Panel Paper: Cannabis Legalization and Traffic Safety in Uruguay: Insights from a Novel Synthetic Control Approach 

Saturday, November 10, 2018
8226 - Lobby Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Steven Davenport, RAND Corporation

Steven Davenport, Pardee RAND Graduate School

Beau Kilmer, RAND Drug Policy Research Center

Motivation. Cannabis policy is a serious topic of discussion in Europe and the Americas. In 2013, Uruguay became the first country in the world to legalize cannabis for nonmedical purposes. Unlike the commercial for-profit model adopted in eight U.S. states, Uruguay has adopted a middle-ground approach which requires residents 18 and older to choose one of three supply mechanisms for legal cannabis: grow at home, join a cannabis social club, or purchase from a pharmacy (Pardo, 2014). In Uruguay and elsewhere, one of the major policy questions about legalization is the effect on traffic safety. While the bulk of the research suggests that driving under the influence of alcohol is more dangerous than cannabis, it also suggests that driving under the influence of cannabis is worse than driving sober. Since cannabis legalization may influence both alcohol and cannabis consumption, objective assessments of legalization on traffic safety will focus on outcomes such as total crashes or total traffic fatalities—not just those involving drivers testing positive for THC, a poor measure of impairment. Methodology. With nearly half of Uruguayans living in one city (Montevideo), this complicates efforts to leverage within-country variation in registered home growers, cannabis social clubs and pharmacies selling cannabis. We propose to use a novel synthetic control approach using data from Montevideo and 11 jurisdictions in Chile: Santiago, Valparaíso, Concepción, Arica, Iquique/Alto Hospicio, Antofagasta, La Serena Coquimbo, Rancagua, Talca, Temuco, Puerto Montt. Following Robbins, Saunders, and Kilmer (2017), we will use calibration to develop weights that exactly match the synthetic control to the treated region. Operationally, this will be done using the new Rmicrosynth package (Robbins and Davenport, 2017). Data. Data on fatal and nonfatal traffic crashes and average precipitation are available for each city on a daily basis. We will also use city-year information on total population, population density, % younger than 25 years old, % male, unemployment rate, and % of the population <9 years of schooling. Due to changes in how the Uruguayan traffic crash data were collected, the analysis is limited to 2012-2017. Results. The data have been collected and analyses are underway. We will have a draft of the paper available in October 2018.