Census tracts are convenient units of analysis because they have similar population sizes, their boundaries align with the physical environment, and they are intended to be homogenous with respect to population characteristics and living conditions . Thus they roughly approximate city neighborhoods. From a routine activities perspective, I argue that it is reasonable to assume in the case of densely populated cities like San Francisco that likely offenders, in choosing whether, where, and when to commit a crime—that is, in weighing the target suitability and guardianship of potential victims—are going to consider targets within an area roughly the size of a census tract. Maps presented in the forthcoming analysis should illustrate the geographic implications of this assumption. Another advantage to using census tracts as the spatial unit of analysis is that there is an abundance of demographic information available at the census tract level via the U.S. Census Bureau and ACS. This provides for an excellent range of control variables. But this approach is not without its disadvantages.The present model does not account for criminal activity in other tracts and therefore misses the “spillover effects” that a land use such as MCDs may have on crime in neighboring tracts. This presents a significant limitation for the present model—although one that could theoretically be corrected for, to some extent, through more sophisticated spatial analyses. Considering the lack of empirical evidence currently available with respect to this issue—and its significant implications for policy making and future research—I argue that, as a preliminary analysis, this study has tremendous value despite this and other limitations. It may not account for inter-tract crime, but it does provide new knowledge about the nature of intra-tract crime.
City residents probably are concerned about businesses in adjacent neighborhoods; but when it comes to crime they are concerned, first and foremost, grow tray with the people next door.In this paper I examine the relationship between MCDs and crime in San Francisco for the year 2010. San Francisco provides an excellent case study for analyzing the social impact of MCDs because, unlike most jurisdictions in which MCDs have emerged in recent years, the local government in San Francisco has been regulating MCDs effectively for years. In some other cities, local governments and MCD operators have undergone heated legal battles with one another. This has resulted in a “regulatory vacuum” with respect to MCDs in most jurisdictions in which they exist . Perhaps the most notable example of this is Los Angeles, where MCDs have been in legal limbo for years. In early 2012 the Los Angeles Times editorial board, responding to a recent motion to ban all MCDs in the city, described the L.A. city government’s approach to dispensary regulation as a “roller-coaster ride”. At one point there were an estimated 1,000 MCDs operating throughout the city, leading to the observation that medical cannabis was more popular in L.A. than Starbucks. By contrast, in San Francisco, MCDs are governed by a comprehensive municipal ordinance that has survived judicial scrutiny to date. The San Francisco “Medical Cannabis Act” sets up a permitting system for MCDs and places certain restrictions on their location and operation. The Act requires that all MCDs comply with California state law as well as guidelines written in 2008 by then-Attorney General Jerry Brown entitled “Guidelines For The Security And Non-Diversion Of Marijuana Grown For Medical Use”. Brown’s guidelines state that “a properly organized and operated collective or cooperative that dispenses medical marijuana through a storefront may be lawful under California law” if they meet certain requirements, including: operating on a not-for-profit basis, obtaining all of the relevant permits and licenses from state and local government, taking steps to verify that their members are qualified patients under state law, acquiring and distributing only marijuana that has been cultivated legally, prohibiting sales to non-members, and providing adequate security .
Regarding security, Brown states that MCDs must “provide adequate security to ensure that patients are safe and that the surrounding homes or businesses are not negatively impacted by nuisance activity such as loitering or crime.” Expanding on this theme, the San Francisco Medical Cannabis Act prohibits “any breach of peace… or any disturbance of public order or decorum by any tumultuous, riotous or disorderly conduct” within permitted MCDs. MCDs are required to submit security plans as part of their permit application. This study examines whether such security protocols amount to capable guardianship, which is an effective deterrent against crime according to routine activities theory . San Francisco is not the only municipality that has regulated MCDs. Other California cities have enacted similar ordinances, including two prominent examples that can be found directly across the water from San Francisco in the cities of Berkeley and Oakland. But as a case study San Francisco has several advantages over these and other alternatives. First and foremost, it is a major city with a large sample of MCDs in the year for which data are collected. By comparison, Berkeley and Oakland have smaller populations and “hard caps” on the number of dispensaries allowed. So although they present interesting pieces of the legal, social, and political puzzles presented by California’s medical cannabis law, their small sample size limits the extent to which they are useful cases for empirical study. In cities that do not regulate MCDs, it is difficult to determine when and where dispensaries operate, and for how long. It is also much harder, in the case of unregulated dispensaries, to infer whether MCDs implement security protocols amounting to capable guardianship against crime, from a routine activities perspective . Simply put, San Francisco is the largest California to have enacted meaningful legislation with respect to MCDs. It has done so in a way that reasonably controls for crime, at least in theoretical terms. Thus it provides an excellent case study for analyzing the spatial relationship between crime and locally regulated MCDs.
Crime data were collected from the San Francisco Police Department in late 2011. The Crime Analysis Unit provided lists of serious crimes reported in 2010 along with the date and approximate location of each crime. Here, “serious crimes” refer to those classified as Part I offenses by the Federal Bureau of Investigation in its Uniform Crime Reports. The crime variables used in this analysis include measures of “violent crime” as well as “property crime” ; both as total counts and as rates per 1,000 residents. The lists of crimes and addresses were geocoded and aggregated into census tracts using ArcGIS software. Geocoding refers to the process by which tabular data are attributed spatial components by a geodatabase. Geocoding resulted in a successful match for more than 98% of all crimes, which were the aggregated into census tracts using a “spatial join” analysis. The remaining 1-2% of crimes were discarded from analysis. In addition, some reported crimes were removed from analysis because their geocoding confidence ratings were below 95%. In the end 43,688 reported crimes were analyzed out of the original 44,422 for which the San Francisco Police Department provided 2010 data. For the purpose of analysis, these crime measures were aggregated together by census tract and transformed by natural logarithm to correct for a right-skewed distribution. Descriptive statistics for these various categories of crime are presented in their original form in Table 4.1. The primary independent variable under review is the density of MCDs. Lists of MCD names and addresses were compiled using information provided by the San Francisco Department of Public Health. These MCDs were located across 16 census tracts primarily in the downtown area, as illustrated by Maps 4.1 and 4.2. The MCD addresses were geocoded to 100%. As with crime frequency, data for MCDs are presented in two forms. Descriptive statistics presented in Table 4.1 include MCD density as the number of dispensaries per square mile in a given census tract. For the regression analyses presented by Table 4.2, this variable is transformed by natural logarithm to address a right-skewed distribution. Crime rates by census tract are presented in Map 4.1 and Map 4.2 on the following pages. MCD locations are marked by green crosses. Property crimes include arson, burglary, larceny-theft, vandalism, and vehicle theft. Violent crimes include assault and robbery. Census tracts are assigned to one of five classes based on their crime rates. In addition to the primary variables already discussed, hydroponic trays data were also collected for several neighborhood characteristics that could potentially confound the relationship between MCDs and crime. These neighborhood characteristics are drawn from social disorganization theory, which associates higher rates of crime with socioeconomic disadvantage, family disruption, residential instability, and population heterogeneity . From these, the present study examines the criminogenic effect of poverty, unemployment, percent of single-parent households, percent of housing units that are vacant, and percent of the population between the ages of 18 and 24. Demographic data are collected from the American Community Survey database of the United States Census Bureau via the American FactFinder website, as well as the Demographic Research Unit of the California Department of Finance. With regard to the census data, variables are constructed from the ACS 5-year estimates for the year 2010.
Criminological research has found that indicators of socioeconomic disadvantage—including poverty and unemployment—have been associated with higher crime rates . In the present study economic data are collected from the ACS. The U.S. Census Bureau calculates the poverty status of individuals based on whether their total income in the past 12 months falls below the applicable poverty threshold, which is determined by age, family size, and family composition . The 2010 poverty thresholds range from $11,139 for a single individual living alone to $42,156 for a family of eight or more people living in the same household. For the present analysis, “poverty” means the number of individuals with incomes under their applicable poverty threshold in the past twelve months, divided by the total number of people for whom poverty status is calculated within a given tract.According to the United States Census Bureau, an individual is considered unemployed if he or she did not have a job and has been actively looking for work during the last four weeks and was available to start a job at the time of the survey . “Unemployment”, in the present analysis, means the unemployment rate in each tract as estimated by the ACS. Research examining crime rates in the United States during the 1990’s suggests that the job market can provide powerful explanations for criminal behavior . Poverty and unemployment are important measures in the model currently being tested, as they control for varying levels of socioeconomic disadvantage across city neighborhoods, which according to social disorganization theory affect crime rates in significant ways .In this study I use “family stability” as an inverse measure of family disruption. I calculate family stability by taking the number of individuals living in married couple family housing and dividing it by the number of people living in single-parent family households. Scholars of both routine activities theory and social disorganization theory predict that higher concentrations of married-couple families are associated with lower crime rates in urban areas, because more parents can provide more supervision and therefore more social control. From a routine activities perspective, both “family stability” and “residential stability” correspond with the notion of capable guardianship. According to social disorganization theory, residential instability weakens a community’s social cohesion and therefore its ability to deter and prevent crime within its territory. In this study I use vacancy rates as measure of residential instability. I calculate “vacancy” by dividing the number of vacant housing units within a census tract by the total number of units within that tract. Data for this measure comes from the Census 2010 Redistricting Plan. Another indicator of residential turnover discussed in the criminological literature is the percent of the sample population that is young . The idea is that neighborhoods with a high concentration of young adults will have correspondingly fewer older adults and children, which results in a lack of social cohesion and crime-preventive capacity much in the same way as the other precursors of “social disorganization” already discussed. The variable “percent young” was constructed using ACS population estimates by dividing the total number of individuals between the ages of 18 and 29 by the total tract population.