (SE) per year (p < .05). Industry share of top contributions exceeded 5% in only 7 cases, but 6 of these were in 2015–16 (appendix table A1). Industry contribution share of top contributions differed significantly by initiative type, but no significant differences were observed in post hoc testing when adjusted for multiple comparisons (Table 2).
Factors Associated with Voter Support and Initiative Passage
Initiative year, turnout, and population born before 1946 (Silent Generation) were significantly associated with voter support (table 5). Each additional year was associated with a 1.40% ± 0.38% (SE) increase in voter support (p < .01). A 1% increase in turnout was associated with a 0.38% ± 0.15% (SE) increase in voter support (p < .05). A 1% increase in Silent Generation population share was associated with a 1.35% ± 0.47% (SE) increase in voter support (p < .01).
Initiative year, turnout, and advocate funding advantage were significantly associated with increased odds of passage (table 5). Each year was associated with 1.90 times higher odds of passage (OR 1.90; 95% CI: 1.17–3.08; p < .05). A 1% increase in turnout was associated with 1.20 times higher odds of passage (OR 1.20; 95% CI: 1.03–1.39; p < .05). A 1-point increase in advocate funding advantage score (representing a 10% shift toward advocates) was associated with 2.23 times higher odds of passage (OR 2.23; 95% CI: 1.00–4.94; p = .049).
However, when Florida's 2014 initiative (57.6% vote in favor but failed due to 60% threshold under state law) was counted in the analysis as passing because it received over 50% of the vote, as required in most elections, only year was significant, with each year associated with 2.02 times higher odds of passage (OR 2.02; 95% CI: 1.22–3.35; p < .01).
Existing studies on cannabis initiatives have assessed public opinion over time and have drawn associations to factors that inform our analysis (e.g., age, political ideology), but do not tie findings to electoral outcomes (Toch and Maguire 2014) or are limited to a single state (Collingwood, O'Brien, and Dreier 2018). Another study assessed historical trends and predicted future electoral outcomes based on generational turnover, election cycle, and changing public opinion, but did not address campaign funding (Caulkins et al. 2012). One study addressed factors influencing legalization ballot initiative success, including funding, but was primarily qualitative and limited to four recreational initiatives from 2010–12 (Leon and Weitzer 2014). That study hypothesized that advocate funding advantage would be a substantive predictor of initiative success but found that it was not determinative. It also analyzed the influence of political climate (election cycle, federal opposition, elected officials' positions) and noted the potential significance of voter age distribution.
Newspapers and magazines have directed occasional attention to the identities and relationships of legalization initiative supporters and opponents (Cadelgado and Miller 2016; Graham 2017; Riggs 2012), but typically address only a single initiative or general trends. A white paper produced by an anti–substance abuse group tallied contributions for supporters and opposition but examined only the former in depth (Rusche 2017), and a campaign funding and lobbying research group summarized major contributors prior to the 2016 election cycle (Gurciullo 2015).
This article bridges existing analytical approaches by investigating campaign funding both for and against legalization and examining its interplay with political and demographic indicators across multiple electoral outcomes to better understand the past and future trajectory of cannabis legalization initiatives. Combined with research demonstrating the pro-industry structure of developing state recreational cannabis laws (Barry and Glantz 2018), our results indicate that the cannabis industry will benefit from these initiatives despite limited or attenuated participation in financing their passage. At least to date, the cannabis industry appears to have a different relationship to ballot initiatives compared to other industries.
Initiatives (and related but distinct referenda) arose from late-nineteenth-century Populist and early-twentieth-century Progressive goals of circumventing legislatures deemed beholden to special interest groups (Matsusaka 2018; Laposata, Kennedy, and Glantz 2014). One modern critique holds that because of the influence of campaign advertising, direct democracy may provide wealthy special interests the type of advantages it was intended to prevent (Matsusaka 2018). However, an empirical analysis of state initiatives affecting the energy, finance, and tobacco industries found that such initiatives rarely resulted in pro-business laws (2%), more frequently either failing (74%) or resulting in laws against industry interests (24%) (Matsusaka 2018: 44).
In response to initiatives contrary to industry interest, the tobacco industry, through its political arm, the Tobacco Institute, created a front group and partnered with other “ballot-prone industries” to monitor state ballot initiatives and advocate for reforms that would make the initiative and referendum processes more difficult (e.g., increased signature requirements, shorter time to gather signatures, supermajority vote requirements for tax increases, single-subject restrictions) (Laposata, Kennedy, and Glantz 2014). These efforts achieved only limited initial success, but appear to have amplified existing policy arguments and positions regarding direct democracy that were eventually adopted in some states (Laposata, Kennedy, and Glantz 2014). The tobacco industry has also more recently used a tactic of introducing a competing “look-alike” initiative on the same subject that is more advantageous to the industry through weaker regulation and preemption of stronger local laws (Tung, Hendlin, and Glantz 2009).
Voter Support and Initiative Success
The only factor consistently associated with both voter support and initiative success for cannabis legalization initiatives was election year, with later contests receiving a higher proportion of votes and being more likely to succeed. This is consistent with changes in public opinion polls (ProCon.org n.d.). Nationally, support for legalization has risen sharply from 16% in 1990 to 62% in 2018 (Hartig and Geiger 2018). State polls from 2015 to 2017 found general support for medical legalization (not including specific proposed laws) from sizeable majorities of voters: 61–79% in Utah; 70% in Iowa; 80% in North Carolina; 84% in Ohio and Florida; and 88% in Pennsylvania (ProCon.org n.d.). State polls from 2018 showed substantial support for recreational legalization (though lower than for medical): 59% in Connecticut and Virginia; 61% in Texas; 62% in New Jersey; and 63% in New York (Quinnipiac University n.d.).
Voter support for specific legalization initiatives is often lower than for legalization in the abstract. For example, in Arkansas, polls showed 84% support for medical legalization in 2015 (ProCon.org n.d.), but a 2016 medical initiative passed with only 53.1% of the vote. Gaps between general support and electoral results may indicate that voters support legalization but remain concerned with policy details, including commercialization and industry structure. If so, this represents an opportunity for health advocates to influence legalization approaches to protect public health.
Gaps between general support and voting results may also indicate an enthusiasm gap between legalization opponents and supporters. This gap would be consistent with the observed association between turnout and both initiative support and success, as higher-turnout elections should better reflect population views. Political choices by legalization advocates on where and when to advance initiatives seem to reflect this situation. Since 2010, 14 of 18 initiatives were in presidential cycles, where turnout is typically higher, particularly among younger voters, who are generally more supportive of legalization (Hartig and Geiger 2018).
The surprising positive association between the Silent Generation population and voter support may be explained by the fact that while this group expresses the lowest support for legalization, support among this cohort has still risen over time (Hartig and Geiger 2018), and they remain a demographic minority despite higher voter turnout (Caulkins et al. 2012). (The association between Silent Generation population and initiative passage was not significant, p = .07.)
The lack of significant association between legalization advocate funding advantage and voter support is contrary to existing explanations of legalization campaigns, which attribute success to overwhelming funding by out-of-state donors (Riggs 2012; Rusche 2017). Legalization advocate funding advantage was associated with odds of passage, but the association was just under the threshold of significance (p = 0.049) and became nonsignificant when adjusting for Florida's 2014 election (in which a medical legalization initiative received a majority of votes but failed to reach the supermajority threshold required to amend the state's constitution). Advocates were routinely better funded than opponents (29 of 32 elections), but other factors appear to better explain electoral outcomes.
Despite these findings, funding may still play a significant role in initiative success. Some funding is a prerequisite for placing initiatives before voters (e.g., for signature collection). Legalization campaigns have also likely increased the issue's salience among voters, which may be reflected in time-delayed changes in opinion not captured by our methodology. It is also possible that advocates' persistent funding advantage (29 of 32 elections), especially when combined with the cross-jurisdictional reach of many modern election advocacy tools (e.g., social media), may have had a cumulative effect on electoral outcomes. However, most of the overall funding advantage is attributable to elections in the last two included cycles, 2015–16, and largely to the heavily skewed Ohio and California elections. While we found an association between funding advantage and likelihood of passage, overall our findings do not support the conclusion that campaign funding is the sole or dominant explanation for legalization initiative success.
Major Donors and Funding Concentration
Many of the largest advocate donors limited their contributions to a single state (as did many opposition donors), though their influence may be broader due to donations to major advocacy groups like the MPP and the DPA, which did contribute in multiple states. A small number of individual donors, including George Soros and Peter Lewis, did contribute in multiple states, as did major opposition contributors Sheldon Adelson and Melvin and Betty Sembler. Both advocates and opponents received approximately one-third of their contributions from donors giving over $10 million, though the true share for advocates may be higher when accounting for the additional contributions from large individual donors to major advocacy groups. Overall, contributions for both advocates and opponents were concentrated among a small number of donors, but legalization advocates received larger contributions. Both advocates and opponents also received a substantial share of their contributions (25% and 35%, respectively) from relatively smaller donors (less than $1 million), but, consistent with overall trends, contributions to advocates from this group were far higher in absolute dollars ($35.3 million to $12.9 million).
Cannabis industry growth has not yet translated into a sizeable share of contributions to legalization campaigns. Few of the largest overall contributors had known industry affiliations, other than those contributing to the 2015 Ohio initiative. Industry-affiliated donors were among the largest contributors within some individual elections, but provided over 5% of top contributions in only 7 of 32 cases. However, 6 of these 7 cases were in 2015–16, and the share of contributions attributable to industry-affiliated donors has risen over time, which may reflect the beginning of a more substantial trend. Elections in 2018 and 2020 may clarify whether the industry's funding role in these initiatives is increasing meaningfully.
Because we did not count the MPP or other advocacy groups as industry affiliated, this likely underestimates industry influence. For example, the MPP and the NCIA have had board members in common (Huddleston 2016), and the MPP was the largest contributor toward 13 of 32 initiatives, in many cases serving as the only significant donor. While beyond the scope of this article, the cannabis industry's influence on and within advocacy organizations responsible for drafting initiative language warrants scrutiny, especially in light of the strongly industry-friendly provisions of such initiatives to date, most conspicuously the proliferation of a commercial, for-profit framework. Overall, the role of the cannabis industry in shaping cannabis policy to create a private commercial market that prioritizes business interests over public health presents considerable policy concerns (Barry and Glantz 2016, 2018) and requires continuing study and attention to avoid future public health harms replicating those of tobacco and alcohol.
Nongovernmental health organizations were largely absent from the major donor rolls of both legalization advocates and opponents with the exception of some hospitals and state medical associations opposing legalization in seven elections: Alaska 2004 (Alaska State Medical Association, Alaska Kidney and Diabetes Association), Michigan 2008 (Michigan Health and Hospital Association), California 2010 (California Association of Hospitals and Health Systems), South Dakota 2010 (South Dakota State Medical Association), Alaska 2014 (Alaska Regional Hospital), Ohio 2015 (Ohio Hospital Association, Ohio Children's Hospital Association), and Massachusetts 2016 (Partners Healthcare System, Health Foundation of Central Massachusetts).
As of July 2018, 23 states authorized statutory or constitutional changes via ballot initiative (Initiative and Referendum Institute n.d.). Among these, 8 had recreational and medical cannabis laws, 8 had medical cannabis laws only, and 7 had no cannabis legalization or limited medical laws (National Conference of State Legislatures 2018a). Four states have medical or recreational legalization on the ballot in November 2018 (MPP 2018a). Based on observed trends, advocates will likely propose more initiatives in 2020 and beyond, and many or most will pass.
The most important factor we observed explaining legalization initiative support and success is time, consistent with rising public opinion favoring legalization (Hartig and Geiger 2018). To the extent this is true, future initiatives will be increasingly likely to pass. This change over time also likely reflects changes in advocates' tactics. For example, legalization advocates have altered campaign rhetoric from concentration on personal freedom to emphasis on tax revenue, social justice, and comparisons to alcohol policy, and they have more strategically approached media appearances, political endorsements, and resource allocation (Ingold 2012; Leon and Weitzer 2014). While these changes may influence the association between time and legalization success, none are directly related to funding.
One potential change in policy environments that may slow the spread of initiatives, but not of legalization itself, may be legalization legislation. Vermont legislatively enacted recreational legalization effective in 2018 (Vermont General Assembly, H. 511 (Act 86). 2017–2018, with a system for lawful sales to be determined later) and several other states were considering recreational cannabis legislation as of early 2019. Most of these states do not have a ballot initiative process, but, if other state legislatures follow suit, this may forestall additional legalization initiatives.
Legislative legalization may also allow nongovernmental health groups and governmental health agencies to influence the construction of these laws to better protect public health, because they will be able to do so without committing funding to a ballot initiative, which they may be politically or legally restrained from doing. The general absence of nongovernmental health organizations among top contributors, other than a small number of hospital groups and state medical associations, is consistent with their lamentable absence from the current legalization policy debate (Barry and Glantz 2018). Because the top donors in most cases represented the vast majority of campaign funding for and against initiatives, health groups' absence suggests a lack of significant influence. If health groups continue to sit on the sidelines during development of cannabis legalization laws, such laws will likely continue to prioritize business over health-related goals (Barry and Glantz 2018). To protect public health, health groups must become involved not only in the development of implementing regulations for cannabis laws, but in the shaping of legalization frameworks at the foundational level, including contributing to the drafting of legalization statutes informed by public health best practices.
Whether by initiative or legislation, further state legalization appears likely if public opinion remains in favor. Public health advocates will need to get involved in the policy debate to increase the likelihood that states will regulate cannabis in a way consistent with public health best practices informed by lessons from tobacco and alcohol control (American Public Health Association 2014, Barry and Glantz 2016, 2018; Carnevale et al. 2017; Cork 2018; Mosher 2016; Orenstein and Glantz 2018). Legalization, especially in any specific form or in any particular jurisdiction, is not inevitable. Health advocates have the opportunity—and the responsibility—to work to shape, slow, or stop policy developments as dictated by their assessment of health impacts.
We relied on the FTM database for contribution data, which limited analysis to 2004 and later, and earlier measures may differ. For ballot committees supporting/opposing multiple initiatives in one election, available data did not allow us to distinguish between contributions attributable to each ballot measure. We attributed all contributions to cannabis legalization to reflect available resources, which may have resulted in overcounting.
We also did not include independent expenditures (i.e., expenditures related to the initiative that do not pass through a registered committee). These appeared to be infrequent and typically minimal in comparison to contributions to registered committees according to available data, and in some cases such expenditures were traceable to registered ballot committees, meaning their inclusion could have resulted in double counting. However, it is possible that excluding independent expenditures may have impacted our analysis by omitting or discounting some contributors.
We attempted to capture relationships among individuals, business entities, and political groups, but some may be obscured by opaque ownership structures and the absence of mandatory disclosures for some entities. Our assessment of cannabis industry ties was limited to FTM data, the NCIA's membership directory, and publicly available websites and news sources. Some industry ties may not be captured by this methodology, so our analysis likely underestimates industry contributions. We limited analysis of industry affiliation to the 10 largest donors for each initiative because these donors typically represented the vast majority of contributions and we therefore assumed that smaller-dollar contributors were less influential. However, smaller donors may have other forms of influence, and aggregate industry contributions among these donors may be significant. Moreover, our analysis does not include other important forms of influence, such as the role of industry in shaping the content of initiatives.
Exogenous factors that influence elections may vary substantially between states and election cycles. We attempted to account for these variations, but factors such as other ballot items, political climate, and other trends can influence turnout, electorate composition, political advertising costs, and other variables not captured by our methodology. These and other idiosyncrasies of state elections suggest a cautious interpretation of statistical results and discourage drawing conclusions about applicability to any particular election.
Our analysis was limited by the small number of observations (n = 32), which may limit power and obscure statistical trends. The particularly small number of Decriminalization initiatives (n = 3) may have also obscured distinctions between the three initiative types. Particularly given the small number of observations, the absence of statistical significance should be interpreted cautiously and not as conclusive evidence of the absence of a relationship.
Money matters in politics, and both advocates and opponents of cannabis legalization appear to agree with this, based on the large contributions in legalization ballot measure contests. However, our analysis challenges arguments that fundraising has been the most critical factor in the success of these initiatives. While funding is undoubtedly important, other dynamics are highly relevant, and our findings are more consistent with the electoral consequences of a major shift in opinion among the American electorate and increasingly savvy political choices by legalization advocates about where, when, and how to advance ballot initiatives. If this is the case, legalization measures are likely to continue to be introduced and to pass. It is therefore critical that public health advocates actively participate in shaping cannabis legalization, with a focus on regulating cannabis well, consistent with public health best practices from domestic and international models.
This research was supported in part by National Institute on Drug Abuse grant DA-043950. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The authors thank Dharma Bhatta, Amy Hafez, Lauren Lempert, and Tanner Wakefield for their comments and feedback. The authors also acknowledge the work of the National Institute on Money in Politics, which created the FollowTheMoney.org database used by the authors.
Appendix: Data and Methods
We used Ballotpedia.org (BP) (Ballotpedia n.d.) (BP) and FollowTheMoney.org (National Institute on Money in State Politics 2018) (FTM) to identify 60 cannabis-related ballot measures from 1972–2016. We compared this list to other compilations and analyses (NCSL n.d.-a; MPP n.d.-b; NORML n.d.; Rusche 2017) and identified three additional measures. We did not consider laws adopted via legislatures (MPP n.d.-b; NCSL n.d.-a) (n = 43) or nonstate US territories (e.g., Puerto Rico) due to distinctions in legal structure and absence of records in the FTM database. We ultimately excluded referenda (n = 1), which are distinct because they originate following legislative action, unlike initiatives.
We excluded initiatives that failed to qualify for the ballot (n = 2) because contributions to groups supporting such measures would be limited in the absence of a full-scale campaign, and we excluded initiatives prior to 2004 (n = 23) because FTM data were not available for earlier elections. We then excluded initiatives addressing only changes in tax or fee levels rather than broader changes in the legal status of cannabis (n = 4). We excluded DC's Initiative 71 (2014) and Montana's Referendum IR-124 (2012) because contribution data were not available from FTM or a comparable source for the DC initiative and the Montana measure was not a citizen initiative, but rather a referendum on a gubernatorial veto of legislative changes to a previously enacted medical cannabis law (which was separately included in the analysis). (The referendum was also challenged in court as being confusing to voters.) Our final dataset included 32 state-level cannabis ballot initiatives in 16 states from 2004 to 2016 (appendix table A1).
FTM data were missing or incomplete for three initiatives in 2016. We therefore supplemented FTM with data from state resources for Arkansas (Arkansas Ethics Commission n.d.), Montana (Montana Commissioner of Political Practices n.d.), and Nevada (Nevada Secretary of State 2017).
We classified each initiative as Decriminalization, Medical, or Recreational based on initiative text and descriptions obtained from BP and the National Conference of State Legislatures. We classified as “Decriminalization” initiatives that reduced or eliminated penalties for cannabis possession and/or cultivation without creating a system for lawful sale. While distinctions within this group are important from a policy perspective (Pacula, Chriqui, and King 2003; NCSL 2019), the absence of lawful sales was distinctive for the group as a whole for purposes of this analysis because such frameworks do not present the opportunity for (lawful) profitmaking from cannabis sales. We classified as “Medical” initiatives that authorized individual cannabis possession and/or cultivation based on medical condition and sale by entities or individuals approved or licensed in some fashion. We did not differentiate based on qualifying conditions, program structure (e.g., patient collectives, nonprofit requirements), or the process for approval or licensure (e.g., municipal vs. state licensure). We classified as “Recreational” initiatives that authorized possession by any adult of legal age and sales by licensed retailers. We classified Ohio's Issue 3 (2015), which would have legalized both medical and recreational cannabis, as “Recreational” because it included the more expansive recreational system.
We obtained vote totals for each measure from BP. We obtained voter turnout data from the US Census Bureau except for Maine 2009 (which were not available. To determine Maine's turnout, we used total votes cast regarding Question 5 divided by the average of 2008 and 2010 Maine citizen population over age 18 reported by the Census Bureau.) We obtained population demographics from the Centers for Disease Control and Prevention (CDC n.d.). To assess contemporaneous political environment in each state, we included variables for partisan control of the state legislature and governorship at the time of election, obtained from BP and the National Conference of State Legislatures (NCSL n.d.-c).
To analyze the 10 largest contributors for and against each measure, we used total contributions by each individual or entity and excluded un-itemized contributions (aggregated small contributions below reporting thresholds). Where multiple contributors gave equally in the 10th position, we included each in calculating industry contribution share (resulting in more than 10 entries for some states) to more fully assess industry relationships, but did not include ties in calculating the percentage of total contributions attributable to the 10 largest donors to assess funding concentration.
While the industry has clear ties to major advocacy groups like the Marijuana Policy Project (MPP) (including board members in common between MPP and NCIA [Huddleston 2016; NCIA n.d.-b]), we did not consider MPP or similar major advocacy groups to be part of the industry for purposes of this analysis. If MPP was counted as an industry group, several early pro-legalization efforts would be reported as entirely or nearly entirely industry funded (because MPP was the primary contributor), which would obscure the later entry of other contributors poised to more directly benefit financially from legalization.
For political action committees (PACs), we obtained supplemental contribution records from OpenSecrets.org (CRP 2018) and approximated industry contribution share based on contributors listed as NCIA members, counting their PAC contribution for the applicable year as industry contributions. Because money is fungible, it is not possible to determine which PAC contributions were in turn contributed to any particular election. Therefore, we approximated an industry contribution by calculating the share of total contributions to the PAC attributable to NCIA members and applying this percentage to the PAC's subsequent contribution. For example, in 2016 Privateer Holdings (an NCIA member) contributed $50,000 to New Approach PAC, which received $6,968,000 in total contributions that year (CRP 2018). For each 2016 contribution made in turn by New Approach PAC, we counted 0.72% ($50,000/$6,968,000) as an industry contribution. We did not, however, add PAC contributions to individual or entity totals in determining the top 10 donors because this would result in double counting.
To compare funding of advocates and opponents, we assessed both total contributions and contributions per voter to account for differing costs of campaigning in different states (e.g., due to population concentration or media expense). We also measured relative advocate and opponent funding. Because there were no contributions to opponents in some elections, we created a measure reflecting the share of total contributions represented by pro-legalization contributions (“funding advantage score”), using campaign contributions as ((Total PRO÷(Total PRO + Total CON)) × 100−50) ÷10, with each point representing a 10% funding advantage. Positive values indicate a funding advantage for legalization proponents, negative values indicate a funding advantage for opponents. To aid interpretation, we converted the data to a scale of −5 to 5, where 0 represents equal contributions to advocates and opponents (50% each); positive values represent advocate advantage and negative values opponent advantage, and each point is equal to a 10% increase. For example, a score of 1 means advocate contributions were 60% of total contributions, and a score of 5 indicates that advocate contributions were 100% of total contributions.
We began with the following 15 independent variables:
Initiative year (baseline 2010)
Election cycle (1 = presidential, 0 = nonpresidential)
Voter turnout percentage
Percentage of population born before 1946 (the so-called Silent Generation, which typically demonstrates lowest support for legalization in opinion polls)
Governor political party (1 = Democrat, 0 = Republican)
Legislature political party control (1 = Democrat or split, 0 = Republican)
Total advocate contributions (2016 dollars)
Total opponent contributions (2016 dollars)
Advocate contributions per voter (total advocate contributions divided by total votes cast)
Opponent contributions per voter (total opponent contributions divided by total votes cast)
Advocate contributions per capita (advocate contributions divided by adult citizen population)
Opponent contributions per capita (opponent contributions divided by adult citizen population)
Advocate funding advantage (total advocate contributions divided by total contributions to both advocates and opponents)
Initiative type (two dummy variables with Recreational as reference)
Cannabis industry percentage of top 10 donor contributions
Using these independent variables, we performed backward, forward, and backward/forward stepwise linear regressions with voter support as the dependent variable, and we performed backward, forward, and backward/forward stepwise logistic regressions with initiative success as the dependent variable. We also performed the same logistic regressions with Florida's 2014 initiative counted as a success. Because the logistic regressions did not achieve convergence, we removed advocate and opponent contributions per capita on the basis that per-voter totals adequately captured this variable. We removed election cycle after determining that election cycle and turnout were strongly correlated (r = 0.77) and that turnout was a more useful variable because it varies within election cycle and may reflect allocation of resources targeting voter outreach. After removing these three variables, we reran all stepwise regressions.
We compared the variables with significant associations in each stepwise regression to select variables for inclusion in our final model. Because our data set included only 32 data points, we determined that the model should include no more than 4 variables to avoid overspecification. Only 4 of the 12 variables were statistically significant in at least one of the stepwise analyses: initiative year, voter turnout, advocate funding advantage, and percentage of the population born before 1946. All but advocate funding advantage were significant in multiple analyses. We included these 4 variables in our model. Total opponent contribution was marginally significant in backward/forward logistic regression on initiative success (p = 0.062); however, we did not include this variable in order to limit the model to 4 independent variables and because it is largely captured by advocate funding advantage (r = –0.64), which also incorporates advocate funding data.
Total and per-voter contribution levels were also correlated (r = 0.40 for advocate contributions; r = 0.78 for opponent contributions). As a check, we performed all stepwise regressions with per-voter advocate and opponent contribution measures but excluding total contribution levels. This change did not alter our observations.
We also report univariate analyses of contribution data over time to provide context for discussion. Because contributions (total and per-voter) did not appear normally distributed, we used nonparametric analyses. We used Kruskal-Wallis analysis of variance on ranks to determine whether a difference existed between samples, followed by Holm-Sidak adjusted post hoc Mann-Whitney rank-sum tests for differences between initiative types (decriminalization vs. medical, decriminalization vs. recreational, medical vs. recreational).