Although they are not of particular theoretical interest, the following measures are included as demographic control variables: population size, percent of the population that is male, and percent of land that is commercially zoned. Population size and gender composition are adapted from the 2010 Census. “Percent of land commercially zoned” was calculated in ArcGIS using zoning shape files provided by the San Francisco Planning Department. Table 4.1 presents descriptive statistics for the measures analyzed in this study. Here the crime data are provided as total counts by category, but in the regression analysis that follows the crime variables are transformed by natural logarithm to correct for a right-skewed distribution. All other variables are presented as described in the previous section. A total of 189 census tracts within San Francisco County are analyzed using data for the year 2010. Five census tracts were removed from analysis because only partial data were available; these were low population tracks with no MCDs and therefore their loss is not analytically significant. Of the tracts analyzed, the average population size is 4,234. The average crime rate is 197.38 property crimes per year . The average violent crime rate is much lower: only 28.65 reported instances per year . As conveyed by the results of regression analyses presented in Table 4.2, seedling starter trays the model put forward by this study is better at explaining property crime than it is violent crime. The simplest explanation for this is that substantially more property crimes are committed on a yearly basis than violent crimes.
Descriptive statistics for the independent variables analyzed by this study are also presented in Table 4.1 These include “% Unemployed” and “% Under Poverty” as measures of socioeconomic disadvantage; “% of Housing Units Vacant” and “% of Population Ages 18-29” and measures of residential instability; and “% In Married-Couple Families” as a measure of family stability . These variables are included in the present model to test the explanatory power of social disorganization theory, which associates higher crime rates with “exogenous sources of social disorganization” , including socioeconomic disadvantage, residential instability, and family disruption. Additional control variables presented in Table 4.1 under “Other Tract Characteristics” include population size, population density, percent of land commercially zoned, and percent of the population that is male. For the vast majority of census tracts the value of the primary independent variable, MCD density, is zero. The 26 dispensaries operating in San Francisco in 2010 were largely clustered downtown. MCDs are not uniformly distributed throughout San Francisco. But neither is crime. Most crime also occurs downtown, with a large amount being reported in just a handful of densely populated census tracts. The fact that some of these high crime tracts also contain the majority of San Francisco’s dispensaries make it especially difficult, in the present analysis, to rule out a spurious correlation. This presents a limitation to the present analysis, albeit one that is intimately tied to the nature of the research question. Table 4.2 presents the results of four different linear regression analyses. The independent variables are the same in each, and so are the cases .
What differs is the dependent variable: total property crimes, property crimes per 1,000 residents, total violent crimes, and violent crimes per 1,000 residents. The primary independent variables are MCD density and the five variables drawn from social disorganization theory: poverty, unemployment, family stability, vacancy, and percent of the tract population between the ages of 18 and 29. Additional control variables include population size, percent of the population that is male, and percent of land that is commercially zoned. As Table 4.2 indicates, the four models are fairly consistent in their findings with respect to which variables predict what, and with what power. One interesting difference is that MCD density is a stronger predictor of violent crime than property crime—although the property crime models have slightly better explanatory power over all . The relationship between MCD density and crime is weak but statistically significant, according to this model. Of the other independent variables under review, the strongest predictors of crime—much more so than MCD density—are poverty and family stability. Poverty provides the strongest explanation for differences in violent crime rates across tracts, while family disruption provides the strongest explanation for property crime rates. As with MCD density, poverty is found to be a better predictor of violent crime than property crime. The opposite is true for family stability. As family stability increases, crime rates decrease. This holds across all four models. But the strength of that correlation for violent crime is nearly twice what it is for property crime . These findings are consistent with past research on social disorganization theory . Findings suggest that residential instability—as conceptualized by the present study—is not as strongly correlated with crime as the other “exogenous sources of social disorganization”. Like family stability, vacancy is a stronger predictor of property crime but not violent crime.
But unlike family stability, vacancy is statistically insignificant with respect to violent crime. These findings should be considered in light of the limitations faced by the present study, which are several. Foremost among these is the narrow conception of the question. The link between MCDs and crime is an intriguing puzzle academically, and one that deserves greater attention. But for policymakers it is just one piece of a much larger puzzle. There are a multitude of factors to consider when deciding whether and how to regulate MCDs. Public safety is, of course, a necessary consideration. But in California so is the provision of state law, passed by a majority of voters in 1996, that qualified patients should have legal access to medical cannabis. And so are a number of other factors, ranging from potential effects on children and teenagers to the budgetary impact of MCD regulation and taxation by local governments. Thus, this study should be considered in light of its narrow academic approach to a broad real world problem. To the extent that this study says anything about the relationship between MCDs and crime, it must also be noted how little it has to say about these other factors. In addition, there are certain technical limitations to this study that merit discussion. This study lacks breadth in both the quantity and quality of its case selection, which means that one should be cautious in generalizing from these findings. Moreover, the research design is purely observational . This makes causal identification impossible—observed associations could run in either direction, or they could be spurious. This study does not account for longer term trends in crime frequency or neighborhood characteristics, a problem that could be corrected by future research that incorporates longitudinal data. If, for example, the crime data obtained for this study could also be obtained for 1990 and 2000—years in which MCDs were nonexistent and virtually non-existent, respectively—then a “before and after” analysis could be conducted for San Francisco and other cities. The neighborhood data examined in this study also have their own limitations. While census tracts are convenient units of analysis, they are not perfect. They are imagined boundaries. The demographic data provided by the ACS provide a reasonably accurate picture of the people who live in a census tract, botanicare trays but do not account for the fact that people—and criminals in particular—tend to move from tract to tract. Nor do they account for tourists, transients, or anyone else unlikely to respond to census surveys. The MCD data are also limited. It is possible that some of the MCDs included in this analysis were not open for the entire calendar year in 2010, or that there were some MCDs operating in 2010 without recognition from their local government . There is also the question of other cannabis distributors not classified as MCDs. These include illicit dealers, medical cannabis delivery services, and small patient collectives without storefront locations.
The crime data suffer from limitations as well. This study inherits all of the imperfections that accrue in reporting, recording, and transmitting data within and from police departments. Furthermore, because they are reported by approximate location , these crime data suffer from a lack of precision. Furthermore, because this study limits its analysis to a subset of crimes classified as Part I offenses by the FBI, it is not a very broad measure of crime. Policymakers might also be interested in less serious categories of crime when making decisions related to MCD regulation. In this paper I have sought to contribute towards a more informed understanding of the spatial relationship between MCDs and crime. Findings suggest that there may be a relationship between the number of crimes reported in a neighborhood and the density of MCDs within the surrounding census tract. But this correlation does not take us very far in understanding the nature of this link. Do MCDs cause crime, or do they simply tend to locate in higher crime areas? Regression analysis casts doubt on the allegation that MCDs are “magnets for crime” by suggesting that some neighborhood characteristics are better predictors of crime than MCD density. The neighborhood characteristics tested by this study are drawn from the literature on social disorganization theory . Of the measures analyzed, the best predictors of high property crime rates are, in descending order of strength and significance: family instability, poverty, percent of housing units that are vacant, percent of individuals between the ages of 18 and 29, and unemployment. The best predictors of violent crime rates are poverty and family instability. MCD density is a better predictor of violent crime than it is property crime. In many cases the relationship between one or more of these factors and crime is stronger than the relationship between MCDs and crime. Thus it may be the case that certain neighborhoods would have even higher rates of crime if not for the presence of MCDs—ongoing longitudinal research should test whether this is the case. And generally speaking, future research should address the uncertainty in these findings by collecting larger bodies of data across longer periods of time. As more thorough and better-controlled analyses are conducted across the many jurisdictions currently grappling with MCD regulation, a clearer picture will emerge of the relationship between medical cannabis dispensaries and crime.Criminologists call crimes that have occurred, but that are not recorded or reported, the “dark figure of crime”, and they form a group of important missing statistics in understanding crime. Ever since crime statistics began being formally collected in the 19th century, this group of missing statistics has been a problem that has plagued law enforcement and criminologists. This dark figure exists for two main reasons: victims fail to report crimes , and law enforcement agents are unable to detect crimes . There have been many attempts by law enforcement and criminologists to better estimate crime and diminish this dark figure through improved and new types of surveillance, anonymous reporting systems and victimization surveys, like the National Crime Survey . More recently, law enforcement at international, national and regional levels has attempted to detect crime by using remote sensing technologies. Using imagery collected remotely, from sensors onboard aircraft, unmanned aerial vehicles and satellites, law enforcement agents have been able to assess where and when certain kinds of crimes have taken place. The use of remote sensing, the “observation of earth’s land and water surfaces by means of reflected or emitted electromagnetic energy” or, more simply, a method of “acquiring data about an object without touching it”, for surveillance and analysis has obvious benefits for law enforcement agencies . It greatly expands the supervision of agents of the law in often remote or inaccessible places, reduces the exposure of these agents to dangerous circumstances on the ground and may make up for a lack of manpower . At the same time, using remote sensing has at least three serious limitations. First, and perhaps most obviously, remotely sensed images that are gathered from overflying helicopters, aircraft or satellites can only detect crimes or crime’s impacts that are visible from above and for sustained periods of time. For example, remote sensors can identify illegal logging, large-scale drug production, and trails in the desert but they would be much less likely to detect murder, assault, homicide, robbery, or other small-scale, undercover, rapid actions, though some attempts have been made to capture the lasting effects of these things, for examples, see Pringle and others. Second, remote sensing cannot record the social, political, economic and historical context of landscapes and the actions that take place within them.