Assessing options for cannabis law reform: A Multi-Criteria Decision Analysis (MCDA) with stakeholders in New Zealand

Background: A number of jurisdictions are considering or implementing diﬀerent options for cannabis law reform, including New Zealand. Multi-Criteria Decision Analysis (MCDA) helps facilitate the resolution of complex policy decisions by breaking them down into key criteria and drawing on the combined knowledge of experts from various backgrounds. Aims: To rank cannabis law reform options by facilitating expert stakeholders to express preferences for projected reform outcomes using MCDA. Methods: A group of cannabis policy experts projected the outcomes of eight cannabis policy options (i.e., pro- hibition, decriminalization, social clubs, government monopoly, not-for-proﬁt trusts, strict regulation, light regulation, and unrestricted market) based on ﬁve criteria (i.e., health and social harm, illegal market size, arrests, tax income, treatment services). A facilitated workshop of 42 key national stakeholders expressed preferences for diﬀerent reform outcomes and doing so generated relative weights for each criterion and level. The resulting weights were then used to rank the eight policy options. Results: The relative weighting of the criteria were: “reducing health and social harm ” (46%), “reducing arrests ” (31%), “reducing the illegal market ” (13%), “expanding treatment ” (8%) and “earning tax ” (2%). The top ranked reform options were: “government monopoly ”(81%), “not-for-proﬁt ”(73%) and “strict market regulation ”(65%). These three received higher scores due to their projected lower impact on health and social harm, medium reduction in arrests, and medium reduction in the illegal market. The “lightly regulated market ” option scored lower largely due its projected greater increase in health and social harm. “Prohibition ” ranked lowest due to its lack of impact on reducing the number of arrests or size of the illegal market. Conclusion: Strictly regulated legal market options were ranked higher than both the current prohibition, and alternatively, more lightly regulated legal market options, as they were projected to minimize health and social harms while substantially reducing arrests and the illegal market.


Introduction
The enactment of recreational cannabis legalization in Uruguay, Canada and 18 U.S. states (plus Washington DC) has reignited the tivism against the prohibition of cannabis and related enforcement operations, such as the annual cannabis crop eradication operation, has been going on for decades. Yet unlike neighboring Australia, where several states implemented cannabis decriminalization during the 1980s and 1990s (Hughes, 2020), the prohibition of cannabis has remained largely unchanged in New Zealand, apart from minor cannabis offences being included in a general criminal justice movement away from prosecution to use of diversion and formal warnings for minor offending ( Wilkins & Sweetsur, 2012 ).
In 2017, the New Zealand Green Party, which had advocated for cannabis law reform for many years, secured a commitment to hold a referendum on the legal status of cannabis in New Zealand as part of its support for the new Labor government. A national referendum on a proposal to legalize recreational cannabis use and supply was held in 2020 (NZ Herald, 2017 ) and did not pass . The New Zealand referendum was notable as it was the first to be held at the national level and to involve voting on a detailed proposal setting out a specific framework to establish a legal market for cannabis rather than a general question about whether cannabis should be legal or not ( Wilkins & Rychert, 2020a ). Despite the rejection of the legalization referendum, debate about the legal status of cannabis has continued in New Zealand and alternative proposals have been put forward ( Wilson, 2020 ), including a petition to introduce decriminalization ( Heatherington, 2020a,b ).
Comparing different cannabis law reform options involves assessing multiple outcomes and trade-offs, and there are important gaps in data and lack of consensus regarding the longer term outcomes of different policy approaches or specific regulatory components ( European Monitoring Centre for Drugs and Drug Addiction (EM-CDDA), 2020 ; Fischer et al., 2020b ;Hall et al., 2019 ;Leung et al., 2018 ;Leung et al., 2019 ;Smart & Pacula, 2019 ). Preliminary evaluations of recreational cannabis legalization in select U.S. states, Uruguay and Canada have concluded that it may take a decade or more to fully understand the consequences of legalization ( European Monitoring Centre for Drugs and Drug Addiction (EMCDDA), 2020 ; Fischer et al., 2020b ;Leung et al., 2018 ;Leung et al., 2019 ;Smart & Pacula, 2019 ). This inconclusiveness reflects a range of factors including time taken to establish retail cannabis outlets and markets; pre-existing cannabis trends and regimes for medicinal cannabis; slow evolving social norms related to cannabis use; time lags between cannabis use and a range of health and social outcomes, and yet to be felt impacts of the maturation of legal cannabis markets and related commercial industry ( European Monitoring Centre for Drugs and Drug Addiction (EMCDDA), 2020 ; Leung et al. 2019 ;Smart & Pacula, 2019 ;Leung et al., 2018 ). The evaluation of cannabis legalization in U.S. states is further complicated by the continuing federal prohibition and related regulatory gaps and barriers to doing business. In addition, it has been recognized that how individuals and groups normatively value different cannabis reform objectives and outcomes influences their preferred reform option ( Rogeberg, 2018. Drug policy researchers have long pointed out there are a range of alternative 'middle ground' options to cannabis policy reform between prohibition on the one hand and unrestricted commercial markets on the other ( Caulkins et al., 2015 ;Decorte et al., 2017;Decorte, Lenton, & Wilkins, 2020b ;Queirolo, 2020 ;Queirolo et al., 2016 ;Wilkins, 2018 ). These include home cultivation, social clubs, not-for-profit community trusts, government monopolies, heavily regulated markets and "for-benefit " companies ( Caulkins & Kilmer, 2016 ;Decorte et al., 2017 ;Fischer et al., 2020a ;Queirolo, 2020 ;Queirolo et al., 2016 ;Wilkins, 2018, Barry et al., 2016, Barry et al., 2018, Cerda & Kilmer, 2017. However, these middle ground approaches have received less mainstream media and political attention than the familiar alcoholstyle regimes implemented in U.S. states. This is the case even when middle ground approaches have been implemented for many years. For example, Uruguay passed a largely non-commercial supply system for recreational cannabis use in 2013, although it has taken some time to establish the registration system and pharmacy retail supply ( Queirolo, 2020 ).
Multi-Criteria Decision Analysis (MCDA) is a group decision making method used to help resolve complex and controversial public policy decisions characterized by uncertainty and incomplete information ( Department for Communities & Local Government, 2009 ). For example, it has been used to assess nuclear waste management and the riskbenefit ratio of prescription drugs ( Hughes et al., 2016 ;Morton et al., 2009 ). MCDA has previously been employed in the drug policy field to assess the harms of different drug types and different policy approaches to drug access and control ( Bonomo et al., 2019 ;Nutt et al., 2010 ;Rogeberg et al., 2018 ;van Amsterdam et al., 2015 ). MCDA allows decision-makers to break down complex problems into simpler and essential components, facilitates the combining and integrating of the knowledge of many decision-makers with a range of expertise, and supports collaborative problem-solving and consensus building, particularly where there is conflicting experiences or perspectives . While MCDA has been utilized in the drug policy field, it is also acknowledged that the outcome of an MCDA analysis is influenced by the composition of the decision-making group . MCDA of drug policies to date have largely involved small groups of health and academic experts. As a result, there have been calls for broadening these decision-making groups to provide wider perspectives on policy reform Rolles & Measham, 2011 ).The aim of this study was to inform the debate leading up to the New Zealand cannabis referendum by conducting an MCDA with a range of key national stakeholders to identify their preferences for cannabis reform outcomes and in doing so rank cannabis policy reform options.

Methodology
A central aim of our study was to involve a broader range of stakeholders in the MCDA assessment as suggested in recent critiques of MCDA. While these stakeholders possessed practical experience of different aspects of cannabis outcomes, they did not necessarily have detailed knowledge of the evidence related to the outcomes of cannabis policy reforms implemented overseas, or the wider research literature on the health and social risks of cannabis use. They also had limited time to digest this research literature and attend a multi-day decision-making workshop [previous MCDA workshops on drug policy have involved two-day meetings (e.g., Rogeberg et al., 2018 ;van Amsterdam et al., 2015 )]. Consequently, to ensure as many stakeholders as possible were able to attend the workshop, much of the development of the MCDA model was completed in advance via a collaborative process with domestic and international cannabis policy experts (and the 1000minds TM MCDA software developers). This preliminary work allowed the MCDA workshop to be completed in a single day as stakeholder participants were able to go straight to normative trade-offs assessing projected outcomes from different reform options. From the perspective of the stakeholders, the MCDA was therefore largely a group discrete choice experiment where they were asked to make comparisons between two alternative projected outcomes that differed in a specified way on a key criterion (e.g., Fig. 1 ) and reach consensus on the preferred outcome.
Developing an MCDA model for evaluating cannabis policy reform options entails four broad stages: (1) defining the policy reform options, (2) specifying the decision-making criteria and their outcome levels, (3) determining the relative weights for the different decision-making criteria and related levels, and (4) ranking the policy reform options based on the relative weighting of the criteria and levels .

Defining the policy reform options
The cannabis policy reform options in our MCDA included a number of "middle ground " approaches adapted to the New Zealand context

Table 1
New Zealand cannabis policy reform options.

Reform option Brief description
Prohibition with criminal penalties Current New Zealand policy. Use and supply is a criminal offence with the possibility of conviction and in rare circumstances imprisonment Prohibition with fines ( "like a speeding ticket ") Commonly known as "decriminalization'. Use and growing a small number of plants for personal use is a civil offence ordinarily resulting in a fine.

Cannabis social clubs and home grow
Home growing of small number of plants for personal use and social sharing is legal (no sales). Individual plant provisions can be combined for members of a social club.

Government monopoly and home grow
Cannabis is sold from government owned shops. There are age limits, no advertising and promotion, high excise taxes, restrictions on potency, plain packaging, mandated health warnings and limited number of outlets. A set number of plants are permitted to be grown at home for personal use.

Not-for-profit trusts with home grow
Community elected trusts sell cannabis and are required to spend a proportion of the revenue on community services, such as sport and arts. There are age limits, restricted advertising and promotion, high excise taxes, packaging restrictions, mandated health warnings and limits on number of outlets. A set number of plants are permitted to be grown at home for personal use.

Strict market regulation ( "like tobacco ") and home grow
Cannabis is commercially sold with a high excise tax, age restrictions, no advertising or promotion, plain packaging, some restrictions on potency, mandated health warnings, and sales from licensed sellers only. A set number of plants are permitted to be grown at home for personal use.

Light market regulation ( "like alcohol ") and home grow
There is largely a commercial cannabis market with limited restrictions on advertising and promotion, a small excise tax, age restrictions and some licensing and product labelling requirements. A set number of plants are permitted to be grown at home for personal use.

Unrestricted market ( "like soft drinks ") and home grow
There is a commercial market for cannabis with only generic consumer goods regulation around product safety and retail selling. There is no excise tax, restrictions on advertising and promotion, restrictions on potency, limited labelling requirements, and no limits on number of outlets. A set number of plants are permitted to be grown at home.
(see Table 1 ). All reform options presented included the "home cultivation of cannabis " as the New Zealand Government considered this a core aspect of all reform options ( Office of the Ministry of Justice, 2019 ). It is important to note that the New Zealand cannabis referendum debate was specifically concerned with the legal status of cannabis use and supply, not wider policy responses and settings in relation to cannabis harm, such as level of spending on treatment and prevention, although these factors are naturally part of the discussion of different reform approaches as reflected in our MCDA criteria.

Specifying the decision-making criteria and outcome levels
The two cannabis policy researchers based in New Zealand (CW and MR) initially identified the five key cannabis policy reform decision making criteria (i.e., health and social harm, illegal market size, reducing arrests, earning tax income, expanding treatment services) and related levels, by drawing on New Zealand Government documents discussing the overarching policy objectives of cannabis reform ( Office of the Ministry of Justice, 2019; Parliamentary Counsel, 2020 ), previous New Zealand reports estimating cannabis harms and assessing law reform options, available New Zealand statistics on cannabis use and harm, and the policy evaluation literature from overseas cannabis reforms (McFadden Consultancy, 2016 ; New Zealand Institute of Economic Research ( NZIER), 2020 ; Sense Partners, 2018 ).

Criteria 1 -Health and social harm
The baseline figure of the "health and social harm " of cannabis use in New Zealand used in the MCDA model of $1.3 billion per year (NZD) was taken from the 2016 New Zealand Drug Harm Index (NZ-DHI) commissioned by the New Zealand Ministry of Health (McFadden Consultancy, 2016 ). The NZ-DHI provides a comprehensive evaluation of the costs of cannabis use in New Zealand, including the personal and community harm of use and related costs of interventions (e.g., hospitalizations, treatment & counselling, drug prevention). The NZ-DHI was written in collaboration with the New Zealand Ministry of Health and other government agencies who provided national statistics on drug use and harm for the report. The methodology was based on the earlier Australian Drug Harm Index methodology (originally commissioned in 2001 and then subsequently updated in 2003, 2006, and 2008). This methodology was adapted to New Zealand in 2008 and updated in 2016.
The NZ-DHI authors acknowledge a number of gaps and limitations of official statistics which require simplifying assumptions to complete the estimates, including difficulty assigning harms from poly-drug use to individual drug types (e.g., cannabis and methamphetamine co-use), quantifying the harm of drug dependency on families, and establishing the causal role of drug use in crime (McFadden Consultancy, 2016 ). A limitation specific to cannabis was New Zealand health statistics did not separate natural cannabis from synthetic cannabinoids, recording them together as a combined "cannabinoid " category (Mc-Fadden Consultancy, 2016 ). Despite these issues, the NZ-DHI provides the official estimates of the social harm of drug use used in New Zealand.
Preliminary evaluations of the U.S. state "for profit " cannabis legalization reforms have raised concerns about declining prices, highpotency products, marketing and promotion practices, and industry influence over regulation development ( Caulkins & Kilmer, 2016 ; Colorado Department of Health & Environment, 2021 ; Davenport, 2019 ; European Monitoring Centre for Drugs and Drug Addiction (EMCDDA), 2020 ; Hall et al., 2019 ;Hunt & Pacula, 2017 ;Smart et al., 2017 ;Subritzky et al., 2016 ;, Orens at al., 2018 and some of these concerns have been reflected in subsequent post legalization findings from both Canada and the US ( Hammond et al., 2021, Mahamad et al., 2020Rup et al., 2020 ).
There is evidence from U.S. national data of increasing adult use, frequency of use and cannabis use disorder (CUD) in legal recreational cannabis states ( Cerdá et al., 2020 ). It has subsequently been noted that only four US states (i.e., Colorado, Washington, Oregon, and Alaska) had established cannabis retail outlets during the period of this study (i.e., 2012 to 2016), and those had only been open for a relatively short period of time, suggesting that the longer-term impact of commercial legalization and higher potency legal products on harm is yet to be fully experienced ( EMCDDA, 2020 ). Evaluations of the impact of cannabis legalisation on youth in the early U.S. recreational legalization states (i.e. Colorado, Oregon, Washington State) who implemented largely commercial regimes have found mixed results with meta-analyses suggesting small increases in use ( Melchior et al., 2019, Smart & Pacula, 2019, Dilley et al., 2019. The longer timeframe of cannabis legalization in Uruguay has allowed more evaluation, utilizing neighboring countries as control groups, and has found only minimal impacts on adolescent use and perceived risk of use to date Laqueur et al., 2020 ;Schleimer et al., 2019 ).
Drawing on the extensive research literature of effective public health regulation of alcohol and tobacco markets (e.g. Gartner & Hall, 2020 ;Pacula et al., 2014 ;Stockwell et al., 2020Subritzky & Lenton, 2016, we assumed that more restrictive regulatory approaches, such as government monopoly, non-commercial supply and strict market regulation, are likely to be more effective at reducing the health and social harms of legal cannabis use than lightly regulated commercial markets. Proponents of cannabis legalization have argued that a legal cannabis market could reduce the overall harm of cannabis, even if the total number of users increase, by improving the safety of legal cannabis products (e.g., via manufacture and potency controls) and by reducing barriers related to help seeking for dependent users. This may well be the case, but at the time of writing this article, we do not yet have the data, either from overseas jurisdictions or New Zealand, to support this argument.

Criteria 2 -Reduce arrests
All of the legalization options in our MCDA model projected large reductions in cannabis arrests following legalization regardless of the strictness of the regulatory framework. This is based on the understanding that following legalization cannabis related arrests would be limited to underage use, illegal supply, public disorder, and drug driving. This assumption is supported by evidence from U.S. states that have legalized cannabis where substantial reductions in arrest rates for cannabis have been achieved (e.g., arrest rates in Alaska, Washington, D.C., and Oregon have fallen by 90%) (Adinoff and Reiman 2019). Drawing on official New Zealand Police apprehension statistics, we used a baseline figure of 10,500 arrests per year for cannabis under prohibition in New Zealand ( Statistics NZ, 2016 ). Note, this number includes all situations where police have contact with cannabis offenders regardless of whether this results in a formal arrest, prosecution, or conviction (i.e., some receive informal warnings, pre-charge warnings or diversion) ( Wilkins & Sweetsur, 2012 ).

Criteria 3 -Reduce the size of the illegal market
All the legalization options in our MCDA projected significant reductions in the illegal cannabis market, while not entirely eliminating the illegal trade. Again, this reflects experience from legal cannabis jurisdictions in the U.S. and Canada where the illegal cannabis trade has been much reduced, but nonetheless has persisted [e.g., two years after legalization, 37% of Canadian users reported having purchased cannabis from an illegal source ( Government of Canada, 2021 )]. The NZ-DHI estimated the annual revenue of the illegal market for cannabis to be $548M based on aggregate cannabis consumption (McFadden Consultancy, 2016 ).

Criteria 4 -Earn tax revenue
The tax earnings figures in our MCDA assumed that the more heavily regulated legal market options (e.g., "government monopoly', "not-forprofit " and "restricted markets') are likely to impose higher excise rates and other taxes and hence collect more total tax revenue. Estimates of the magnitude of the economic benefits of a legal cannabis sector in New Zealand have varied. For example, an economic consultancy firm estimated the government revenues from cannabis legalization in New Zealand would be $191-$249M (NZ$) ( Sense Partners, 2018 ). The New Zealand Institute of Economic Research (NZIER), projected annual tax revenues from legal cannabis sales in New Zealand of $490M (New Zealand Institute of Economic Research (NZIER) 2020). Most recently, the economic consultants BERL has advised the New Zealand Ministry of Justice that a legal cannabis sector in New Zealand would generate $923M annually in taxation and licensing fees ( BERL, 2019 ). Our projections are based on the actual tax earnings reported from legal cannabis sales in Colorado (i.e. $200-300M), which has a comparable population to New Zealand ( Colorado Department of Health & Environment, 2021 ). It is also important to note these are estimates of gross tax income and public expenditure is required to implement, regulate and enforce legal cannabis markets (e.g., the Colorado Marijuana Enforcement Division's annual budget in 2020-21 is $25M USD).

Criteria 5 -Expanding drug treatment
The most restrictive legalization options (e.g., "government monopoly ", "not-for-profit " and heavily regulated markets) were projected to invest more in drug treatment capacity as they would collect more cannabis excise tax, although ultimately these remain discretionary policy decisions. The estimate of current unmet demand for cannabis treatment in New Zealand came from the New Zealand National Committee for Addiction Treatment (NCAT) who calculated that only 50% of the people who want help for substance abuse are currently receiving it due to service capacity limits ( Sense Partners, 2018 ). The New Zealand Drug Foundation has called for the doubling of the treatment budget to meet this unmet demand paid for by taxes from cannabis legalization ( Sense Partners, 2018 ).

Review of criteria and levels by international cannabis policy experts
Five international cannabis policy experts (Tom Decorte, Benedikt Fischer, Beau Kilmer, Simon Lenton, Rosario Queirolo) with expertise and experience of the implementation of cannabis reforms in their own countries, were invited to review the MCDA model, firstly individually in advance of the MCDA workshop, and then collaboratively as a group during a face-to-face meeting one day before the workshop. This involved reviewing the policy decision making criteria and the projected outcomes assigned to each policy reform option. The final model comprised of the eight cannabis policy reform options, and five decision making criteria and projected outcome levels discussed above ( Table 2 ).

Determining the weights for the different decision-making criteria and related levels
The PAPRIKA MCDA method The MCDA model was developed using the 1000minds TM MCDA software platform (www.1000minds.com) which utilizes the PAPRIKA (P otentially A ll P airwise R an K ings of all possible A lternatives) method (for technical details, see Hansen & Ombler, 2008 ). Briefly, the PA-PRIKA method asks decision makers to express their preferences for decision making criteria and projected outcome levels by ranking undominated outcomes defined on just two criteria at-a-time (and assuming all other criteria's categories are pairwise identical) (see Fig. 1 ). The trade-off questions are repeated with different pairs of criteria and outcome scenarios. Each time the decision-maker answers a trade-off question, all other scenario pairs that can be pairwise ranked by applying the logical property of 'transitivity' are identified and eliminated ( Wesselbaum, 2022 ). Simulations have shown that if the decision-maker explicitly ranks undominated pairs defined on just two criteria at-a-time, the overall ranking of alternatives produced by the value model is very highly correlated with the true ranking of all possible trade-offs ( Hansen & Ombler, 2008 ).

MCDA workshop
Forty-five key national stakeholders from government agencies, health, NGO, law enforcement, academia, pro-legalization activism, and industry attended the MCDA workshop on 16 th October 2019 ( Table  A.1 ). These participants represented all the NZ government agencies involved in the cannabis reform debate (i.e., Ministries of Justice, Health, Social Development, Police, Health Promotion Agency, Customs and National Drug Intelligence Bureau), a number of health providers and professional societies (e.g., District Health Boards, Mental Health Foundation, New Zealand Nurses Organization, Mental health, addiction and disability workforce development), NGO's concerned with drug policy (e.g., NZ Drug Foundation, Public Health Association of New Zealand), cannabis industry (i.e., Hikurangi Enterprises Ltd, New Zealand Medicinal Cannabis Council), cannabis activism (i.e., NORML NZ) and academics from the addictions, medicine, public health and criminology fields. A leading anti-legalization advocacy group was also invited to the workshop but declined to attend. The workshop included representatives from the Ministry of Justice and Ministry of Health team (3 peo-ple) tasked with developing the referendum proposal, but they recused themselves from voting due to conflict of interest (leaving a voting group of 42). The meeting was facilitated by the 1000minds TM software developers.
In the morning of the MCDA workshop, the five invited international cannabis policy researchers each made seminar presentations on the cannabis reform options implemented in their countries (i.e., Simon Lenton -Australian state decriminalization; Tom Decorte -Spain and Belgium social clubs; Rosario Queirolo -Uruguayan state-controlled market; Benedikt Fischer -Canadian strictly regulated market; Beau Kilmer -U.S. lightly regulated legal market). Marta Rychert presented on the "not-for-profit " trust model for cannabis based on working examples of alcohol licensing trusts and gaming machine gambling trusts in New Zealand . The morning seminars introduced the workshop participants to the policy reform options that were part of the MCDA model, including summarizes of known outcomes and outstanding gaps in data and understanding of consequences. The international presenters were encouraged to group the outcome data around the MCDA model criteria (i.e., "health and social harm ", "number of arrests ", "reducing the illegal market ", "expanding drug treatment " and "earning tax "). Each seminar was followed by an extended Q&A session where participants were encouraged to ask further questions about the reform approach and related outcomes and unknowns.
The MCDA decision-making workshop was conducted in the afternoon and involved 42 participants who were invited to anonymously vote on each trade-off question using their mobile phone (i.e., Option A, Option B, and Option C "they are equal ") (see Fig. 1 ). The results of each vote were displayed on a large screen at the front of the meeting room. If one of the three possible answers received more than 50% of the votes it was accepted as the final answer. If there was no such 'majority winner' after the vote, the participants discussed the question and voted again. If necessary, this process was repeated until a majority winner emerged. Each time the participants answered a question (after voting), the method adapted the selection of the next question based on their preceding answers (always one whose answer was not implied by earlier answers). A total of 27 trade-offs questions were answered by the stakeholder group during the MCDA workshop. Based on these answers, the PAPRIKA method calculated the final weights for the criteria and levels. Note: The bolded values represent the relative weights of the criteria overall (i.e., bolded values sum to 100%).

Ranking the policy reform options
From the participants' answers to the above trade-off questions, the 1000Minds software calculated the relative weights on the criteria and levels and from these ranked the policy options (for technical details, see Hansen & Ombler, 2008 ). The final ranking of the policy options was shown to the group, and this was followed by a group discussion of the rankings. Table 3 presents the relative weights of the five criteria and the levels within each criterion. Controlling the "health and social harm " of cannabis use, and "reducing cannabis arrests " received the highest relative weightings from workshop participants. In contrast, earning "tax and community funding " from cannabis sales received the lowest weighting. Table 4 illustrates the relative importance of the different criteria and levels. For example, a small reduction in the "health and social harm " from cannabis use was judged to be over 20 times more important than a very large revenue from cannabis sales. A small reduction in the "health and social harm " from cannabis use was also considered to be over three times more important than a very large reduction in the illegal market for cannabis. Expanding drug treatment for cannabis use was considered to be four times more important than a very large revenue from cannabis sales. Table 5 presents the ranked cannabis policy options with their respective performances on the five decision-making criteria. "Government monopoly " received the highest total score (81%), followed by "not-for-profit trusts " (73%) and "strict market regulation " (65%). "Government monopoly " scored higher than the second ranked option of "not-for-profit " trust almost entirely due to its projected better performance on the highly weighted "health and social harm " criteria (i.e., small decrease vs. no change) (45.9% vs. 40.5%), and slightly better performance on "drug treatment provision " (8.2% vs. 6.8%) and "tax revenue " (2.0% vs. 1.5%). The "strict market " option scored lower entirely because it scored relatively lower on controlling the "health and social harm " of cannabis use (31.6%) compared to the "government monopoly " and "not-for-profit " options. The "lightly regulated market " option scored lower due to its relatively poor performance on controlling the "health and social harm " of cannabis use (17.5%), while performing better on reducing arrests (23.0%) and the illegal market (11.7%). Prohibition with civil fines (i.e., commonly known as decriminalization) scored relatively well on controlling the "health and social harm " of cannabis use (31.6%), but had relatively low scores on reducing arrests (7.6%) and the illegal market (5.1%). It also provided no tax revenue (0.0%) and related improvement of drug treatment services (2.8%). Cannabis social clubs scored exactly the same as prohibition with civil fines (i.e., performing relatively well at controlling "health and social harms ", but having limited impact on reducing arrests and the illegal market and offering little improvement in treatment capacity). The unrestricted market option (i.e., regulated like soda drinks), scored highest at reducing arrests (30.6%) and the illegal market (13.3%), as it involves only generic consumer goods regulation and hence involved very few regulations that could result in users' arrest or opportunities where illegal supply could out compete legal sales, but achieved the lowest with respect to controlling "health and social harm " of cannabis use (0.0%). "Prohibition with criminal penalties " received the lowest ranking as, although it scored relatively well at controlling the health and social harm (40.5%), largely due to containing cannabis use, it resulted in no change in the number of cannabis arrests (0.0%), drug treatment capacity (0.0%), or the size of the illegal market (0.0%).

Discussion
The New Zealand national stakeholders ranked the strictly regulated legal cannabis market reform options (i.e., "government monopoly ", "not-for-profit ", "strict market like tobacco ") higher than the prohibition-based options ( "prohibition ", "decriminalization ") as the former were projected to substantially reducing cannabis arrests and the size of the illegal market, and also contribute to more funding for drug treatment, while minimizing any increase in the health and social harm of cannabis use. The strictly regulated legal cannabis market reform options were also ranked higher than the more commercial legal market options ( "lightly regulated ", "unrestricted " market) largely because they were projected to provide greater control of the social and health harms of cannabis use, while also offering substantial reductions in cannabis arrests and the illegal market. Rogeberg et al. (2018) employed an MCDA process with experts to evaluate policy options for cannabis and also found state control (of production and sale) to be the preferred reform option, and conversely prohibition to be the lowest ranked option. While Rogeberg et al.'s (2018) decision making criteria were not directly comparable with ours (and used a different MCDA software package -Catalyze Hiview 3®), controlling health, social and public impacts were influential in favoring state control over the free market option.
The stakeholders also considered reducing arrests for cannabis a relatively high priority outcome of reform. This may reflect long standing concern in New Zealand about the higher conviction and imprisonment rates for M āori (the indigenous people of Aotearoa/New Zealand) compared to other ethnicities, and the fact that M āori are three times more likely to be arrested and convicted for cannabis offences than non-M āori ( Fergusson et al., 2003 ; Office of the Ministry of Justice, 2019 ). There may also be a wider moral view that a criminal conviction is a disproportionately severe penalty for minor cannabis  offending. Rogeberg et al.'s (2018) MCDA also found experts placed a high relative value on reducing the criminalization of cannabis users and the illegal market as important criterion in evaluating reform options. Sheehan et al. (2021) examined racial differences in cannabis arrest rates among adults and youths after state-wide cannabis decriminalization, legalization and no policy change in the U.S. from 2000 to 2019, and found there was a substantial decrease in absolute cannabis arrests among Black adults as an immediate response to decriminalization, but the relative racial disparity remained, perhaps reflecting the greater difficulties of poorer individuals at paying civil fines. In U.S. states that enacted cannabis legalization, there was a clear reduction in arrests among both Black and White adults as well as a reduction in the racial disparity. However, as the reduction trend began before the implementation of cannabis legalization, this reduction was likely not solely attributable to legalization. The stakeholders placed the lowest relative importance on "tax earnings " as a reform outcome, even when the tax revenue presented involved seemingly substantial sums (i.e., $200-$350M per year NZD). Cannabis industry and business lobby groups often emphasize the potential tax earnings and economic development opportunities of cannabis legalisation as compelling reasons for law reform, although estimates of these financial returns have varied considerable between New Zealand commentators (e.g., Sense Partners, 2018, NZIER, 2020, BERL, 2019 and they are rarely offset with a discussion of additional health and social service spending which may emerge from increased legal cannabis use and harm. Similarly, the experts in Rogeberg et al.'s (2018) MCDA placed a relative low value on earning state revenue and reducing public spending on law enforcement as a factor influencing the cannabis reform option.
The stakeholders also gave the reduction in the illegal market a relatively low weighting, even when substantial reductions in illegal market activity were being projected (i.e. $100-$250M or 50%-80% reduction of the illegal market). This is surprising as reducing the illegal market is often presented as a critical objective of cannabis legalization ( Oregon Liquor Control Commission, 2019 ), and indeed, higher legal cannabis taxes, potency controls and retail restrictions are sometimes opposed on the basis these will handicap the legal cannabis sector from competing against the illegal trade (New Zealand Institute of Economic Research (NZIER) 2020). It may be the case that the low priority given to reducing the illegal market for cannabis by the New Zealand stakeholders reflects a belief that much of the illegal cannabis market in New Zealand is relatively innocuous, and indeed, may provide much needed seasonal income in some economically depressed rural regions ( Walker et al., 1998 ). In contrast, the argument about reducing illegal market activity has been particularly relevant for other countries, such as Uruguay and Mexico, where cannabis legalization has been proposed as a means to reduce levels of violence and instability related to wider narcotrafficking ( Queirolo et al., 2019 ).
Increasing drug treatment capacity also received a low relative importance by the stakeholders. This may reflect a perception that only a minority of cannabis users are ever likely to require drug treatment. This perception is largely supported by a recent systematic review that found overall 22% of cannabis users experienced CUD (range 10-30%), with 13% suffering the more serious cannabis dependence (largely under prohibition regimes) ( Leung et al., 2020 ).

Limitations
We acknowledge a number of limitations with our MCDA. First, as underlined by the domestic and international cannabis policy experts at the MCDA workshop, considerable gaps remain in relation to understanding of full consequences of the different cannabis law reform options implemented overseas, particularly with regard to the recent legalization of recreational cannabis in select jurisdictions of the Americas. There is further uncertainty concerning how these overseas reforms might be translated and implemented in New Zealand with its distinct socio-cultural, political and economic ecology ( Fischer et al., 2020a ;Wilkins & Rychert, 2020b ). The projected outcome levels of our MCDA model are designed to communicate broadly sketched plausible scenarios rather than precise estimates, and this limitation was clearly communicated to the stakeholders involved in the MCDA exercise both during the morning seminars and afternoon MCDA workshop. While New Zealand boasts several world renown longitudinal studies of city based birth cohorts that have tracked the lifetime impact of cannabis use, there are some significant gaps in national population level statistical data on cannabis use and harm ( BERL, 2019 ). For example, the last population level data on frequency of cannabis use, dependency and related harms was collected in 2007/08 ( BERL, 2019 ). The handful of existing attempts to model outcomes from cannabis law reform in New Zealand have faced significant domestic and international data limitations, most importantly the limited data on cannabis legalization implemented overseas ( Sense Partners, 2018 as critiqued in Wilkins et al., 2019 ).
As described earlier, we utilized the NZ-DHI estimates of the health and social harm of cannabis as the starting point for our MCDA model. While this report had access to national statistics on cannabis use and harm, there remained gaps in data and, as a result, simplifying assumptions had to be made to complete the estimates (McFadden Consultancy, 2016 ). A number of these assumptions are controversial, such as the decision to split the remaining 64 drug related deaths that could not be assigned to opioid overdose (i.e., 10 deaths) or psychedelics (1 death) evenly between amphetamines and cannabinoids (i.e., 32 deaths each). This decision may be linked to the inability at the time to separate natural cannabis from synthetic cannabinoids in official statistics, as synthetic cannabinoids have been linked to a number of overdoses in New Zealand ( Wilkins & Rychert, 2018 ). Furthermore, all the acquisitive crime committed by people with cannabis dependence is assigned to cannabis harm without determining the causal role cannabis use may have played in motivating these offences.
Second, as noted at the beginning of this paper, one of the recommendations from the recent MCDA of drug policy is to broaden the decisionmaking group Rolles & Measham, 2011 ). Our MCDA achieved this objective to an extent by including officials from a range of government agencies and NGO workers concerned with drug and health issues, including M āori, cannabis legalization activists, medicinal cannabis industry, drug treatment and mental health, and law enforcement organizations. This could be taken further to include those most affected by the current cannabis prohibition and reform options, including youth, M āori, Pasifika, parents of adolescents, and those living in high deprivation communities. To ensure the views of these affected groups are not swamped by the majority view in a combined workshop, a series of MCDA workshops could be conducted with each entirely compromising members of one of the affected groups (e.g., a M āori MCDA workshop). While we were not able to convince the main anti-cannabis legalization activist group to attend the MCDA, we had a range of government and NGO stakeholder participants who may have been ambivalent, or even opposed, legalization (e.g., police, NGO from mental health, health practitioners, and drug treatment). The refusal of the main anti-leglization group to attend may reflect their view, wrongly held in this case, that the forum is not interested in their perspective or values. Indeed, the rejection of the NZ cannabis referendum has in part been explained by a failure to engage with conservative right-wing voters about how cannabis legalization may align with their political values (such as promoting personal freedom and responsibility, eliminating inefficient government spending, and reducing government power over citizens) , Wilkins et al, 2022. Anti-legalization groups may well have supported the current prohibition approach and more investment in drug treatment and prevention. This option was not part of our MCDA policy options as we were primarily aiming to inform the debate around the cannabis referendum which was specifically concerned with the legal status of cannabis use and supply, not wider policy settings addressing cannabis harm under the current prohibition, such as increasing the level of funding of treatment and prevention.
Third, MCDA is an instrumental group decision making tool that focuses on tangible outcomes of policy decisions. It does not directly address or debate the moral basis of policy positions. Moral views of cannabis use were found to play an important part in how people voted in the NZ cannabis referendum  and likely influenced how stakeholders voted in the MCDA trade-offs. Yet, as outlined earlier in this paper, the purpose of MCDA is to facilitate group decision making concerning controversial policy issues by asking participants to consider tangible trade-offs in outcomes to reach pragmatic compromises. The closeness of the final NZ cannabis referendum result (48.4% for, vs. 50.7% against) illustrated both significant support for reform and concern about the specific reform proposal put forward for the referendum vote (i.e., the Cannabis Legalization and Control Bill or CLCB). The CLCB most closely resembled the "strict market like tobacco " option in our MCDA. Our MCDA results suggest a higher stakeholder support for two even more restrictive legal market options, "government monopoly " and "not-for-profit " trusts.

Conclusions
The MCDA of cannabis policy options completed by a range of New Zealand stakeholders underlined the high importance placed on control-ling the health and social harm of cannabis use, while also reducing the number of cannabis arrests and the size of the illegal market. Strictly regulated legal market options were ranked highest as they were projected to minimize the health and social harms of cannabis use via strong public health regulation, while also substantially reducing arrests and the size of the illegal market by legalizing the use and sale of cannabis.

Role of Funding Source
This research was supported by a Massey University Strategic Investment Fund (SIF) grant and a Marsden Royal Society of New Zealand grant ( MFP-MAU1813 )

Contributors
All authors materially participated in the research and the writing and revisions of the manuscript. All authors have contributed to and agreed with the final manuscript.

Declarations of Interest
No conflict to declare.