County-Level Differences in Support for Recreational Cannabis on the Ballot

Abstract

Cannabis is traversing an extraordinary journey from an illicit substance to a legal one, due in part to an unprecedented wave of bottom-up law reform through successful citizen ballot initiatives. Yet, even in states that have legalized recreational cannabis, there is substantial geographic variability in support of cannabis legalization. Geographic variability in voter support for cannabis legalization is impactful (e.g., county moratoriums/bans) yet poorly understood. This paper demonstrates the consequences of county-level population demographics, sociopolitical factors, and community differences in experience with criminalization of cannabis possession for understanding county-level variation in support of recreational cannabis law reform on (un)successful ballot measures in California (2010), Colorado (2012), Washington (2012), and Oregon (2014). OLS regression analyses of nearly 200 counties indicate that differences in racial and ethnic composition (% Black, Hispanic), political affiliation (% Republican), past criminalization, gender composition, and higher education level of residents all predict county-level variation in support for liberalization of cannabis law. Stronger Republican political leanings and/or higher percentages of Black or Hispanic residents were associated with reduced support, whereas higher education, male sex composition, and greater past criminalization were associated with increased support for cannabis legalization across counties. Religiosity and rurality were inconsequential as predictors of county-level voting patterns favoring recreational cannabis. The substantial geographic variability in voter support for cannabis legalization has significant implications for policy implementation and effectiveness.

Keywords

cannabis geography policy quantitative research social policy

Introduction

In the second decade of the 21st century, cannabis has traversed an extraordinary journey from an illicit substance to a legal one, due in part to an unprecedented wave of bottom-up reform. Movement toward legalizing recreational cannabis use has come from ballot initiatives spearheaded by citizens and passed by the states through popular vote. In 11 states (Alaska, California, Colorado, Illinois, Maine, Massachusetts, Michigan, Nevada, Oregon, Vermont, and Washington, as well as DC), recreational cannabis is now legal (C. J. Mosher & Akins, 2019); only Illinois and Vermont legalized cannabis via state legislative action. Those in government, industry, and academia are increasingly interested in the underlying processes driving these citizen-initiated changes, including the process involved in the passage of cannabis-related legislation.

The success of these ballot-based initiatives points to an increase in support for cannabis law reform, as do national surveys of public opinion. As of January 2018, 61% of Americans favored legalizing cannabis, a number that has steadily increased over the last decade (Pew, 2018). Desire for cannabis law reform has been spurred, in part, by changing opinions about the effects of criminalization on local communities and a demand to address racial inequality in enforcement of drug laws. This concern is also reflected in recent bills (re)introduced to U.S. Congress, such as the Marijuana Justice Act in January 2019. This bill would remove cannabis from the list of Schedule 1 drugs and expunge past convictions for possession (congress.gov, 2019). Discussion around this bill has focused on the disproportionate effect of criminalization on disadvantaged communities, but current research has only begun to investigate the effect of criminalization/desire for social justice on state-level votes about cannabis law reform (Collingwood et al., 2018).

Attempts at legalization have revealed a patchwork of public opinion, with slow (if any) change in receptivity to cannabis in certain places, revealing within-state divisions underlying American society. Even within a state like Washington, for example, where cannabis was legalized by 56% of residents voting in support, the majority of voters in 19 of 39 counties voted against the measure (C. J. Mosher & Akins, 2019). Focusing on state-level patterns of variation in the legal status of cannabis covers up marked polarization in opinion at local levels, which is made more visible in the context of ballot initiatives where individuals are asked to vote for law change, effectively translating their opinions about legalization to actions in voting behavior. Public opposition in some areas remains strong, even after legalization, translating into local moratoriums on voter approved cannabis law reform (Dilley et al., 2017).

County-level voting results on social-issue ballot measures provide a unique window into how public opinion changes. While individual opinions on these issues are often addressed using survey methods, patterns identified at the individual level do not always translate to aggregate-level voting behavior, necessitating studies of predictors at the aggregate level to truly understand voting patterns and policy change. As cannabis use is normalized and legalized, public opinion is changing at different rates in different geographic areas and among demographic groups. To some extent, variation in demographic composition of counties (alone) might produce variation across places in aggregate opinion and, consequently, voting behavior.

This study is among the first to assess sources of variation in county-level support for cannabis legalization. We start with demographic/compositional effects derived from important correlates identified in individual-level public opinion research. Local demographics and other community-level sociocultural dimensions remain to be explored as significant sources of county-level variation in support for legalization. Second, this study examines how a community’s past experience with criminalization of cannabis use has shaped local voter preference on ballot initiatives to legalize, above and beyond community compositional factors.

In this article, we use individual-level correlates identified in public opinion/survey research to inform aggregate level analyses of variation in county-level voting patterns based on demographic or compositional effects, and we employ arrest data on cannabis possession from the Uniform Crime Reports to quantify past criminalization of cannabis use. Votes were analyzed from four recreational initiatives between 2010 and 2014, including three successes and one failure. Using county-level census and voting data, this article aims to answer the questions: To what extent do demographic-compositional factors and sociopolitical population characteristics account for county-level variability in voter support for recreational cannabis legalization? To what extent is varied community experience with the criminalization of cannabis use related to county-level differences in voter support for recreational cannabis legalization?

Literature Review

Recreational cannabis legislation makes cannabis legal, within limits, for any adult to purchase and consume. More than 30 states and DC have legalized cannabis for medical use, but 14 additional states (a total of 45 states) allow at least the prescription and sale of medicines made from CBD-rich cannabis.¹ That leaves only five U.S. states (Idaho, Kansas, Nebraska, South Dakota, and Texas) that have not legalized the use of cannabis in any form. In January 2018, Vermont became the first state to legalize recreational cannabis through the legislature, although like DC, recreational sales were not allowed (Zezima, 2018). In early 2020, the Vermont House approved legislation that will allow the establishment of retail cannabis outlets in the state by 2022 (Landen, 2020). Figure 1 shows that reform is moving in from the coasts, with support concentrated on the west coast and in the northeast, and opposition holding in the Midwest and the South. What’s more, these changes are occurring rather quickly following decades of steady opposition for recreational legalization across most of the country.

Figure 1.

Legal status of cannabis by state, February, 2018. Source: Norml State Info (www.norml.org/states)

While change in aggregate opinion by state is happening quite quickly, that change is not universal within states. Even in what are considered the most pro-cannabis states, there are counties, cities, and towns that maintain staunch opposition. Some of these places have prohibited law implementation, disallowing recreational cannabis sales and effectively creating “cannabis deserts” within legal states (Dilley et al., 2017). This divide also appears in the distributions of votes for and against recreational cannabis. Table 1 details the range of variation across counties in the four votes included in the current analysis.

Table 1.

Ballot Measures in Recreational Sample, Percentage of County Voting Yes.

State	Year	Outcome	% Yes	Lowest Rate of Support		Highest Rate of Support
California	2010	Defeated	47	32%	Colusa	90%	Los Angeles
Colorado	2012	Approved	55	32%	Kiowa	79%	San Miguel
Oregon	2014	Approved	56	31%	Malheur	71%	Multnomah
Washington	2012	Approved	56	38%	Garfield	68%	San Juan
Total	—	—	54	31%	Malheur, OR	90%	Los Angeles, CA

The divide can also be perceived in the bans and moratoriums on implementing cannabis sales that happen after legislation is passed. For example, while Washington and Colorado both passed legislation in 2012, bans or moratoriums currently exist in 43 (67%) of Colorado’s 64 counties (Mitchell, 2018), whereas 9 (23%) of Washington counties still have bans or moratoriums in place (Municipal Research and Services Center, 2018). As is evident from Figure 2, even 5 years later, places with lower rates of support are more likely to have bans or moratoriums in place. More bans and moratoriums also exist at the city level, even in counties with high rates of support (Dilley et al., 2017). Bans and moratoriums delay legalization and impact the ability of the Washington State Liquor and Cannabis Board to achieve their goal of supplanting remaining black market distribution channels. Bans also impact the amount of tax revenue received by city and county jurisdictions. Clearly, it is important to understand the extent and nature of within-state variation in support for legal cannabis on the ballot.

Figure 2.

Bans and moratoriums in Washington, 2018. Source: Municipal Research and Services Center (2018). Note. Stars denote counties with bans or moratoriums or that have taken no action toward regulation.

Prior Research

Understanding what factors influence attitudes about and votes for cannabis legalization is important. The shift in recent years toward reforming drug policy through ballot measures makes both public opinion and patterns in voting about cannabis central to determining drug policy outcomes. One popular explanation of voting trends on cannabis law reform focuses on self-interest. Personal experience using cannabis remains one of the most significant predictors of support for recreational legalization. Some 70% of people who have tried cannabis support legalizing it (Pew, 2013). A full 89% of people who have used cannabis within the past year support recreational legalization. Conversely, some 74% of people who have never tried cannabis say they would feel uncomfortable around cannabis users, whereas only 25% of those who have tried it would say the same (Pew, 2013). While self-interest likely plays an important role in determining rates of support for recreational cannabis legalization, no county-level data exist to allow tests of the effect of self-interest on votes. However, analyses of recent polls show that cannabis use and attitudes vary widely across demographic groups and that many more people than only cannabis users support legalization. Self-interest is only a partial explanation of attitudes; there are people who support legalization that likely will not use cannabis even if the substance is legalized (McCarthy, 2015; Pew, 2013, 2014).

To better understand bottom-up reform, and to investigate factors beyond self-interest, Leon and Weitzer (2014) contextualize cannabis legalization as a case of campaigns to legalize vices. Such campaigns (e.g., to decriminalize prostitution, gambling, drugs) have been shown to be highly susceptible to emotional appeals, moral panic claims (in this case impact on youths), and social justice and civil rights claims. They identify a number of conditions that increase the likelihood of vice legalization. These include the following: (1) significant numbers of people engage in the vice, with many of them having “conventional lifestyles”; (2) belief that legalization is less harmful than criminalization; (3) production and distribution can be controlled by government authorities; (4) decriminalization or legalization is supported by law enforcement, politicians/legislators, and/or business leaders; and (5) potential for significant revenues for governments (Leon & Weitzer, 2014, p. 197). Other important factors include the status of proponents and opponents, the extent and content of media coverage and editorials regarding the issue, national versus midterm elections, threat of government intervention, and campaign spending. More specific to cannabis legalization measures, Leon and Weitzer also note that the age distribution of voters in a particular election can affect the outcome of legalization measures, given that younger voters are more likely to be in support.

Socia and Brown (2014) find that demographic characteristics predict voter support for medical cannabis. They also link this variation in support to moratorium passage after the law was passed. Specifically, “race, ethnicity, education, age, and political affiliation may all influence support for medical marijuana” (Socia & Brown, 2014). Prior work has provided a plethora of information about who wants to legalize, but individual-level opinion polls and surveys may or may not predict aggregate-level voting trends.

Support for recreational cannabis varies significantly across sociodemographic groups. Most of our knowledge about these patterns comes from nationally representative polls, which have monitored public opinion about cannabis for decades. In short, supporters are more likely to be young, nonreligious, White, liberal, or Democratic and are more educated, have higher incomes, and have used cannabis. Opponents tend to be over age 65, highly religious, Hispanic, conservative, or Republican, have less education and lower incomes, and have not used cannabis (Pew, 2018). These demographic effects on opinions have been observed at city, county, and state levels.

Recent research has also revealed an additional geographic component in votes on cannabis law reform. Prior research has shown that voters rarely have encyclopedic information on ballot measures, but use other cues to help decide their votes, such as elite endorsements, economic indicators, perceived costs and benefits, and racial and ethnic context. Beyond the influence of demographics and social climate factors believed to influence votes on many social issue ballot measures, including those involving cannabis, local levels of criminalization are thought to impact opinions and votes. Having had encounters with the criminal justice system has been shown to predict higher rates of support for cannabis law reform, demonstrating that cannabis legalization is being driven not only by “marijuana enthusiasts” but also by those who are concerned about social justice and the negative impacts of the war on drugs on disadvantaged communities (Collingwood et al., 2018). Additionally, a race-specific study found that Blacks living in cities with the highest levels of black drug arrest rates have the highest rates of support for cannabis law reform (Thornhill, 2011).

Hypotheses

Our first four hypotheses focus on demographic characteristics known to influence opinion about cannabis policy: age, gender, race, and ethnicity. The age gap in attitudes about cannabis, where seniors are more negative, is persistent and significant (Alwin, 1998; Hall & Schiefelbein, 2011; Leon & Weitzer, 2014; Socia & Brown, 2014; Toch & Maguire, 2014). Only 35% of the silent generation (currently aged 70–90) support legalization, whereas 70% of millennials say the same (McCarthy, 2015; Motel, 2015; Pew, 2013, 2014). In terms of gender differences in attitudes, women are less likely to have tried cannabis or to support recreational cannabis legalization (Grella et al., 2015).

Differences by race and ethnicity in attitudes about and use of cannabis are complex. Historically, Blacks have been less likely than Whites to support cannabis legalization (Meares, 1997). Yet newer studies have found Blacks, and non-Whites more generally, to be less punitive than Whites, due to beliefs about racial bias in the criminal justice system (Bobo & Johnson, 2004; Toch & Maguire, 2014). Hispanics are less likely than any other ethnic group to support recreational cannabis (Grella et al., 2015). And counties with more Black or Hispanic residents had lower rates of support for medical cannabis in 2012 (Socia & Brown, 2014). Accordingly, we hypothesize regarding voter support for recreational cannabis legalization, that:

Hypothesis 1: Counties with a higher proportion of older residents will have lower rates of support.

Hypothesis 2: Counties with a higher proportion of male residents will have higher rates of support.

Hypothesis 3: Counties with a higher proportion of Black residents will have lower rates of support.

Hypothesis 4: Counties with a higher proportion of Hispanic residents will have lower rates of support.

Our second set of hypotheses focus on sociocultural effects: local levels of education, religiosity, political climate, and rurality. Polls show that lower levels of education and income are generally associated with negative attitudes about cannabis (Grella et al., 2015; Hall & Schiefelbein, 2011; Leon & Weitzer, 2014; Socia & Brown, 2014). The more often someone attends church, the less likely that person is to report using cannabis or support legalizing it (McCarthy, 2015). Political ideology and political party affiliation are also strong predictors of attitudes about cannabis, with conservative Republicans least likely to support legalization of cannabis (Newport, 2011; Pew, 2014; Saad, 2014). In addition, more urbanized states have been more likely to pass medical cannabis laws, an effect we also expect to see with recreational measures (Hall & Schiefelbein, 2011). Based on extant individual-level findings, at the aggregate level for recreational cannabis legalization, we expect that:

Hypothesis 5: Counties with more residents with a bachelor’s degree will have higher rates of support.

Hypothesis 6: Counties with a higher proportion of religious adherents will have lower rates of support.

Hypothesis 7: Counties with a higher proportion of Republican voters will have lower rates of support.

Hypothesis 8: Counties with a higher proportion of rural residents will have lower rates of support.

Our final hypothesis addresses criminalization of cannabis use. Simkins and Geiger-Oneto (2015) found that perceived likelihood of decreased crime was a significant predictor of support for recreational cannabis. In addition, because legalization campaigns (particularly in Washington State) have emphasized social/racial justice issues (C. J. Mosher & Akins, 2019), we expect that counties with more cannabis possession arrests will have higher rates of support for legalization. Diverse local experiences with cannabis criminalization, resulting from geographic differences in enforcement of drug laws, likely led to differences in the resonance of social justice framing, making it more impactful in some places than others. We posit that for cannabis legalization:

Hypothesis 9: Counties with more cannabis possession arrests will have higher rates of support.

The present study fills gaps in the literature on support for cannabis legalization. Analyzing votes, as this article does, is advantageous because until attitudes are expressed through votes, they do not affect policy outcomes or access broader impacts associated with state-level legalization of recreational cannabis. Polls show that public opinion about cannabis is split across class, gender, age, religious, racial/ethnic, and political lines and that self-interest is an important predictor. This article goes beyond polling to assess within-state differences in voting patterns, which are not well understood. This research identifies whether and to what extent geographic clustering of like individuals is associated with county-to-county differences in preferences for drug policy. Such geographic clustering has important policy implications given moratoriums and the resulting unequal access to dispensaries, tax revenues, and employment in the cannabis industry.

Data and Method

To assess which county-level characteristics predict votes about cannabis, the nine hypotheses detailed above were generated based on the findings of prior research. The hypothesized relationships were tested through a series of ordinary least squares (OLS) regressions.

Data

To test these hypotheses, county-level votes on four recreational ballot measures were analyzed for: California 2010, Colorado 2012, Washington 2012, and Oregon 2014. Although data were available for earlier recreational ballot measures (eight failures between 1972 and 2006), the current analysis was limited to votes between 2010 and 2014 to control for period effects, particularly due to differences by presidential administration. Data were collected for all votes on recreational cannabis between 2010 and 2014 where county-level results were available, and one vote from each state was selected for analysis. Alaska’s 2014 vote was excluded due to the differences in data between counties and boroughs. All four included states had previously passed medical legislation at the time of their vote. Three of the four recreational measures passed, with support ranging from 47% to 56% voting yes. California’s 2010 ballot measure (analyzed here) was defeated, but a similar measure was later passed in 2016. Data for independent variables were taken from (1) David Leip’s Atlas of U.S. Presidential Elections, (2) demographic data collected in the 2010 national census, (3) Brown University’s Longitudinal Tract Data Base, (4) The Association of Religion Data Archives’ Churches and Church Membership in the U.S. data sets, and (5) Uniform Crime Reports from the Inter-University Consortium for Political and Social Research (ICPSR) for the years 2004–2012.

Recreational Ballot Measures

California 2010

In November 2010, California Proposition 19, the Marijuana Legalization Initiative to legalize various cannabis-related activities, was defeated on the ballot, with only 47% support. Opponents were concerned about backlash from the federal government, lack of a statewide standard, and some argued that allowing recreational cannabis would exacerbate abuse of the extant medical marijuana system. Governor Schwarzenegger decriminalized possession of small amounts of cannabis in mid-2010 (Leon & Weitzer, 2014).

Colorado 2012

The Colorado Marijuana Legalization Initiative, Amendment 64, legalizing use and possession of cannabis and allowing regulation and taxation of retail sales, was approved in November 2012 by 55% of state voters. Beginning on January 1, 2014, adults in Colorado also could grow a small number of cannabis plants (out of public view) for personal use. In Colorado, the overarching frame was that cannabis is less harmful than alcohol and should be regulated as such (Leon & Weitzer, 2014).

Oregon 2014

In November 2014, The Oregon Legalized Marijuana Initiative, Measure 91, was approved with 56% voting yes. The initiative legalized cannabis for people 21 and older, allowing adults to possess up to 8 ounces of dried cannabis and four plants. It also regulated sales through dispensaries. Recreational sales of cannabis in Oregon began on October 1, 2015.

Washington 2012

In November 2012, the Washington Marijuana Legalization and Regulation, Initiative 502, regulating legal production, possession, delivery, and distribution of cannabis and sales of small amounts to people 21 and older, was approved with 56% voting in favor. In Washington, a major emphasis in the pro-legalization campaign was how legalization would promote social justice by removing cannabis cases from the criminal justice system and reducing racially disparate outcomes. The Washington medical cannabis movement opposed the measure, not to protect their profits, but because they saw it as being too restrictive.

Variables

The dependent variable in this study was county-level support for cannabis legalization. Support was operationalized as percentage of county voting yes to legalize recreational cannabis. County-level election results for each ballot measure were collected from David Leip’s Atlas when available and from Secretary of State websites when unavailable.

Gender composition was operationalized as percentage of male residents in each county. Age composition of the county was operationalized as percentage over age 65. Decennial census data for these variables were obtained from American Fact Finder for 2010. Racial composition was operationalized as percentage non-Hispanic Black residents. Ethnicity was operationalized as percentage Hispanic or Latino residents. Census data for these county variables were obtained from Brown University’s Longitudinal Tract Data Base, for 2010.

Education was operationalized as percentage of residents 25 years and over with bachelor’s degrees in each county. Data for this variable came from American Community Survey 5-year estimates. Religiosity was operationalized as percentage religious adherents in each county. Data for this variable were obtained from the American Religious Data Archive for 2010; adherents include standardized counts of members and other participants, according to a complete census of religious congregations/denominations (Jones, 2002). Political party affiliation was operationalized as percentage of the county voting for the Republican candidate in the most recent presidential election (e.g., for cannabis votes in 2010 and 2012, this variable indicates percentage voting for Republican John McCain in 2008).

Rurality was operationalized using county-level rural–urban continuum codes for 2013 from the U.S. Department of Agriculture. The continuum ranges from 1 (Metro—counties in metro areas of 1 million population or more) to 9 (Nonmetro—completely rural or less than 2,500 urban population, not adjacent to a metro area). Criminalization of cannabis was operationalized as rate of arrest for cannabis possession per 1,000 residents. Uniform Crime Report data for this variable were obtained through ICPSR for 2006–2012, to calculate 3-year lagged averages of arrest rates per 1,000 residents.

A series of OLS regressions were performed to test the above-stated hypotheses. Only one vote per state was included, so each county represents an independent observation, though analyses were still subject to clustering of counties within states.² Variables were log transformed as necessary to reduce skewed distributions and to ensure linear relationships between each independent variable and percentage voting yes for cannabis law reform.³ The distributions of total population, percentage Black, and cannabis possession arrest rates were highly skewed, so these variables were log transformed. First, a correlation matrix was generated. Bivariate correlations were assessed to detect potential multicollinearity between independent variables.⁴ Then, after detecting no issues, three regression equations were estimated: Model 1 includes only the hypothesized demographic effects (age, sex, race, ethnicity). Model 2 adds sociocultural variables (political party distribution, religiosity, rurality, and levels of education). The full model presented in Model 3 includes a variable representing criminalization via arrests for cannabis possession. All three models include dummy variables for state (not shown), using California as the reference category. There is moderate evidence of heteroskedasticity in Model 3, but the χ² statistic is small, and errors are normally distributed.

Findings

Table 2 presents results of the regression of independent variables on percentage of a county voting yes for recreational cannabis. The first model includes county demographic-compositional variables, the second model adds sociopolitical factors, and the final model includes past criminalization via cannabis possession arrest rates of a county. (State dummy variables are included but not shown.) Model 1, assessing compositional effects of county-level demographics on support for recreational cannabis, demonstrates that counties comprised of higher percentages of older (aged 65+), Black, or Hispanic residents were significantly less apt to support recreational cannabis legalization. For example, a 10% difference across counties in share of older residents was associated with 6% fewer voting in support of legalization; every 10% difference in share of Hispanic residents was associated with 2% fewer residents voting in support. Demographic characteristics alone, though shown to have significant effects, explain only about 20% of observed county variation in votes. In addition, Model 1 shows evidence of omitted variables.

Table 2.

Ordinary Least Squares (OLS) Regression Coefficients: Percentage Voting Yes for Recreational Cannabis.

Variables	(1)	(2)	(3)
Variables	Demographics	Sociopolitical	Criminalization
Percentage over age 65	−0.636*** (0.00)	−0.001 (0.99)	0.023 (0.79)
Percentage male	−0.249 (0.15)	0.451*** (0.00)	0.466*** (0.00)
Percentage Black (ln)	−0.054** (0.47)	−1.360*** (0.00)	−1.217*** (0.00)
Percentage Hispanic	−0.188*** (0.00)	−0.117*** (0.00)	−0.113*** (0.00)
Percentage college graduates		0.073* (0.09)	0.079* (0.07)
Percentage religious		−0.026 (0.09)	−0.025 (0.23)
Percentage Republican		−0.574*** (0.00)	−0.569*** (0.00)
Rurality		−0.084 (0.61)	−0.084 (0.60)
Arrests for possession (ln)			0.624* (0.07)
Constant	71.239*** (0.00)	51.285*** (0.00)	49.481*** (0.00)
Observations	194	194	194
Adjusted R ²	.191	.859	.860

Note. All models include dummy variables for state with California as reference category (not shown). p values in parentheses (one-tailed tests).

***p < .01. **p < .05. *p < .10.

Model 2 accounts for effects of sociopolitical characteristics of counties on voting in support of cannabis legalization—educational attainment, religiosity, political orientation, and rurality. Neither rurality nor religiosity within a county yielded significant effects on voting patterns. However, counties with a higher percentage of college graduates evinced somewhat greater support for cannabis legalization. And exerting a strong and significant effect, percentage Republican in a county was associated with much reduced voter support for recreational cannabis. Counties comprised of 10% more Republicans are expected to evince nearly 6% lower rates of voter support for recreational cannabis. Notably, a county’s political leaning fully accounted for any age-related demographic effects on county voting patterns, rendering the coefficient for percentage over age 65 nonsignificant. (Racial and ethnic composition effects persisted, independent of county politics.) The addition of sociocultural variables in Model 2 dramatically increases the adjusted R ² from .19 to .86. Further inspection of sociocultural factors revealed that a county’s political leaning accounted for much of the increase in county-to-county variance explained.

In the full Model 3, sociodemographic characteristics (race, ethnicity, gender composition) and sociocultural variables—namely political climate, but also educational attainment—continued to exert significant effects on voter support for legalization. Importantly, past criminalization, measured via county arrest rates for cannabis possession, was associated with significantly greater levels of support for cannabis legalization, but effect strength was rather weak. The addition of past criminalization adds little to the amount of variance explained. Model 3 shows no evidence of omitted variables or specification error. Based on coefficients for the independent variables in the full model predicting percentage voting yes for recreational cannabis (Table 2, Model 3), and holding all else equal, counties with 10% more male residents average 5% higher rates of support, more Black or Hispanic residents average 1% lower support, more college graduates average 1% higher support, more Republicans average 6% lower rates of support, and more cannabis possession arrests average under 1% higher rates of support for recreational cannabis legalization.

In short, the regression models presented in Table 2 indicate that counties with more college graduates, more arrests for cannabis possession, and more male residents were likely to have higher rates of support for recreational cannabis. Counties with more voters that are Republican, more Black residents, and more Hispanic residents were likely to have lower rates of support, all else being equal. Somewhat surprisingly, religiosity and rurality did not exert significant effects. Higher cannabis-related arrest rates in the past were related to increased voter support for legalization. Although many of the effect sizes seem small, the mean percentage voting yes was 49%, so even small effects have the potential to change resulting policy outcomes.

Discussion

Based on county-level voting records collected for each state, there is substantial geographic variability in support for cannabis legalization in states that considered such a measure on the ballot (see also Dilley et al., 2017). Descriptive statistics on the nature and extent of variability in county-to-county voting patterns on recreational cannabis legalization demonstrated significant differences within each of the four states voting on legalization between 2010 and 2014. At least 44% of voters in each of the four states analyzed voted against legalization, and even within states that legalized, some counties evinced low levels of support. In our data, nearly 70% of Malheur County, OR, voted against legalization, compared to 44% statewide, for example. What’s more, these opponents were not randomly distributed throughout states.

This article also sought to identify county-level characteristics that predicted support for cannabis law reform. This research demonstrated that county-level distributions of gender composition, racial and ethnic composition, higher education, political affiliation, and criminalization all predicted variation in support for recreational cannabis law reform on ballot measures, contributing to the patchwork patterning. Although demographic composition of a county is an important factor in predicting recreational cannabis voting outcomes, county demographics accounted for only about one fifth of observed variation to be explained. In terms of sociocultural effects, however, the political leaning of a county evinced one of the strongest aggregate-level effects on voter support for legalization and accounted for a substantial share of variation to be explained. Also exerting a small but significant effect, counties with more arrests per capita for cannabis possession prior to legalization tended to have higher rates of support; these criminalization effects on county-level voting patterns persisted above and beyond demographic compositional and sociocultural factors, suggesting that further study is merited.

The regression analyses showed that the spatial distribution of social and demographic compositional characteristics is related to geographic variability in voter support for legalization. To the extent that people continue to sort into increasingly similar “silos” in which to live, we can expect greater unevenness in policy implementation within states. The patterning of this variation has significant policy implications. Provisions in Washington, Oregon, and Colorado allow local jurisdictions to ban growing or selling of cannabis, which many of the counties with low rates of support have done. This in turn creates a patchwork of implementation, unevenly distributing access to cannabis and the economic benefits of the cannabis industry within states.⁵

For example, in Washington State, the majority of voters in 19 of the 39 counties voted against the cannabis legalization measure (Dilley et al., 2017), and several jurisdictions in the state prohibited retail sales of cannabis. The second (Pierce) and fifth (Clark) largest counties initially prohibited retail sales of cannabis, although both counties eventually overturned their bans (C. J. Mosher & Akins, 2019). In Oregon, Measure 91 included provisions to allow cities and counties to prohibit marijuana production and sales within their jurisdictions (Berg, 2017). As of early 2016, more than 100 cities had prohibited retail sales of cannabis, with most of these jurisdictions in the eastern part of the state (Selsky, 2016). In California, while large jurisdictions such as Los Angeles, San Francisco, San Jose, and San Diego have laws permitting cannabis businesses (Fuller, 2019), as of 2019, only 161 of the state’s 482 municipalities and 24 of its 58 counties allowed commercial marijuana activity of any type (Schroyer & McVey, 2019). Implications of such uneven implementation warrant further study.

Limitations and Future Directions

Like all studies, ours is not without limitations. Votes about cannabis are influenced by larger political, social, and economic forces, and these effects are likely to have changed over time and to differ across geographic contexts. Votes about recreational cannabis occur at the state level, so each county-level vote is influenced by the characteristics of that state, its history, and the specific framing and provisions of the measure appearing on their ballot (Bestrashniy & Winters, 2015; Hall & Schiefelbein, 2011; Leon & Weitzer, 2014; Pew, 2013). The framing of ballot measures has important impacts on rates of support across the state (Long, 2014). A full review of the literature on public opinion and framing is outside the scope of this article, but Weakliem (2005) reviews the current state and history of public opinion research. Of particular relevance to public opinion about cannabis is the discussion of framing and ideology, including the civil rights frame discussed by Zaller (1992), which could be invoked by recreational ballot measures. Future research could further explore justice-related variables such as the relationship between racial and ethnic disparities in drug arrests and support for legalization. Other state-level factors of importance include status of leading advocates and opponents, media coverage, timing (presidential elections, youth turnout), and threat of federal government intervention if passed (Leon & Weitzer, 2014).

Analyses showed that the political divide is the main driver of geographic difference in voting patterns. So, to the extent there are political divides (and activists cannot find themes that resonate with conservative voters), we can perhaps expect a continued patchwork of implementation, with more cannabis deserts within “legal states.” The ecological fallacy limits the conclusions that can be drawn from county-level data. While it would be interesting to know whether, for example, Democrats in primarily Republican counties were more apt to vote against legalization due to contextual effects, or whether voting occurred along party lines, that was not the intent of this study on how county-level demographics and social correlates were related to aggregate-level voting outcomes.

The inability to control for rates of cannabis use is another limitation of this study. There were no such data available at the county level to measure self-interest related to personal cannabis use. Although no statistical evidence of omitted variables was found, poll results indicate that in addition to the independent variables analyzed here, intent to use cannabis if the proposed legislation passes would also be a significant predictor of variation in public opinion. Such methodological issues have no simple solutions using currently available data. To assess predictors of votes about cannabis more accurately, effects would need to be modeled at the individual, county, and state levels. This strategy would be made possible only if individuals in multiple states were surveyed and data were collected on demographics, attitudes about recreational cannabis, and voting patterns. Without access to such data and advanced analytic techniques, these limitations of the present analysis must be acknowledged, but accepted.⁶

Finally, there is some evidence that community differences in experience with criminalization of cannabis use were related to voter support for legalization. The effect demonstrated here is small but significant. Future studies should continue unraveling the relationships between criminalization, race and ethnicity, desire for social justice, rates of use, and support for legalization. This analysis was also limited in its ability to differentiate between the effects of race (or ethnicity) and the effects of racial prejudice. Rates of use and attitudes about cannabis vary only a little by race, but enforcement of cannabis laws is highly racially biased (American Civil Liberties Union, 2013). Stereotypes in the media affect attitudes about race and crime, but the effect interacts with local neighborhood context; Whites living in homogenous neighborhoods respond to stereotypes by endorsing more punitive policies to address crime and expressing more negative attitudes toward Blacks. In contrast, Whites living in heterogeneous neighborhoods are either unaffected by exposure to stereotypes or move in the opposite direction, supporting less punitive policies and expressing more positive attitudes toward Blacks (Gilliam et al., 2002). This means that racial prejudices may influence public opinion on cannabis legalization. In short, county-level measures of racial composition cannot differentiate between the effects of differences in support by race and the effects of racial prejudice. Counties with more Black residents had lower rates of support, but we cannot know whether this was due to (a) Blacks voting no or to (b) White voters, threatened by the proportional size of minority populations, voting no due to racial prejudice. Future studies, especially those that model effects at multiple levels and over time, will be able to better assess these issues.

Conclusion

Attempts to change cannabis laws are taking the nation by storm. These reforms, when successful, have important impacts on criminal justice and mass incarceration, as well as providing states with much needed tax revenue. They may also have unintended (negative) consequences, and people have justifiable concerns about what the long-term effects of reform may be. Even when states, as a whole, vote to legalize cannabis, there are pockets of resistance or varying spatial geographies in which many residents do not support liberalization of cannabis policy. This article identified several county-level characteristics associated with variation in support for cannabis law reform, finding differences in how demographic and sociocultural factors influence support for recreational cannabis ballot measures.

Few prior studies analyzed votes rather than opinions, and there is a gap in current research regarding geographic differences in support for the legalization of recreational cannabis. The current research also addressed the call to investigate how community differences in criminalization of cannabis use was related to legalization voting outcomes. The results of previous empirical articles and national public opinion polls were detailed and used to generate county-level hypotheses. Possession arrest data were used to ascertain the extent of criminalization. These hypotheses were tested on a sample of county-level votes on recreational ballot measures between 2010 and 2014, using a series of OLS regressions. The demonstrated aggregate effects on voting patterns supported expectations based on individual-level survey data, clarifying the relationship between community demographic and social characteristics, and differences in drug policy preferences between counties. As well, the results support that communities affected by mass criminalization were more prone to support legalization, net controls for demographic and sociopolitical characteristics.

Stronger Republican political leanings and higher percentages of Black and/or Hispanic residents were associated with reduced county-level support, whereas higher education, male composition, and greater past criminalization were associated with increased support for cannabis legalization across counties. These factors explained a large share of county-to-county variation within states voting (not) to legalize cannabis.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Jennifer Schwartz

Notes

References

Alwin

D. F.

(1998). The political impact of the baby boom: Are there persistent generational differences in political beliefs and behavior? Generations, 22(1), 46–54.

American Civil Liberties Union. (2013). The war on Marijuana in black and white. https://www.aclu.org/sites/default/files/field_document/1114413-mj-report-rfs-rel1.pdf

Berg

(2017). Legal weed: Idaho consumers breathe life into tiny Oregon town, irk Otter. Idaho Statesman. https://www.idahostatesman.com/news/local/article136552113.html

Bestrashniy

Winters

K. C.

(2015). Variability in medical Marijuana laws in the United States. Psychology of Addictive Behaviors, 29(3), 639–642. https://dx-doi-org.web.bisu.edu.cn/10.1037/adb0000111

Bobo

L. D.

Johnson

(2004). A taste for punishment: Black and White Americans views on the death penalty and the war on drugs. Du Bois Review, 1(1), 151–180.

Collingwood

O’Brien

B. G.

Dreier

(2018). Evaluating ballot initiative support for legalised marijuana: The case of Washington. International Journal of Drug Policy, 56, 6–20.

Congress.gov. (2019). H.R.4815 - marijuana justice act of 2018. Congressional Research Service.

Dilley

J. A.

Hitchcock

McGroder

Greto

L. A.

Richardson

S. M.

(2017). Community-level policy responses to state Marijuana legalization in Washington State. International Journal of Drug Policy, 42, 102–108.

Economist. (2018). Why America still has “dry” counties. https://www.economist.com/the-economist-explains/2018/06/05/why-america-still-has-dry-counties

10.

Fuller

(2019). ‘Getting worse, not better’: Illegal pot market booming in California despite legalization. New York Times. https://www.nytimes.com/2019/04/27/us/marijuana-california-legalization.html

11.

Gilliam

F. D.

Valentino

N. A.

Beckmann

M. N.

(2002). Where you live and what you watch: The impact of racial proximity and local television news on attitudes about race and crime. Political Research Quarterly, 55(4), 755–780.

12.

Grella

C. E.

Cochran

Mays

(2015). Does health status influence attitudes about and use of medical marijuana? Findings from a general population survey in California. Drug and Alcohol Dependence, 146, 130–131.

13.

Hall

Schiefelbein

(2011). The political economy of medical Marijuana laws. Atlantic Economic Journal, 39(2), 197–198. https://doi.org/10.1007/s11293-011-9266-2

14.

Hennessy-Fiske

(2012). Alcohol sales fuel spirited debates in East Texas. Los Angeles Times. https://www.latimes.com/world/la-xpm-2012-feb-23-la-na-texas-alcohol-20120223-story.html

15.

Jones

D. E.

(2002). Religious congregations & membership in the United States 2000: An enumeration by region, state and county based on data reported for 149 religious bodies (Vol. 1). Glenmary Research Center.

16.

Landen

(2020). Vermont house backs legal market for marijuana sales. VTDigger. https://vtdigger.org/2020/02/26/vermont-house-backs-legal-market-for-marijuana-sales/

17.

Leon

Weitzer

(2014). Legalizing recreational Marijuana: Comparing ballot outcomes in four states. Journal of Qualitative Criminal Justice and Criminology, 2(2), 193–218.

18.

Long

(2014). Jokers, smokers and midnight tokers? How television news framed pro-pot legislation in Oregon and Colorado before the November 6, 2012 Elections [Thesis]. University of Missouri-Columbia. https://hdl.handle.net/10355/44303

19.

McCarthy

(2015). More than four in 10 Americans say they have tried Marijuana. http://www.gallup.com/poll/184298/four-americans-say-tried-marijuana.aspx

20.

Meares

T. L.

(1997). Charting race and class differences in attitudes towards drug legalization and law enforcement: Lessons for federal criminal law. Buffalo Criminal Law Review, 1(1), 137–174.

21.

Mitchell

(2018). Which Colorado counties allow Cannabis sales? https://www.westword.com/marijuana/colorado-counties-allowing-marijuana-sales-10046595

22.

Mosher

Akins

(2014). Drugs and drug policy: The Control of consciousness alteration. Sage.

23.

Mosher

C. J.

Akins

(2019). In the weeds: Demonization, legalization, and the evolution of us marijuana policy. Temple University Press.

24.

Motel

(2015). 6 facts about Marijuana. http://www.pewresearch.org/fact-tank/2015/04/14/6-facts-about-marijuana/

25.

Municipal Research and Services Center. (2018). Marijuana regulation in Washington State. http://mrsc.org/getdoc/8cd49386-c1bb-46f9-a3c8-2f462dcb576b/Marijuana-Regulation-in-Washington-State.aspx

26.

Newport

(2011). Record-high 50% of Americans favor legalizing Marijuana use. http://www.gallup.com/poll/150149/record-high-americans-favor-legalizing-marijuana.aspx

27.

Pew. (2013). Majority now supports legalizing Marijuana. http://www.people-press.org/2013/04/04/majority-now-supports-legalizing-marijuana/

28.

Pew. (2014). America’s new drug policy landscape. http://www.people-press.org/2014/04/02/americas-new-drug-policy-landscape/

29.

Pew. (2018). About six-in-ten Americans support marijuana legalization. http://www.pewresearch.org/fact-tank/2018/01/05/americans-support-marijuana-legalization/

30.

Project-CBD. (2015). What is CBD? https://www.projectcbd.org/what-cbd

31.

Saad

(2014). Majority continues to support pot legalization in U.S. http://www.gallup.com/poll/179195/majority-continues-support-pot-legalization.aspx

32.

Schroyer

McVey

(2019). Chart: Most California municipalities ban commercial cannabis activity. https://mjbizdaily.com/chart-most-of-california-municipalities-ban-commercial-cannabis-activity/

33.

Selsky

(2016). The tides are shifting in Oregon over legal marijuana. The Cannabist. https://www.thecannabist.co/2016/05/16/tides-shifting-oregon-marijuana-legalization/54172/

34.

Simkins

Geiger-Oneto

(2015). A social identity perspective on the legalization of Marijuana in the United States. Robert Mittelstaedt Doctoral Symposium Proceedings. University of Nebraska. http://cba.unl.edu/academic-programs/departments/marketing/about/robert-mittelstaedt-doctoral-symposium/docs/2015_SymposiumProceedings.pdf

35.

Socia

K. M.

Brown

E. K.

(2014). Up in smoke: The passage of medical Marijuana legislation and enactment of dispensary moratoriums in Massachusetts. Crime & Delinquency. https://doi.org/10.1177/0011128714557024

36.

Thornhill

T. E.

(2011). African Americans and the marijuana legalization paradox: Do race-specific murder victimization rates and race-specific drug arrest rates explain it? Journal of Ethnicity in Criminal Justice, 9(2), 110–135.

37.

Toch

Maguire

(2014). Public opinion regarding crime, criminal justice, and related topics: A retrospect. Journal of Research in Crime and Delinquency, 51(4), 424–444. https://doi.org/10.1177/0022427813520444

38.

Weakliem

D. L.

(2005). Public opinion, political attitudes, and ideology. In Janoski

(Ed.), The handbook of political sociology: States, civil societies, and globalization (pp. 227–246). Cambridge University Press.

39.

Zaller

(1992). The nature and origins of mass opinion. Cambridge University Press.

40.

Zezima

(2018). Vermont is the first state to legalize Marijuana through legislature. https://www.washingtonpost.com/news/post-nation/wp/2018/01/23/vermont-is-the-first-state-to-legalize-marijuana-through-legislature/?utm_term=.91247257bcf1