Literature review: Direct democracy and policy diffusion

There is much scholarship on initiatives and referenda, but most studies focus on two broad areas of research. One stream addresses whether direct democracy generates higher voter awareness, competence (Nicholson, Reference Nicholson2003), and turnout for ballot propositions, as well as the vote for candidates being elected on the same ballot (Biggers, Reference Biggers2011; Tolbert et al., Reference Tolbert, Bowen and Donovan2009). The other stream addresses two primary questions (Matsusaka, Reference Matsusaka2018): First, do states with initiatives and referenda have different policy outputs—for example, on taxes and spending—compared with states without direct democracy? A variant of this research asks whether initiatives and referenda are employed to enact anti-minority policies, unlike enactments in states that rely solely on institutions of representative government (Gamble, Reference Gamble1997; Lewis, Reference Lewis2013). Second, do public policies that result from initiatives and referenda correspond more with majority opinion compared with those enacted in states without direct democracy? However, no previous study of plebiscitary democracy has related it directly to the diffusion of policy innovations.

Nor has this research question been considered in the larger body of research on policy diffusion (Graham et al., Reference Graham, Shipan and Volden2012; Karch, Reference Karch2007; Shipan & Volden; Reference Shipan and Volden2012). Overviews of the policy diffusion scholarship rarely, if ever, make any mention of a role for direct democracy (Dolowitz & Marsh, Reference Dolowitz and Marsh1996; Graham et al., Reference Graham, Shipan and Volden2012; Karch, Reference Karch2007). One extensive literature review identified six actors that facilitate policy transfer among governments (Dolowitz & Marsh, Reference Dolowitz and Marsh1996, p. 345): elected officials, political parties, bureaucrats/civil servants, pressure groups, policy entrepreneurs/experts (see also Mintrom, Reference Mintrom1997), and supranational institutions. The primary focus of diffusion research, therefore, has been the institutional policymakers in government. As Pacheco (Reference Pacheco2012) rightly concluded, conventional policy diffusion models “focus largely on the decision making of state officials” (p. 187), not public opinion.

Yet 23 state constitutionsFootnote ¹ allow statutes or constitutional amendments to be initiated through voter petitions, and 14 of the 30 states with medical marijuana laws did so through popularly initiated referenda. Apart from the direct use of referenda to bypass the state legislature, moreover, there is evidence of indirect effects: the mere existence of popular initiative in a state may prompt legislators to act in order to preempt the prospect of direct democracy (Arceneaux, Reference Arceneaux2002; Gerber, Reference Gerber1996). This scenario occurred in Ohio. In 2016, Ohio legislators enacted a medicalized marijuana law after a citizen petition was initiated. The drive was organized by Ohioans for Medical Marijuana with backing from the Marijuana Policy Project. Once legislation was signed into law in June 2016 by Republican governor John Kasich, the petition drive was suspended.

In their study of medical marijuana policy diffusion, Hannah and Mallinson (Reference Hannah and Mallinson2018) included “initiative availability” as a dummy variable in their event history analysis to measure “states’ institutional capacity” insofar as “[t]he ability to use the direct initiative should increase the likelihood of a state passing a law” (p. 414). However, presumably the presence of direct democracy would facilitate policy adoption by legislative enactment, because Hannah and Mallinson do not operationalize the direct enactment of medicalized marijuana by citizen I&R. In other words, its positive statistical effect in their model is simply more evidence of an indirect effect of direct democracy on policy adoption by legislative enactment.

As originally formulated (Berry & Berry, Reference Berry and Berry1990, p. 400; see also Berry & Berry, Reference Berry, Berry, Sabatier and Weible2014), their theoretical framework posited the need to include variables that operationalize the “motivations” of political decision-makers to innovate, the “resources” that facilitate policy adoption, and the “obstacles” that sought to block those enactments.

For our purposes, the most important “resource” favoring policy adoption (Berry & Berry Reference Berry and Berry1990, p. 403) would be the positive experiences of other states, which are manifested as the direct effects of horizontal diffusion. As to why states would be influenced by the policy adoptions of other states, there are three prominent explanations in the policy diffusion literature (Karch, Reference Karch2007): emulation, competition, and imitation. Emulation means that policies diffuse because officials view a successful adoption elsewhere and want to copy it. Competition occurs when officials fear that failing to adopt a policy will put their state at a competitive disadvantage with other states. Imitation fuels policy diffusion because officials “believe that they share a policy-relevant characteristic with a jurisdiction that has already enacted it and, therefore, believe that they ought to enact it also” (Karch, Reference Karch2007, p. 59).

Most often, previous adoption by “neighboring” states is included (Mooney, Reference Mooney2001), but limiting the states to only those with a common border would necessitate omitting Alaska and Hawaii from this analysis (see Berry & Berry, Reference Berry and Berry1990). A more relevant consideration is that states with I&R are clustered across the Inter-Mountain and Far West, with none in the South or in the Northeast and Midwest corridors. The use of U.S. Census Bureau regions would include all I&R states in the same geographical area. The regional spread of a larger number of states also raises “the possibility of imitation based on similarities not necessarily found in the most proximate, neighboring states” (Karch, Reference Karch2007, p. 60). Partisan and ideological similarities would be examples (Grossback al., Reference Grossback, Nicholson-Crotty and Peterson2004; Volden, Reference Volden2006), but extending that analogy would suggest that activists in states with direct democracy would be encouraged to imitate policies adopted by activists in other states who employed direct democracy.

A clue to the importance of I&R diffusion through imitation is provided by Figure 1, which shows the cumulative frequency of adoptions of (1) all 30 medical marijuana laws, (2) the 14 medical marijuana laws adopted through referenda, and (3) 16 adoptions that occurred through statutory enactments. Figure 1 does not remotely resemble the conventional S-curve pattern for policy diffusion across states (Berry & Berry, Reference Berry, Berry, Sabatier and Weible2014, p. 315). In fact, the flattest pattern for referendum adoptions of medical marijuana laws is highly suggestive that factors intimately related to direct democracy had unique effects on the overall pattern of policy diffusion. For our purposes, most revealing about Figure 1 is that adoption of medical marijuana through citizen initiatives and referenda began much earlier than statutory adoptions, so we anticipate that previous episodes of adoption by direct democracy should precipitate later adoptions by that method and perhaps have an indirect effect on future statutory adoptions as well. Therefore, Hypothesis 1 posits that medicalized marijuana will be more rapidly diffused when previous states with direct democracy adopted this legislation compared with adoptions by legislation.

Figure 1.

Cumulative distribution of adoption of medical marijuana.

Berry and Berry (Reference Berry and Berry1990) also categorized unified party rule as a “resource” that facilitates policy diffusion, whereas divided-party government, they argued, would frustrate adoptions (p. 403). But the use of direct democracy instead of legislative enactment for policy adoptions begs the question of why marijuana activists might bypass the formal institutions of government. What aspect of the political opportunity structure would block legislative enactments? A partisan caveat must be added to Berry and Berry’s (Reference Berry and Berry1990) original formulation because research on other morality policies, notably abortion and gay rights, found that Democrats were more supportive than Republicans (Fleischmann & Moyer, Reference Fleischmann and Moyer2009; Rhodebeck, Reference Rhodebeck2015; Roh & Berry, Reference Roh and Berry2008; Roh & Haider-Markel, Reference Roh and Haider-Markel2003). Partisan bias may be less obvious with drug policy, but Meier (Reference Meier1992) also assumed that Democrats were less punitive than Republicans in their approach to the drug problem, though at that time party proved to have “little impact” on the pattern of marijuana laws (pp. 43, 53). The “war on drugs” was inaugurated by President Richard Nixon and reignited under the Ronald Reagan administration, though Democratic president Bill Clinton also waged a campaign against drugs (Ferraiolo, Reference Ferraiolo2007, pp. 161, 163; McWilliams, Reference McWilliams1991; Sharpe, Reference Sharpe1992). Thus, Hypothesis 2 posits that medicalized marijuana legislation is less likely to be adopted by legislative enactment in states during periods of unified Republican or divided-party rule. However, a corollary, as Hypothesis 2a posits, is that medicalized marijuana legislation is more likely to be adopted through citizen-initiated referenda in states during periods of unified Republican or divided-party rule. Footnote ²

More importantly, an essential variable that is missing from Hannah and Mallison’s (2018) study is that they did not specify how direct democracy would overcome a specific political obstacle to legislative adoption. It was for that reason that an “interaction” term is needed similar to what was employed by Schildkraut (Reference Schildkraut2001) to explain why states with larger populations of foreign born (whose opposition would pose an electoral threat to state legislative enactments) were nonetheless able to enact official English laws through citizen-initiated referenda (p. 448). Similarly, in our analysis, to more precisely specify the political conditions that would prompt the use of direct democracy to adopt medicalized marijuana, we need to include an interaction term between Republican rule or divided government (meaning partial Republican control) and the availability of citizen-initiated referenda.

Because Hypotheses 1 and 2 are the primary drivers of this analysis, six other “control” variables are included, though we do not hypothesize about their probable impact on adoption. Medicalized marijuana adoptions hold the potential for “vertical” diffusion effects given that state laws legalizing marijuana use are constitutionally suspect. Marijuana is classified as a Schedule I drug, like heroin and cocaine, under the Controlled Substances Act (CSA) of 1970, and the U.S. Supreme Court twice upheld federal authority to enforce the CSA in states with conflicting drug laws, in United States v. Oakland Cannabis Buyers’ Cooperative (2001) and Gonzales v. Raich (Reference Raich2005). To resolve this legal thicket, the Barack Obama administration announced in October 2009 that the U.S. government would not prosecute in states with medical marijuana laws (Stout & Moore, Reference Stout and Moore2009), and in August of 2013, it expanded that policy to include individuals who possess small amounts of marijuana for personal use in states which fully legalized marijuana (Dennis, Reference Dennis2013). The Donald Trump administration revoked both those policies in January 2018 (Savage & Healy, Reference Savage and Healy2018), but this decision came after the time frame (1996–2017) of this analysis.

Variables suggested by Berry and Berry (Reference Berry and Berry1990) that indicate “motivation” to adopt legislation would apply to political elites in states where medicalized marijuana was enacted legislatively and may be relevant to referendum states where citizen activists orchestrated petition campaigns. Virtually all the promotional literature by marijuana activists publicized their desire to provide relief for patients suffering from incurable diseases or the side effects from medical treatments (Ferraiolo, Reference Ferraiolo2007); for examples, see the supporting statements that accompanied the medical marijuana referenda held in California, Colorado, Massachusetts, Nevada, and Washington State (Wendell & Tatalovich, Reference Wendell and Tatalovich2020, Table 4F). Among those with afflictions who may benefit from medicalized marijuana, the largest potential constituency would be cancer patients. In 2000, 21 state medicalized marijuana laws specifically included cancer in the listed of approved indications for its therapeutic use (Pacula et al., Reference Pacula, Chriqui, Reichmann and Terry-McElrath2002, pp. 428–430).

An additional motivation would be the desire of public opinion, or constituencies within the polity, to legalize marijuana use. The combined results of three Gallup Polls showed higher support among self-described liberals (Carroll, Reference Carroll2005). Lax and Phillips (Reference Lax and Phillips2012) constructed a public opinion liberalism score to measure state-level attitudes on 39 policies, which confirmed that liberal policies correlated with liberal public opinion. On the specific policy of medicalized marijuana, all 26 states with such legislation had enacted “liberal” laws and every enactment corresponded with majority opinion in those states (Lax & Phillips, Reference Lax and Phillips2012, p. 154). A much earlier study by Meier (Reference Meier1992) similarly determined that liberalism, defined as the 1972 vote percentage for Democratic presidential candidate George McGovern, predicted “those states with lighter penalties for marijuana offenses” (p. 53). In addition to the broad preferences of public opinion, polling (Pew Research Center, Reference Center2015) suggests a relationship between individuals’ age and education and their attitudes toward marijuana.

The last category of variables used by Berry and Berry (Reference Berry and Berry1990) include those factors that act as “obstacles” to policy adoption. One predictable source of opposition would be evangelical Protestants, as there is a substantial body of research on state adoptions of other morality policies such as abortion (Meier & McFarlane, Reference Meier and McFarlane1993), same-sex marriage (Haider-Markel, Reference Haider-Markel2001), prohibition of alcohol (Frendreis & Tatalovich, Reference Frendreis and Tatalovich2010), and lotteries (Mooney, Reference Mooney2001), which include a variable for the presence of evangelical Protestants.