Who Influences Whom? Elite and Citizen Preferences Regarding Marijuana and Drug Legalisation in Latin America

Abstract

In Latin America, public opinion support for drug legalisation is generally low. Yet since the 2010s, drug policy debates have grown, and some politicians became reform advocates. This may suggest a disconnect between elites and citizens. We tentatively explore the relationship between elites’ and citizens’ opinions using PELA and Latinobarómetro surveys. The findings suggest a more reciprocal relationship than we originally assumed, as the percentage of population agreeing with marijuana legalisation is associated with higher legislator support for drug legalisation, and presidential and legislator support for drug legalisation are associated with higher – but not more intense – citizen support for marijuana legalisation. Although better data is needed to further probe our findings, we contend that increases in elite and public support can move each other toward reform, albeit in small proportions. Based on these results we suggest a hypothesis for future research: although drug legalisation is a moral issue where voter predispositions are strong, it can also be conceptualised as a “hard issue” given its regulatory complexity, making opinions sensitive to debates and framing.

Introduction

In December 2013, Uruguay became the first country in the world to regulate the production, sale, and use of recreational cannabis. By 2020, 20 countries in Latin America and the Caribbean had adopted medical cannabis stipulations, many more than a decade before (Igarapé, 2023). These developments marked a shift in a region that for most of the twentieth century adhered to the prohibitionist spirit of international drug control. This trend is puzzling because it unfolded amid low public support for drug legalisation, raising questions about the relationship between elites’ and citizens’ preferences for drug reform.

This research note analyses the connection between citizen and elite preferences using data from the 2015 Latinobarómetro survey on citizen support for marijuana legalisation and data from 27 out of 39 iterations of the PELA-USAL (Parliamentary Elites of Latin America) survey on legislator support for drug policy. Given the timing of PELA surveys (one per legislative cycle) we included surveys closest to 2015, taking advantage of variation in timing (before or after 2015) to tentatively explore the direction of the relationship between elites’ and citizens’ opinions. Given different question wordings, we present tentative results highlighting the importance of generating better data on drug policy preferences. Our findings suggest a more reciprocal relationship between elites and citizen preferences on drug legalisation than often assumed, less dependent on the overall intensity of support for legalisation in each group, and more dependent on increases of support within the population and the existence of sizeable (but not necessarily majority) groups of legislators supporting reform. In both directions, increases in support seem to move opinions towards supporting drug reform, albeit in small proportions. We suggest that this can reflect that although legalisation is a moral issue where preferences are difficult to change, it is, compared to other moral issues, a “hard issue” (Carmines and Stimson, 1980) sensitive to political discussions, given the complexity of regulatory options and its connection to several policy areas including health, education, and security.

We proceed as follows. The first section reviews the literature on the relationship between public and elite opinion focusing on drug policy and presents the hypotheses. The second describes data and methods. The third discusses the findings. We conclude suggesting future research avenues.

Public Opinion, Elite Opinion, and Drug Legalisation

Scholarship on the relationship between public and elite opinion on drug policy in Latin America is limited, although a few recent works are notable. Mendiburo-Seguel et al. (2017) analyse public attitudes toward drug policy, highlighting significant variation across countries. On the elite side, recent work using PELA data examines how ideology, religiosity, and institutional factors shape legislators’ positions on moral issues, including drug legalisation (Bohigues and Fernández-i-Marín, 2022; Bohigues et al., 2022). These studies offer valuable insights, although they do not explicitly examine the interaction between elites and citizens.

Support for drug and marijuana legalisation in Latin America is low among elites and citizens – lower than support for abortion or same-sex marriage. In 2015, 24.6 per cent of the region's population supported same-sex marriage, but only 16.4 per cent backed marijuana legalisation (Latinobarómetro). PELA surveys show that while 52 per cent of legislators support same-sex marriage, only 37 per cent support drug legalisation. Strong support is also lower for legalisation (9.6 per cent) than for same-sex marriage (23.6 per cent) or abortion (14 per cent).

Yet, despite limited support, policy and discourse have shifted. From 2000 to 2010, only 12 marijuana-related bills were introduced in the region; this rose to 137 between 2011 and 2021, though some countries saw no such initiatives (Durán-Martínez and Pennell, 2024). Drug policy changes – mostly involving medical marijuana or decriminalisation – have come from elites via legislation, executive or ministerial decrees, or court rulings that mandate cannabis access for claimants, setting precedents. This contrasts with the United States, where many recreational marijuana reforms emerged through citizen initiatives rather than elite-driven processes. These initiatives challenged the assumption that voters primarily follow partisan cues as public opinion changed more drastically and earlier than legislator opinion, and instead point to the influence of secularisation and shifting public attitudes (Bradford and Bradford, 2017; Cruz et al., 2016; Dyck and Lascher, 2019).

Some argue Latin American reforms have been top-down, led by charismatic leaders and pro-reform elite coalitions (Durán-Martínez and Pennell, 2024). This suggests a disconnect between elites and public opinion, or that influence flows primarily from elites to citizens. Yet no systematic analysis has tested this. Given other work highlighting bottom-up mobilisation in drug reform (Von Hoffmann, 2020), it is important to examine whether public opinion shapes elite views.

Because no prior studies assess the relationship between citizen and elite preferences on drug legalisation, we draw on broader literature on representation and drug policy, while considering the nature of drug policy as a moral issue. Like abortion or same-sex marriage, drug policy is a moral issue that can be an “easy issue,” as defined by Carmines and Stimson (1980), where voter predispositions are strong and tied to ethical or religious beliefs, unlike “hard issues” which are more difficult to understand and where voter predispositions are less strong. We tentatively suggest that drug policy can also be a “hard issue” given its multiple regulatory pathways and connections to public safety, health, education, and crime. “Hard issues” can be more influenced by framing and public debate (Dyck and Pearson-Merkowitz, 2019). This motivates our use of indicators capturing both the level and intensity of support, as detailed below.

Hypotheses

The first hypothesis connects drug policy literature and research on politicians as informational shortcuts for ideologically aligned citizens. In Latin America, prominent politicians led drug debates in the mid-2010s, and a modest rise in public support occurred during the same period (Cruz et al., 2016). In 2009, three former presidents formed the Latin American Commission on Drugs and Democracy, shifting the discussion of legalisation from activist and academic to political circles. These presidents also advocated to hold an earlier UN General Assembly Special Session on the World Drug Problem, a key venue for drug debates (Idler and Garzón, 2021). It is plausible to think that these elite signals shaped public preferences.

There is little research on elite influence on drug policy public opinion in Latin America; the most rigorous study is Cruz et al. (2016), who found that views of President José Mujica influenced support for marijuana legalisation in Uruguay, but did not find similar effects in the United States. Given that most drug reforms in Latin America have been top-down and that some presidents championed reform in the 2010s, it is reasonable to expect that presidential endorsement could increase citizen support, by providing cues to co-partisans and signalling broader social normalisation. While this may polarise opinions among ideological opponents, the stigma traditionally surrounding drug policy may mean that presidential support makes openness to reform more socially acceptable. Thus, we hypothesise:

H1. Presidents’ public support for drug policy reform may increase levels of public opinion support.

The relationship between legislators and public opinion can go in different directions. The literature on elite-citizen connections in the United States suggests that citizens generally follow partisan cues, and parties tend to respond to their constituencies, but this depends on issue salience and partisan divisions (Lax and Phillips, 2012; Pomirchy and Schonfeld, 2019). Not all legislators reveal their preferences publicly, or only a handful of them do. This may depend on issue salience, defined as the prominence of an issue within the public or policymakers’ consciousness and often operationalised through the frequency of judicial and legislative discussions, or media attention (Dennison, 2019, Helbling and Tresch, 2011). As Figure 1 shows, increasing media attention suggests that drug debates became salient in the 2010s across Latin America, possibly making legislators more likely to voice their preferences, or to reveal them through voting of legislative proposals, thus providing information shortcuts that influenced public opinion. Yet because drug policy debates were not equally salient across countries, this effect may vary across countries. We thus hypothesise:

H2: Legislators’ preferences on drug reform are likely to influence public opinion, particularly in countries where drug policy debates are salient.

Figure 1.

Salience of Drug Policy (News Articles Mentioning Legalisation per Year Across Latin America).

Legislators may follow public opinion to enhance electoral prospects. Yet, in Latin America this influence is expected to be weaker than in consolidated democracies because some party systems may be more unstable and personalistic. Giles et al. (2008) argue that perceptions of whether the population strongly opposes or supports an issue can influence elite decisions even when tenure does not depend on public opinion, as is the case of the U.S. Supreme Court. Recent studies using PELA data (Bohigues and Fernández-i-Marín, 2022; Bohigues et al., 2022) find that public perceptions of insecurity mediate the effects of individual-characteristics on legislators’ preferences, suggesting some responsiveness to public sentiment. Drawing on these insights, we hypothesise that even in Latin America's varied, and sometimes volatile party systems, legislators’ preferences can be associated to increases in support to drug legalisation, more than to the overall intensity of citizen support, which is low.

H3: Legislators’ preferences are more likely to be connected to increases in the percentage of citizens who express support for legalisation than to the average intensity of citizen support.

Next, we explore whether ideology and religion mediate the relationship citizens-politicians. Existing scholarship suggests that the influence of religion on citizens’ electoral choices varies across contexts and religious types and forms of organisation. Religious organisations, particularly Evangelical (Pew, 2014), organise members to shape electoral outcomes as occurred in Brazil's 2018 elections, or in Mexico's policy discussions on abortion (Reuterswärd, 2021; Smith, 2019). Such influence can be significant even when churchgoers lack strong feelings on a policy. For instance, in the vote for the referendum approving Colombia's peace agreement in 2016, opposition parties organised a campaign linking the referendum to sensitive social issues, successfully mobilising religious voters who did not have initial strong opinions (Dávalos et al., 2018). Citizens’ religious and ideological beliefs may be strongly ingrained, which can weaken the influence of general elite preferences. Religious individuals may be less likely to follow overall elite cues but can be more likely to follow religious’ legislators. Likewise, there is consistent evidence of the effect of ideology on citizen preferences for legalisation (Cruz et al., 2016; Sánchez and Ortiz, 2008) leading us to hypothesise:

H4. The preferences of highly religious and strongly ideological citizens are less likely to be connected to overall legislators’ preferences than those of non-religious or weakly ideological citizens.

Analyses of legislator opinion on moral issues like abortion, drug legalisation, and LGBTQ rights find that individual religiosity and ideology (Bohigues et al., 2022), and a country's overall religiosity (González-Rostani and Morgenstern, 2023) are predictors of legislator opinion, particularly for highly religious and right-leaning legislators. Legislators’ strong ideological and religious beliefs weaken their connection to citizens, because their electoral appeal is likely connected to these strong beliefs (Bohigues and Fernández-i-Marín, 2022). We thus hypothesise:

H5. The preferences of highly religious and right-wing legislators are more likely to be strong (located to the extremes of the scale of support) than those of non-religious and centrist legislators, making them potentially more difficult to change and less likely to be connected to citizens’ preferences.

Data and Methods

PELA-USAL is a survey of parliamentary elites in 18 Latin American countries, conducted by the University of Salamanca since 1994. In 2012, PELA began asking: “On a scale from 1 to 10 (1 = strong disapproval, 10 = strong approval), how strongly do you approve or disapprove of drug legalization?.” This question appears in 39 surveys across 18 countries; we use data from 27 waves closest to the 2015 Latinobarómetro survey.

Systematic public opinion data on drug legalisation is limited. Institutions such as Asuntos del Sur (2015) and Brazil's Datafolha field surveys on drug policy, but infrequently or covering few countries. The Latin American Public Opinion Project included marijuana-related questions in Colombia (2014), El Salvador (2014), and Uruguay (2014, 2017), but not elsewhere. To our knowledge, the 2015 Latinobarómetro is the only region-wide survey asking about drug policy. It asked: “Do you strongly agree, agree, disagree, or strongly disagree with the legalisation of marijuana?.” We reverse-coded responses (1–4) to associate higher values with greater support. Latinobarómetro asks about marijuana legalisation, while PELA-USAL asks about drug legalisation and this distinction likely affects responses. Public support tends to be higher for marijuana than for all “drugs,” which often implies harder substances, and preferences may vary by policy. These differences in wording are meaningful. We do imply that these questions are interchangeable, but that given imperfect existing data, they can inform further research.

The data reveal interesting patterns. The lowest legislator support appears in Nicaragua’17 (2.7/10), Panama’14 (2.1/10), and Panama’18 (2.0/10); the highest in Mexico’18 (5.7/10), Colombia’18 (6.2/10), and Uruguay’15 (7.2/10). In South America, Uruguay’15 shows the highest support (7.28/10), while Bolivia ranks lowest (2.12/10). In Central America, Costa Rica’18 leads (3.7/10). Patterns of polarisation also vary: for example, 29 per cent of Chile’14 and 28 per cent of Paraguay’19 legislators strongly oppose legalisation, but Chile's higher average score suggests more polarised positions (Figure 2).

Figure 2.

Legislator Support for Drug Legalisation.

Of the 14 countries where more than one PELA wave contains the drug legalisation question, there is increasing support over time in Argentina, Bolivia, Colombia, Costa Rica, Ecuador (2017–2021), El Salvador (2012–2015), Honduras (2014–2018), Mexico, Panama, Paraguay, and Republica Dominicana suggesting a regional tendency towards growing legislator support. Decreases in support appear in Chile, Ecuador (2013–2017), El Salvador (2015–2018), Honduras (2018–2022), Peru, and Uruguay (2015–2022).

Regarding citizen support, Latinobarómetro shows the highest percentages of population agreeing with marijuana legalisation in Chile (53 per cent) and Uruguay (33.2 per cent) and the lowest in Nicaragua, Guatemala, and Venezuela. Figure 3 illustrates how although legislator opinion on drug legalisation and public opinion on marijuana legalisation correlate, legislator support varies more. We cannot assess changes in public opinion because Latinobarómetro has not asked this question again, yet other sources suggest that public opinion changes occur. In Uruguay when the regulation of recreational cannabis was approved most people opposed it, but support increased shortly after (Cruz et al., 2018).

Figure 3.

Legislator and Citizen Support for Drug Legalisation.

Methods

We use multilevel regression and use the timing of PELA surveys to tentatively explore the direction of the relation between citizen and elite preferences. We use original categorical variables and binary recodings where 0 corresponds to opposing drug legalisation and includes any response of 4 or below (PELA) or strongly oppose and oppose (Latinobarómetro), and 1 corresponds to 6 or above (PELA) and strongly agree or agree (Latinobarómetro). This is a more conservative operationalisation of support than the one used by Bohigues and Fernández-i-Marin (2022) in analysing PELA drug responses, but the appendix shows results with their measure and another alternate coding of support. We use both categorical and binary variables to assess both the openness to, and the intensity of, support.

To assess H1, H2 and H4 we estimate models where the DV is citizen agreement with marijuana legalisation. The main independent variable is a second level predictor corresponding to the country mean score of support among legislators, which indicates the average intensity of legislator support. We include a variable coding a President's support for legalisation ranging from 0 to 2: 0 indicates strong opposition; 1 corresponds to cases where Presidents expressed openness to drug policy reform but not necessarily legalisation, or cases where they expressed support for aspects of drug reform (such as regulating medical marijuana) but opposed full legalisation; 2 corresponds to cases where Presidents openly critiqued the War on Drugs and/or expressed support for specific reforms (frequently marijuana legalisation). ¹ To assess issue salience, we include a dummy variable capturing whether news articles with the keywords “legalización,” “marihuana” and derivates, appeared in each country in a search using Nexis Uni between 2000 and 2020.

These regressions use legislator surveys conducted before ² the 2015 wave of Latinobarómetro. First-level (individual) controls include income, education, religiosity, ideology, fear of crime, crime victimisation, trust in democracy, urban resident. Second level (country) controls include homicide rates, GDP, schooling, infant mortality, population, and an indicator on the type of electoral systems created by Crisp et al. (2020), which ranges from 1 to 6, where 1 represents “weak systems” that tend to generate proportional representation, and 6 represents “strong” systems where proportionality is low (Crisp et al., 2020).

To assess H3 and H5, the DV is individual legislator support for drug legalisation (binary and scale) including similar first- and second-level predictors as in the first set of regressions. The main IV is a second-level predictor corresponding to the average score of citizen support per country or the percentage of respondents declaring support (binary) to marijuana legalisation. We use different codings of the IV to explore whether the average intensity of support or the percentage of support drive the relationship between public opinion and elite opinion.

Results and Analysis

To assess H1 and H2 we use citizen support for marijuana legalisation as DV in multilevel ordered logit regressions where the unit of analysis is the individual respondent in Latinobarómetro. Results appear in Table 1 with exponentiated coefficients (values greater than 1 mean positive associations). In Models 1–2, the DV is the indicator on a scale of 1–4 of whether a respondent agrees or disagrees with marijuana legalisation, in Models 3–4, the DV is a binary indicator of support for marijuana legalisation. Models 5–6 split the sample between countries where media salience of legalisation is high/low. All models include the same controls. Results are consistent with H1: across all models, Presidential support for legalisation is positively associated with citizen support for marijuana legalisation. This association is only significant when President support is moderate (1 in our coding), but stronger levels of President support have significant impact among those approving the President's performance (interaction between Presidential approval and President's position). The results suggest that moderate Presidential support for legalisation is associated with higher levels of citizen support for marijuana legalisation across the board, but strong levels of President support may simultaneously strengthen support among those who approve of the president while somewhat increasing opposition among those who don’t approve. This effect on opposition is less significant (regressions with opposition as the DV appear in the appendix).

Table 1.

Effect of Legislator Support for Drug Legalisation on Citizen Support for Marijuana Legalisation.

DV: Citizen supportMarijuana legalisation	(1)	(2)	(3)	(4)	(5)	(6)
	Support legalisation 1–4	Support legalisation 1–4	Support legalisation 1–0	Support legalisation 1–0	Support legalisation 1–4	Support legalisation 1–4
Main independent variables					Media salience	No media salience
Legislator support legal (>5)	1.019**		1.005***		1.021***	1.019
	(2.08)		(3.27)		(3.84)	(1.47)
Legislator mean support leg.		1.174		1.056***
		(1.38)		(2.73)
President support legal (1)	2.096***	2.127**	1.154***	1.154***	1	7.438***
	(2.66)	(2.52)	(2.97)	(2.74)	(.)	(4.29)
President support legal (2)	0.582*	0.661	0.894**	0.903*	1	1
	(−1.65)	(−1.16)	(−1.97)	(−1.65)	(.)	(.)
President approval × president support legal (1)	1.120 (1.06)	1.121 (1.07)	1.061*** (2.95)	1.061*** (2.94)	1	1.434 (1.62)
President approval × president support legal (2)	1.435*** (3.03)	1.438*** (3.04)	1.106*** (4.55)	1.105*** (4.54)	1.410** (2.54)	1.259 (1.27)
First-level controls
Age	0.840***	0.840***	0.973***	0.973***	0.793***	0.899***
	(−7.09)	(−7.09)	(−5.97)	(−5.98)	(−6.54)	(−3.08)
Education	1.088***	1.088***	1.016***	1.016***	1.091***	1.088***
	(5.48)	(5.47)	(5.61)	(5.59)	(3.82)	(4.02)
Female	0.742***	0.742***	0.944***	0.944***	0.676***	0.814***
	(−6.69)	(−6.69)	(−6.94)	(−6.94)	(−6.17)	(−3.26)
Ideology	0.962***	0.962***	0.992***	0.992***	0.940***	0.983
	(−4.70)	(−4.68)	(−5.40)	(−5.36)	(−5.11)	(−1.48)
Religiosity	0.832***	0.832***	0.964***	0.964***	0.786***	0.912**
	(−6.70)	(−6.71)	(−7.16)	(−7.17)	(−6.34)	(−2.29)
Urban resident	1.042***	1.042***	1.009***	1.009***	1.058***	1.022
	(3.72)	(3.72)	(4.42)	(4.43)	(3.69)	(1.33)
Income level	1.112***	1.111***	1.016***	1.016***	1.170***	1.051
	(3.64)	(3.62)	(2.90)	(2.89)	(3.69)	(1.24)
Victimisation	1.120*	1.121*	1.027**	1.027**	1.091	1.168*
	(1.94)	(1.96)	(2.44)	(2.44)	(1.07)	(1.83)
Fear of crime	0.973	0.973	0.996	0.996	0.921**	1.031
	(−1.26)	(−1.27)	(−0.98)	(−0.98)	(−2.57)	(1.01)
Report drug use friend	1.299***	1.300***	1.045***	1.046***	1.226***	1.368***
	(5.25)	(5.26)	(4.73)	(4.77)	(3.06)	(4.15)
Trust pol. parties	1.068**	1.069**	1.009*	1.009*	1.088**	1.059
	(2.38)	(2.39)	(1.82)	(1.81)	(2.06)	(1.49)
Belong pol. party	0.924*	0.924*	1.008	1.008	0.919	0.935
	(−1.73)	(−1.73)	(0.97)	(0.97)	(−1.27)	(−1.04)
Satisfaction democracy	1.056*	1.056*	1.016***	1.016***	1.075*	1.036
	(1.80)	(1.80)	(2.87)	(2.86)	(1.66)	(0.83)
Gangs present in neighb.	0.947	0.946	0.999	0.998	0.975	0.921
	(−1.07)	(−1.10)	(−0.14)	(−0.20)	(−0.36)	(−1.12)
Drugs present in neighb.	1.131**	1.133**	1.024**	1.024**	1.114	1.156*
	(2.18)	(2.21)	(2.24)	(2.27)	(1.30)	(1.86)
Presidential approval	0.936	0.933	0.961***	0.961***	1.001	0.916
	(−0.85)	(−0.89)	(−2.73)	(−2.72)	(0.01)	(−1.09)
Second-level controls
Homicide rates	1.016***	1.016***	1.002***	1.002***	0.978***	1.021***
	(4.24)	(3.85)	(3.38)	(2.94)	(−2.58)	(6.03)
GDP per capita	1.000***	1.000***	1.000***	1.000***	1.000***	1.000***
	(4.71)	(4.93)	(3.19)	(3.42)	(10.18)	(5.33)
Mortality rate	0.782***	0.807***	0.946***	0.951***	0.545***	0.579***
	(−3.25)	(−2.74)	(−4.22)	(−3.69)	(−10.17)	(−6.08)
Population	1.000***	1.000***	1.000***	1.000***	1.000	1.000***
	(−4.25)	(−3.76)	(−4.40)	(−3.81)	(0.87)	(−3.81)
N	7705	7705	7705	7705	3742	3963

Note: Exponentiated coefficients; t statistics in parentheses.

*p < .10, **p < .05, ***p < .01.

Regarding H2 we find a positive and significant association between legislator support and citizen support. Results suggest that the percentage of legislator support is associated with whether citizens support or oppose marijuana legalisation more than with how strongly they do so (mean support), and the effect is stronger where media salience is greater (Model 5). Positive coefficients in models with the binary variable as the outcome suggest that citizens move towards more support when elite support increases, with the effect being more consistent on whether citizens support than on how intensely they support. The appendix presents models with legislator opposition, median support for drug legalisation, and alternative codings of binary legislator support as IVs, as well as OLS specifications, with similar results.

To assess H3 the next regressions use legislator support for legalisation as DV with the individual legislator as the unit of analysis. The DV is a categorical variable (1–10 scale of support, Models 1–2, Table 2) or a binary indicator of whether a legislator supports legalisation (Models 3–4). In Models 1 and 3, the main predictor is the percentage of citizens who agree with legalisation; in Models 2 and 4, the main predictor is the mean of the scale support for marijuana legalisation. Results are partially consistent with H3, showing that as the percentage of citizens who declare agreeing with marijuana legalisation increase, the odds of legislator support for drug legalisation increase (results are robust to different codings of legislator support and OLS models, see appendix). Yet mean citizen support has no association and in one model the coefficient is negative. This suggests that the intensity of support influences legislators less than whether there is increasing population support which may move legislators towards supporting some reform even if the population at large remains opposed. Results in Tables 1 and 2 suggest a complex relation between citizen and legislator preferences, with the percentage of support among citizens – but not the intensity of population support – influencing legislators support, and legislator support showing an association with whether citizens are likely to support marijuana legalisation, but not with how strongly they do so. Regressions reported in the appendix suggest that the percentage of citizen opposition has an association with legislator support, but the effect size is smaller, suggesting that legislators are more sensitive to increases in support.

Table 2.

Effect of Citizen Support for Marijuana Legalisation on Legislator Support for Drug Legalisation.

DV: Legislator support for drug legalisation	(1)Support 1–10	(2)Support 1–10	(3)Support 1–0	(4)Support 1–0
Main independent variables
% citizen support marijuana legalisation	11291.6** (2.29)		3.799*** (3.09)
Mean citizen support marijuana legalisation		0.191 (−1.13)		0.715** (−2.29)
Individual legislator (first) level controls
Female	0.996	1.005	1.005	1.010
	(−0.03)	(0.03)	(0.19)	(0.37)
Age	0.985***	0.985***	0.997***	0.997***
	(−2.63)	(−2.69)	(−2.69)	(−2.98)
Education	1.276***	1.263***	1.023	1.020
	(3.08)	(2.95)	(1.62)	(1.40)
Income	1.002	1.003	1.006	1.008
	(0.03)	(0.04)	(0.39)	(0.50)
Ideology	0.902***	0.902***	0.989*	0.990*
	(−3.17)	(−3.18)	(−1.89)	(−1.75)
Religiosity	0.690***	0.691***	0.936***	0.936***
	(−6.25)	(−6.24)	(−5.86)	(−5.94)
Concern w. insecurity	1.043	1.045	1.002	1.002
	(1.09)	(1.13)	(0.23)	(0.31)
New legislator	0.998	1.006	0.979	0.984
	(−0.01)	(0.05)	(−0.93)	(−0.70)
Opposition legislator	1.509*	1.541*	1.072	1.088*
	(1.77)	(1.85)	(1.62)	(1.93)
President support drug legal (1)	1.436	1.271	1.022	1.043
	(0.74)	(0.41)	(0.35)	(0.62)
President support drug legal (2)	1.506	2.104	1.045	1.136*
	(0.86)	(1.20)	(0.72)	(1.70)
Opposition legis* president support 1	0.623	0.606*	0.930	0.911*
	(−1.59)	(−1.68)	(−1.31)	(−1.65)
Opposition legis* president support 2	0.799	0.790	0.976	0.943
Country (second) level predictors
Homicide rate	1.003	1.002	1.000	1.000
	(0.46)	(0.28)	(−0.33)	(−0.38)
GDP per capita	1.000	1.000	1.000	1.000
	(−0.38)	(0.06)	(−1.04)	(0.33)
Mortality	0.844	0.985	0.995	1.006
	(−0.93)	(−0.08)	(−0.27)	(0.31)
Population	1.000**	1.000**	1.000***	1.000***
	(2.08)	(2.06)	(5.36)	(4.65)
Electoral system	0.884	0.970		0.977
	(−0.51)	(−0.11)		(−0.78)
N	879	879	879	879

Note: Exponentiated coefficients; t statistics in parentheses.

* p < .10, ** p < .05, *** p < .01.

Source: Authors’ elaboration.

To assess H4 and H5 we include interaction terms between citizen/legislator support, and religiosity/ideology. Figure 4 plots marginal effects of the interaction for ease of interpretation (tables appear in appendix). Figure 4, left panel, shows results for citizen support for marijuana legalisation as the DV, plotting the interaction between percentage of legislator support and citizen ideology (top left) and religiosity (bottom left). The results align with the expectation that as the percentage of legislator support for drug legalisation increases, citizens declare higher support for drug legalisation at all levels of ideology and religiosity, except for citizens on the right of both scales, for whom there is no significant effect of legislator support on their preferences (results hold when using support by religious legislators as IV).

Figure 4.

Interactions between Citizen/Legislator Support, and Ideology/Religiosity (Marginal Effects).

Figure 4, right panel, displays results for models where legislator support is the dependent variable, plotting the interaction between percentage of citizens supporting marijuana legalisation and legislator ideology (top right) and religiosity (bottom right). The results show that as population support for marijuana legalisation increases, legislators declare higher support for drug legalisation at all levels of ideology, except for legislators on the right, for whom there is no significant effect of citizen support. Results for religiosity are similar: as the percentage of population support increases, legislators are more likely to support legalisation, except for highly religious legislators. These results support H4 and H5, although the moderating effect of extreme religiosity and religious beliefs seems stronger for legislators than citizens and for extremes of the scale.

These results suggest a nuanced relationship between citizen and elite preferences. Legislators seem to be influenced by public opinion more than one would expect from anecdotal evidence, and citizens’ tendency to support (rather than the intensity of support) appears influenced by overall legislator support. Yet, the fact that the most significant results appear when using the scale of legislator support as the DV and the percentage of citizen support as IV may indicate that when pockets of support emerge among the population, legislators move between margins of support for certain drug reforms without radically changing positions. For example, more legislators are now open to supporting regulating medical marijuana, while still opposing recreational cannabis. The fact that the percentage of population agreeing is the most significant variable suggests, as some studies posit, that social mobilisation may move drug reform debates forward even if the population's overall mood remains opposed to legalisation, assuming that strong supporters are more likely to mobilise. These results suggest that small differences in levels of support are important. Citizen support for marijuana legalisation may increase as the overall legislature mood becomes more supportive of legalisation and when Presidents declare moderate support. Results for H4 and H5 suggest mediating effects of religion and ideology on the relation between citizens and legislators’ opinions for highly religious and rightist legislators and citizens, but not for centrist, left leaning, or non-religious citizens. Less determinate effects for high, but not extreme, conservatism and religiosity suggest that religion and ideology's relation with drug policy preferences may be more malleable than for other moral issues.

Conclusion

We found evidence consistent with the idea that legislators and citizens influence each other's preferences on drug and marijuana legalisation, although in nuanced ways. Citizen preferences influence legislator preferences on drug legalisation, except for highly religious and extreme right legislators. The evidence is consistent with the possibility that the percentage of population agreeing with legalisation at some level influences legislators more than the strength of citizen support. We also found that Presidential and legislator support for drug legalisation is associated with higher citizen support (though not necessarily with how strongly citizens support marijuana legalisation). Based on these results, we propose an argument that should be subject to further scrutiny: when there are growing pockets of support within the population, or among legislators, legalisation debates may advance, even if the average intensity of support is low and the population remains sceptical of reform. If legislators express support – even if moderate – this can move public opinion towards acceptance of – moderate – policy reforms. This suggests that framing, policy changes, and debates may increase support for moderate reforms, but radical preference changes from strong opposition to strong support remain difficult.

While drug legalisation is a moral issue, we suggest that it may also exhibit characteristics of a “hard issue” (Carmines and Stimson, 1980), and preferences may thus appear responsive to elite signalling and political framing. We do not claim to test the hard/easy issue distinction, but we suggest that the complexity of regulatory options, the overlap with other policy domains (e.g. health, education, security), and the apparent responsiveness of moderate citizens and legislators may indicate a more malleable opinion landscape than typically found in moral debates. This possibility deserves further exploration, in dialogue with scholarship on morality in drug policy and evidence utilisation in reform debates.

To further understand relationships between citizen and legislator preferences collecting more data is necessary. PELA provides invaluable data to assess changes over time that could be improved by including questions regarding different drug policy aspects. Existing public opinion surveys could include fine-grained questions about legalisation consistently to assess temporal variation and complex preferences. Analyzing the influence of public opinion debates, and media coverage, on citizens and legislators, also constitute avenues for future research.

Supplemental Material

sj-docx-1-pla-10.1177_1866802X251369520 - Supplemental material for Who Influences Whom? Elite and Citizen Preferences Regarding Marijuana and Drug Legalisation in Latin America

Supplemental material, sj-docx-1-pla-10.1177_1866802X251369520 for Who Influences Whom? Elite and Citizen Preferences Regarding Marijuana and Drug Legalisation in Latin America by Angélica Durán-Martínez and Renato Fakhoury in Journal of Politics in Latin America