After analyzing the number of MMICs on crime and arrest rates, drug, alcohol, and other mortality rates will be regressed on the number of MMICs issued per county. The point of this is to observe whether or not marijuana has a negative effect on drug and alcohol related deaths, implying that marijuana is a substitute for other drugs. Because cross-sectional data will be used, there are unobservable events that could affect the analysis within that time period. For example, the Great Recession occurred from 2007-2009, which could have possibly increased crime rates. In order to combat time trend errors in the model, I will add annual fixed effects. This will allow the model to absorb any overlooked effects dependent on time. Because California counties are diverse and not all of them implement laws to the same extent, county fixed effects are also necessary for all regressions. By using these fixed effects, we will control for county-specific omitted variables that are time invariant. Relevant county-controlled variables may include the number of police stations or type of legislation implemented within a single county. All regressions in this report will contain both county and year fixed effects.The main data set we will use is the number of MMICs issued each fiscal year per county. This data was collected by the California Department of Public Health when SB 420 was implemented in 2005. The count of MMICs is updated through September 2015, but we will only use the number of cards issued from 2005-2014 since all other data is given annually. The cards issued each year range from zero to 1475. Because each card is only valid for one year, we assume that these annual numbers include renewed cards.
There is a variation in these numbers between counties and time due to the fact that some patients may not have renewed their cards and every county implemented this system at different times. Because it is a voluntary identification system,indoor growing trays any significant results would be under estimated. The MMIC data has been converted into number of MMICs issued per 100,000 people, as shown in 7.1.1 of the Appendix, in order for an easier interpretation between variables. It should be noted that some counties did not participate in some years and many others had zero medical marijuana cards issued at the beginning of 2006. Sutter and Colusa counties still have not applied this system and thus have no observations. Because there is no data on medical marijuana cards issued, Sutter and Colusa counties were omitted from all data sets. Table 4.1 below offers summary statistics for the MMICs issued per 100,000. In order to use unemployment as a right-hand side variable in the models, data from the California Employment Development Department was collected and offers per county unemployment rates from 1990-2014. This data will allow us to have a stronger model when examining the given research questions. Unemployment rates from 2005- 2014 will be used in order to compare it to our MMIC data. Referring again to Table 4.1, we observe a mean unemployment rate of 9.8. All data sets contain 560 total observations from the 56 counties used within the 10-year period.The mortality data to be used in this report comes from the Centers of Disease Control and Prevention . The mortality rates are divided into three categories: Alcohol-Induced Causes, Drug-Induced Causes, and All Other Causes.
These rates are given per 100,000, as shown in 7.1.3 of the Appendix. Estimated population sizes per year are also included in the data set. Like the MMIC data, the crude rates are reported per county, per year from 2005-2014. Within this time period, these crude rates have ranged from 4.9 to 1328.6 per 100,000. Referring above to Table 4.1, the mean alcohol-induced, drug-induced, and other crude rates are 13.5, 15.8, and 759.4, respectively. In addition to the mortality data, arrest rates will be examined to determine if medical marijuana is a substitute for other drugs and alcohol. The arrest data comes from the State of California Department of Justice’s Criminal Justice Statistics Center and includes 76 arrest variables. Of these 76 variables, I will be using 7 of them in my data analysis. These variables include marijuana, drunk, felony drug offenses, narcotics, dangerous drugs, other drugs, and total arrests. Other drugs represent all misdemeanor drug arrests excluding marijuana. However, the marijuana variable used in our data is the sum of both misdemeanor and felony marijuana arrests. As stated by the CJSC, “A felony offense is defined as a crime which is punishable by death or by imprisonment in a state prison. A misdemeanor offense is a crime punishable by imprisonment in a county jail for up to one year.” Full variable definitions are given in Table 7.2.1 in the Appendix. All variables in the data set were given as number of arrests per county, per year again from 2005-2014. As presented in part 7.1.2 of the Appendix, I converted these numbers into arrests per 100,000 so the analysis of all variables could be more easily interpreted. The CJSC has also provided crime data from 2005-2014 to be used in the regressions. Not to be confused with arrest data, the crime data set contains all individuals convicted of a crime, whereas arrests occur when a person is simply taken into custody for a crime.
The crime data presented by the CJSC offers 66 variables, from which I selected the 10 main types of crime, including, violent crime, burglary, larceny/theft, property crime, aggravated assault, motor vehicle theft, robbery, forcible rape, homicide, and total crime. Property crime is the sum of burglaries, larceny/thefts, and motor-vehicle thefts and violent crime is the sum of forcible rapes, homicides, and robberies. For full definitions of crime variables, refer to Table 7.2.2 in the Appendix. The crime data set originally included city and county distinction, but I collapsed the data into strictly per county observations. Computed the same as the MMIC, crude, and arrest rates, the third calculation shown in 7.1.3 of the Appendix was used to convert the numbers into crimes per 100,000 people. Table 4.2 below offers summarized statistics of all data collected from the CJSC.To begin analyzing the effect of medical marijuana in California, all nine individual crime rates and total crime rates were regressed on MMIC rates and unemployment rates with county and year fixed effects. It is necessary to include county fixed effects in the model because there are unobservable factors that could affect crime rates. For example, high-income counties in California may have lower crime rates by being able to afford tighter security. It is also obligatory to include year fixed effects in the crime rates model. This type of fixed effect absorbs any event or time trend that could potentially adjust crime rates. Because the data ranges from 2005- 2014, the housing market crash could have affected crime rates. Referring to Graph 5.1, it is indicated that crime rates don’t necessarily have a linear time trend. Thus, the individual year dummy variables will be the best fit to combat the unobservable events that occur across time. Here we see that for every additional medical marijuana card issued,pipp horticulture rack cost total crime decreases by one and a half crimes. This appears to be a significantly large effect. However, looking at the average MMIC rate of 53 and the average total crime rate of 6,210, it is unlikely that medical marijuana could completely eradicate crime. The estimated results imply that if the mean of MMICs goes up to 54, crime rates will fall to an average of 6,208.5. This is only a decrease of 0.024% of total crime, which is a small, yet reasonable estimate. While this is a small effect on total crime, the 95% confidence level suggests the true estimate is between -2.46 and -0.55. Because these values are negative, it is acceptable to assume medical marijuana will not negatively impact society by increasing crime rates. After observing that medical marijuana has a negative effect on total crime, it can also be seen that medical marijuana also has negative effects on larceny-theft and property crime, with estimates shown in tables 5.4 and 5.5. Table 5.4 indicates that for every additional MMIC issued, larceny/theft declines by about half of a crime, while Table 5.5 suggests that for every additional MMIC issued, property crime decreases by ¾ a crime. Because property crime is defined as the sum of larceny/thefts, burglaries, and motor-vehicle thefts, the effect on larceny/theft is contained within the effect on overall property crime. As these estimates appear to be miniscule, they are both statistically significant at = 0.05 with t-statistics of -3.66 and -3.52, respectively. Many individuals who argue against the legalization of marijuana claim that marijuana usage would increase crime, thereby negatively impacting society. By building a 95% confidence interval it is shown that the true estimates are negative and that 95% of the time, the estimate will fall between -1.18 and -0.33. Thus, medical marijuana will not increase overall property crimes, specifically larceny/thefts.
The other seven crime variables regressed on MMIC, using Equation 5.2, showed no significant effects of medical marijuana on crime. However, vehicle theft showed a statistically significant negative effect at the 90% confidence level. This can be explained by the above regression results on property crime, given that vehicle theft is included in the overall property crime rates by definition. All other crimes displayed zero effect from medical marijuana. While we can comfortably say that medical marijuana does not increase crime rates, there needs to be an explanation for why it has a significantly negative effect on both total crime and property crime. One explanation is that allowing consumers to purchase legally decreases the amount of associated crime that comes with the illegal marijuana market. It is often true that individuals who enact in criminal activity participate in more than one crime. This means when individuals are purchasing marijuana illegally, they are more likely to commit other crimes. Thus, when additional MMICs are issued, individuals are purchasing marijuana legally and are less likely to be crime participants. This effect can be seen in the above regression results where additional MMICs lead to a slight fall in committed crimes. A second explanation could be that there are substitution effects for marijuana and other drugs and alcohol. With evidence of marijuana reducing violent behavior, as explained further below, individuals are less likely to commit crimes. Because many crimes are committed while drunk or intoxicated, an increase in marijuana use with significant substitution effects on other drugs or alcohol could lead to a slight decrease in crime.This brings us to the next two models, created to observe whether or not marijuana is a substitution drug for alcohol and/or other drugs. Equation 5.6 regresses every individual arrest rate on MMICs and unemployment rates, while Equation 5.7 regresses drug-induced, alcohol-induced, and all other mortality rates on MMICs and unemployment rates. These two equations will allow us to examine any substitution effects going on between marijuana and other drugs and alcohol. Both equations are again controlled for county and time fixed effects. This means that 99.9% of the time medical marijuana has a negative effect on drunken arrests. While this indicates that there may be a substitution effect for alcohol, it is a small effect with a 1:4 substitution ratio. For this effect to decrease drunken arrest rates by 1%, MMICs would have to increase by about 20 per 100,00. This could be a possible scenario, given that the standard deviation of MMICs is 95. In the likelihood of this event, medical marijuana could be a significant substitute for alcohol. As briefly mentioned earlier in this analysis, a substitution effect between marijuana and alcohol can justify why we see a decrease in crime. It has been observed by many studies that a large proportion of crimes are committed when an individual is intoxicated. According to the Huffington Post, the National Institute on Alcohol Abuse and Alcoholism “found that 25-30% of violent crimes are linked to alcohol use,” and the journal of Addictive Behaviors performed a study that suggested “cannabis reduces likelihood of violence during intoxication,” thus explaining why an increase in marijuana use can decrease crime rates.By finding a slight substitution effect between marijuana and alcohol, we are able to explain some of the negative effect that marijuana has on crime.