The reliability of illicit drug markets has important implications for any proposed legalization or decriminalization of illicit drugs. The level of violence and cheating in the illicit drugs market will define the magnitude of the social benefits from the legalization and regulation of the illicit market. The reliability of the illicit drugs market also has important implications for any decriminalisation of drugs. A common criticism of decriminalisation is it leaves a ‘harmful’ illicit market intact (see Drug Policy Forum Trust, 1998). If the illicit market is actually relatively reliable, policy makers are able to respond to any public demand for lighter penalties for drug use without any major concern about the role the illicit market plays in violence and crime. This thesis uses property rights economics to model the level of cheating between exchange parties in retail illicit drug transactions. Illicit drug transactions are generally assumed to be unreliable because exchange parties cannot call on the police and courts to enforce the terms of transactions. The reliability of illicit drug transactions is essentially a problem of social cooperation in the absence of third party enforcement. The scope of the model is restricted to exchange parties during retail drug transactions. It does not explain all the cheating and violence which might occur in an illicit drugs market. The model therefore can only partially address the drug control policy issues related to the reliability of illicit drug markets. The decision to cheat on a retail illicit drug transaction is initially modelled as a two person prisoners dilemma game. The “ongoing interaction” solution to the prisoners dilemma is adapted to retail illicit drug transactions by introducing search costs. The decision to cheat involves comparing the expected one-off benefit of cheating with the search costs of locating a new exchange partner. The one-off benefit of cheating is constrained by the low financial value of retail drug transactions. Search costs are high in the illicit drugs market because exchange parties are secretive and suspicious, there is no advertising, information directories, or prominent store locations, to assist exchange parties with finding exchange partners, and search activity can result in arrest and victimisation. The model indicates there are often strong incentives to voluntarily honour the terms of retail illicit drug transactions. Search costs also make it costly for victims of cheating to quit exchange relationships. The decision to quit an exchange relationship involves comparing the search costs of finding a new exchange partner with the losses from continuing to transact with an existing cheating exchange partner. It will usually pay to quit an exchange relationship after cheating because the perpetuity of losses from cheating will generally be greater than the search costs of finding a new exchange partner. However, if the costs of cheating are very small, for example in the case of quality fraud, it maybe rational to persist with a cheating exchange partner - at least in the short term. The assertions of the model are tested in the retail cannabis market in New Zealand. The central assertions are product fraud, robbery, and assault, will be rare, while quality fraud will be relatively common. Sellers will not cheat buyers because transactions are very profitable, reputation is important, new customers are costly to find, and sellers wish to dispose of stocks of drugs as quickly as possible. Data collected in a postal questionnaire of members of the Aotearoa Legalise Cannabis Party (ALCP) was broadly consistent with the model’s assertions (n = 357). A range of statistical analysis was used to test the model’s assertions. Most buyers (78.5%) had never suffered from product fraud. The overwhelming majority of buyers (91-96%) had never been robbed or assaulted. The overwhelming majority of buyers (85-98%) experienced product fraud, robbery, and assault, less than 0.1 times per 100 transactions (less than 1 incident in 1000 transactions). In contrast, 81% of buyers had suffered from quality fraud. Many buyers (41%) experienced quality fraud more than 5 times per 100 transactions. Overall, the average level of cheating (per 100 people) in the ALCP sample was higher than the average level of cheating (per 100 people) in New Zealand as a whole. The important exception was the incidence of assault. There were one-third less assaults (per 100 people) in the ALCP sample than in New Zealand as a whole. The level of robbery (per 100 people) in the ALCP sample was comparable to the level of robbery in New Zealand as a whole. In contrast, there were approximately 100 times more incidents of quality fraud (per 100 people) in the ALCP sample than New Zealand as a whole. The highly positively skewed nature of the data (ie. many people never been cheated) meant the median provided a better description of the sample than the mean. A large proportion of buyers (79-96%) had levels of product fraud, robbery, and assault (per 100 people) below the levels of these offences for New Zealand as a whole. Only a very small proportion of buyers (2-16%) had levels of product fraud, robbery, and assault, above the range of these offences (per 100 people) for New Zealand. In contrast, a large proportion of buyers (79%) had levels of quality fraud (per 100 people) above the level of fraud (per 100 people) for New Zealand as a whole. Sellers were provided with eight reasons why they did not cheat buyers and asked to score the importance of each using the scale, 1 (“not important”) to 7 (“very important”). Four of the reasons were from the model and were based on the profit motive and the need to minimise search costs. They were, “money,” “selling stocks quickly,” “reputation,” and “limited customers.” The other four reasons were traditional explanations for reliable illicit drug exchange based on fear of physical retaliation and the desire to minimise the risk of arrest. They were, “fear of physical retaliation,” “fear of attracting the police,” “fear of betrayal to the police,” and “cheating may result in committing a more serious crime.” Between 61-65% of sellers thought the four reasons used in the model were very important reasons not to cheat buyers (ie. importance score of 6-7). In contrast, only between 33-44% of sellers thought the four traditional reasons were very important (ie. importance score of 6-7). When ranked according to average score of importance the four reasons used in the model were all higher than the traditional reasons. The reasons used in the model occupied the first four rankings, “money” (5.5), “selling stocks” (5.4), “reputation” (5.3), and “limited customers” (5.1). “Fear of physical retaliation” received the lowest average score of importance of the eight reasons provided (3.8). The other three traditional reasons also received relatively low average scores of importance, “fear of betrayal to police” (3.9), “fear of attracting the police” (4.2), and “cheating may result in committing a more serious crime” (4.4). The policy implications that can be drawn from the thesis are extremely tentative. Additional modelling of the illicit drugs market and a greater understanding of the wider consequences of liberalising drug control laws is required before firm policy recommendations can be made. The model and research completed in this thesis suggest the legalization and regulation of the cannabis market will only have a small impact on lowering violence and cheating in New Zealand. The only significant impact the legalization of cannabis would have on levels of victimisation would be to reduce quality fraud. The results from the thesis could be used to support the case for the decriminalisation of cannabis in New Zealand. The thesis suggests the illicit cannabis market does not contribute significantly to violence and victimisation in New Zealand. Decriminalisation could be used to meet any public demand for the reduction in the legal penalties for cannabis use without any major concern about leaving a dangerous illicit market intact.