Low Priority Laws and the Allocation of Police Resources

Transcription

1 Low Priority Laws and the Allocation of Police Resources Amanda Ross Department of Economics West Virginia University Morgantown, WV And Anne Walker Department of Economics University of Colorado at Denver Denver, CO Abstract: There is an ongoing literature in economics examining the deterrent effect of police officers on criminal activity. However, this literature tends to focus on the aggregate number of officers employed versus the relative allocation of an officer s time. In this paper, we examine how the reallocation of police resources affects police behavior and criminal activity using the adoption of low priority initiatives by some jurisdictions. Low priority initiatives mandated that minor marijuana possession offenses be the lowest enforcement priority for police officers. We first test whether adoption of the initiative decreased the arrest rate for minor marijuana possession offenses. If police officers devote fewer resources towards minor marijuana possession crimes, then more resources will be available to deter and solve more serious crimes. This would suggest that if misdemeanor marijuana arrest rates decreased, there may be a reduction in crime rates or clearance rates for more serious crimes, such as murder or robbery. Using city-level data from California, we find that those jurisdictions that adopted low priority laws experienced a reduction in arrests for misdemeanor marijuana offenses. However, we do not find a significant effect of enacting a low priority initiative on the crime rate or clearance rate of more serious felony crimes. Our findings are important for local policy makers, as we do not find evidence that the initiatives had an impact on more serious crimes as was intended by the legislation. Keywords: Police Resource Allocation, Police Incentives, Drug Policy, Low Priority Laws JEL Code: H1, H4, K4, R5 1

2 1. Introduction Becker (1968) proposed a model of rational criminal behavior where individuals weigh the costs and benefits of crime and, based on that calculation, choose to engage in criminal activity or not. This model of criminal behavior predicts that an increase in police resources will increase the probability that a criminal will be arrested. An increase in the probability of arrest will increase the costs of engaging in criminal activity and will reduce crime rates. Since the seminar work of Becker, numerous economists have empirically tested the effect of additional police protection on crime rates within a jurisdiction (Benson and Rasmussen, 1991; Benson, Kim, Rasmussen, & Zuehlke, 1992; Cornwell & Trumbull, 1994; Sollars, Benson, & Rasmussen, 1994; Levitt, 1997; Benson, Rasmussen, & Kim, 1998; Corman & Mocan, 2000; McCrary, 2002; Levitt, 2002). However, much of the existing empirical work examining the relationship between police protection and deterrence has relied on the aggregate number of police officers, not on the relative amount of time officers spend on various tasks. To determine the effect of additional police resources on crime rates, the manner in which officers allocate their limited time is the relevant information needed. Since direct data on the division of time and resources within police departments is not available, policies which mandate a specific allocation of an officer s time provide an opportunity to test the consequences of a reallocation of police resources on crime rates. In this paper, we draw upon such a mandate, known as a low priority initiative, to examine how a change in the relative amount of time police officers spend on specific activities affects criminal activity within a jurisdiction. Low priority initiatives mandate that minor marijuana possession offenses be the lowest 2

3 enforcement priority for local law enforcement agencies. Those cities that have adopted such an initiative believe that by shifting attention away from low-level drug crimes, police officers will be able to focus more time towards more serious crimes, such as murder, rape, robbery, assault, burglary, motor vehicle theft, and larceny. Gil Kerlikowske, the police chief in Seattle in 2003, stated that the issue was one of limited police resources in reference to Seattle s policies on drug enforcement, including a low priority law. 1 While low priority initiatives have been adopted by several localities since 1979, no research thus far has examined the impact of these mandates on police behavior and criminal activity. We begin by estimating the effect of the adoption of a low priority law on arrests using data on the number of misdemeanor marijuana drug arrests in California from 2000 to Using a city-level panel data set, we find that the adoption of a low priority law reduced the number of arrests for misdemeanor marijuana offenses per person. This reduction in the arrest rate is present when we look at the entire state of California, as well as when we exclude small towns and focus on only the larger cities. We also considered the impact of the policy on arrests rates of other crimes to address any concern that there was a change in the arrest behavior of police officers in general versus just for misdemeanor marijuana offenses as intended by the policy. We do not find evidence that the adoption of a low priority initiative affected the arrest rate for any other type of crime, suggesting the policy had the intended impact and only affected arrests of low-level marijuana offenses. Next, we consider the effect of this reallocation of police resources away from low-level 1 3

4 marijuana crimes on more serious offenses. We look at the impact of the adoption of low priority initiatives on both the crime rate and the clearance rate (defined as the number of crimes for which an arrest was made divided by the number of reported crimes) for murder, rape, robbery, assault, burglary, motor vehicle theft, and larceny. Both of these rates could be affected by the additional police resources, as officers could either deter more individuals from committing crimes (which would reduce the crime rate) or solve more crimes (which would increase the clearance rate). We find that adoption of a low priority initiative did not have a significant negative effect on the crime rate nor a significant positive effect on the clearance rate for any of these violent and property crimes. Overall, we find that the low priority initiative had the intended effect of reducing the number of minor drug crime arrests, but we do not find evidence of a deterrent effect for any other serious crime. Our findings have important implications for local policy makers regarding the allocation of scarce police resources since we do not find that the policy had the intended effect of enabling police officers to deter more serious crimes. Some possible explanations for our findings are that not enough free time is freed up by the mandate to have a measurable effect on other types of crime. Alternatively, the mandate may not cause a reallocation of resources, but may simply cause police officers to stop engaging in one activity but do not use that time for another productive activity. We discuss in detail the different mechanisms that may explain the lack of an effect on other types of criminal activity in the conclusion. The remainder of this paper is organized as follows. Section two provides background information on low priority initiatives as well as a brief overview of the literature on the 4

5 economics of crime. Section three describes our empirical methodology while section four discusses the data used in our analysis. Section five presents estimates of the impact of low priority laws on the arrest rate for misdemeanor marijuana offenses, as well as other types of crime, and section six contains estimates of the impact of low priority laws on the crime rate and clearance rate for felony violent and property crimes. Conclusions and policy implications are discussed in the final section. 2. Background Information 2.1. Low Priority Initiatives Low priority initiatives mandate that minor marijuana possession offenses be the lowest enforcement priority for local law enforcement agencies. While there are some differences in the specific law that was implemented in each jurisdiction, a few components are common across all the initiatives. 2 First, the law only affects minor marijuana possession offenses. Felony drug crimes, including felony-level possession and distribution offenses, were not affected by the policy change. Second, the law only affected offenses where marijuana was intended for adult personal use. Possession or selling of marijuana to minors are not affected by low priority initiatives. Finally, the mandate was only intended to affect the private use of marijuana, so any offenses committed on public property were not affected. 3 The first low priority initiative was passed in Berkeley, CA in Since then, 16 other 2 For an example of a specific low priority initiative, the Oakland initiative can be found at the following link: 3 Most of the initiatives also had some language regarding who was responsible for making sure the ordinance was enforced. 5

6 cities have adopted a similar initiative. Table 1 contains a list of all the localities that adopted the policy, as well as the year that the initiative was implemented and whether it was passed through a city council vote or a voter initiative. The last column of Table 1 includes information on the margin by which each jurisdiction passed the law. As we can see in Table 1, these initiatives have become increasingly popular over time, as the margin of individuals who voted for the law has steadily increased. The goal of a low priority initiative is for police to allocate fewer resources towards the enforcement of minor marijuana possession laws so that resources can be allocated towards more serious offenses. For example, the initiative that was passed in Santa Monica noted that the initiative makes marijuana offenses, where cannabis is intended for adult personal use, the lowest priority and that the law frees up police resources to focus on violent and serious crimes, instead of arresting and jailing nonviolent cannabis users. 4 Therefore, the success of the initiative is measured through two mechanisms. First, did the policy cause a reduction in the arrest rate for minor marijuana offenses? Second, after observing a decrease in the number of arrests for minor marijuana possession crimes, was there an effect on more serious crimes? This second question is a direct test of the Becker model of rational criminal behavior and the deterrence hypothesis The Deterrence Hypothesis Becker's (1968) economic theory of crime provides the theoretical foundation for the deterrence 4 See for more details on the Santa Monica initiative. 6

7 hypothesis. Becker models criminals as rational, utility-maximizing individuals. These individuals weigh the costs and benefits of different activities and choose how to allocate their resources between crime, a risky asset, and legitimate income, a riskless asset. Becker's model shows that an increase in the probability of capture and punishment decreases an individual's expected utility of committing a crime. Therefore, an increase in the probability of arrest will reduce the likelihood that an individual commits a crime. Note that this deterrent effect of police officers is present through two separate mechanisms. First, the additional police presence will allow officers to be in the appropriate place to stop a crime that is about to be committed. In addition, the additional officers may cause individuals to choose not to attempt to commit a crime in the first place. The knowledge of the additional officers and the fear of being caught, even if the individual may not be able to see the officer at that moment, may cause individuals to be less likely to commit a crime. Both of these mechanisms would suggest additional police officers reduce crime rates. Becker's theory spawned a huge empirical literature that tests the economic model of crime and its implications. Numerous papers have examined the impact of additional police officers on crime rates. Initially, many researchers found additional police officers had a positive impact on crime rates (see Cameron (1988) for a review). This positive coefficient is likely caused by a reverse causality issue, since areas with higher crime rates will employ more police officers. Levitt (1997) used election years to instrument for police officers and found that increases in police resulted in a substantial reduction in violent crime. While studies questioned the validity of his instrument (McCrary, 2002), research using other instruments has found that additional 7

8 police resources reduce crime rates (Levitt, 2002; Corman & Mocan, 2000). Other studies have used the reallocation of police officers to specific neighborhoods in response to terrorist threats, which is exogenous to local crime rates, to estimate the impact of police presence on crime rates (Di Tella & Schargrodsky, 2004; Draca et al., 2011; Klick & Tabarrok, 2005) Police Resources One criticism of this literature that examines the impact of police on crime rates is that the data on police protection only has information on the aggregate number of police officers, as opposed to the relative allocation of police resources (Benson, et al., 1994; Benson et al., 1995). More specifically, the incentives of police and the discretion of officers in allocating their time cannot be measured when only the total number of officers employed is available. Once the discretion of individual police officers is taken into account, it is no longer clear that increasing the number of police officers will necessarily increase the probability of arrest and reduce crime rates. However, data on how individual police officers allocate their time is not available, making it difficult to determine exactly how police allocate their limited resources among various tasks. Benson et al. (1998) provide empirical support for the claim that it is not simply the aggregate level of police resources that matters, but rather the relative allocation of police resources towards enforcement of particular types of crime. Using the war on drugs during the 1980's as a source of exogenous variation in how police officers allocate their time, the authors 5 In addition, numerous papers have looked at the impact of different sentencing policies on crime rates, specifically the death penalty and more recently Three Strikes Laws and Truth-in-Sentencing Laws (Ehrlich, 1975; Ehrlich, 1977; Shepherd, 2002; Mocan & Gittings, 2003; Dezhbakhsh & Shepherd, 2006; Helland & Tabarrok, 2007; Iyengar, 2008; Ross, 2012). 8

9 test whether the reallocation of police resources towards drug crimes had an impact on violent crime rates (murder, rape, robbery, and aggravated assault). 6 They conclude that changing the incentives of police officers to allocate more time and effort towards drug crimes took resources away from other crimes and caused an increase in violent crime rates. Sollars et al. (1994) also explored the effects of the reallocation of police resources due to drug enforcement, but focused on the resulting changes in the property crime rate (burglary, motor vehicle theft, and larceny). They found that the reallocation of scarce police resources away from property crime and towards drug crime significantly decreased the risks that individuals who commit property crimes faced, causing an increase in the property crime rate. Other empirical studies have also shown that the probability of arrest for property crimes is negatively related to the portion of police effort directed at drug control (Benson & Rasmussen, 1991; Benson et. al., 1992). All of these studies support the notion that the allocation of resources within the police department matters more with regards to deterring crime than does the aggregate amount of resources utilized by the police agencies. We expand upon this literature by using adoption of low priority initiatives to examine the impact of allocating resources away from a specific policing activity. The war on drugs created a situation where police were incentivized to allocate additional time towards major drug crimes and away from deterring property and violent crimes. Low priority initiatives mandate that police officers allocate less time and effort towards arresting individuals for minor drug 6 Mast et al. (2000) shows that the policy changes associated with the war on drugs created an increase in drug arrests in Florida. Their findings suggest that police have considerable discretion when deciding how to allocate their scarce resources. 9

10 crimes. If police heed the mandate and stop arresting individuals for minor marijuana offenses, then additional police time has been created which can be spent deterring more serious crimes. However, just because police officers spend less time arresting individuals for minor possession offenses does not necessarily mean that officers use this additional time towards other productive activities. Our paper contributes to the literature by looking at how removing incentives to prosecute certain types of crimes affects police behavior, versus previous work which has looked at what happens when police are incentivized to prosecute a specific type of crime. 3. Empirical Methodology In this paper, we answer two questions. First, did low priority initiatives decrease arrest rates for minor marijuana possession offenses? Second, if low priority initiatives caused police to devote less time and effort towards these minor drug crimes, did this reallocation of resources have a deterrent effect on more serious crimes? If low priority initiatives resulted in a reduction in the arrest rates for minor marijuana possession offenses, then this would suggest that police have allocated resources away from this activity. Resource allocation decisions by police officers cannot be observed directly, but arrest rates reflect the consequences of that decision. If we find that the low priority initiatives lead to a reduction in the number of arrests for misdemeanor marijuana offenses, then we can evaluate the effect of the implicit reallocation of resources on the deterrence of more serious violent (murder rape, robbery, and assault) and property crimes (burglary, motor vehicle theft, and larceny). The first question considers the effect of low priority initiatives on the number of arrests per 10

11 capita for misdemeanor marijuana offenses. To estimate the effect of the policy on arrest rates, we use the following specification: log(arrest it ) = β 1 LP it + β 2 X it + γ t + α i + v it (1) where arrest it is the number of arrests for minor possession of marijuana per 1,000 people in year t in city i. LP it is a dummy variable that equals one if city i had a low priority law in year t. If the estimated coefficient on the low priority law dummy, β 1, is negative and significant, then the enactment of low priority laws had the intended effect of reducing the number of arrests for minor drug crimes. X it is a vector of control variables including percent male, percent black, median age, percentage households with a female head of household and children, as well as various controls for the education level of the individuals. We also include in X it the number of police officers per 1,000 people. γ t is a year fixed effect to control for year specific shocks that are common to all jurisdictions, α i is a city fixed effect to control for unobserved differences in each locality, and v it is an idiosyncratic error term. If the estimates obtained from the first regressions indicate that a reallocation of police resources occurred, then we can consider if this reallocation affected other types of crimes. To determine the effect of low priority laws on the deterrence of more serious crimes, we estimate the following: log(crime it ) = β 1 LP it + β 2 X it + γ t + α i + ε it (2) The dependent variable in equation (2), log(crime it ), is a measure of criminal activity in city i in year t. We use two different measures of criminal activity as the dependent variable the crime rate and the clearance rate (number of arrests divided by the number of reported crimes). 11

12 Using these different measures will allow us to account for the different mechanisms through which the reallocation of police resources may have a deterrent effect. For example, the reallocation of police resources could deter individuals from committing a crime through increased police presence, which would reduce the crime rate. Alternatively, police could allocate more resources towards solving crimes and thus would increase the clearance rate. Both effects are important to policy makers as both may deter criminals. 4. Data Given the difficulty of finding a nation-wide data set on misdemeanor marijuana crimes, we focus on California a large state that had seven cities implement low priority initiatives. Annual data on the number of arrests for misdemeanor marijuana offenses from 2000 to 2009 in California was obtained from the California Department of Justice. 7 Annual data on the Index I violent and property crimes, as well as the number of police officers per capita, were obtained from the Federal Bureau of Investigation s (FBI) Uniform Crime Reports (UCR). Data for the socio-economic control variables used in our analysis was obtained from the City and County Data Books. Combining these various data sets, we create a panel data set on California cities from 2000 to One concern when considering the impact of the policy on arrest behavior as well as the crime rates is that those areas which adopted the policy are different than those that did not. Since we are using the non-adopting cities to model what would have happened in the absence of 7 See the following website for information on our arrest data: 12

13 the policy, it is important that the adopting and non-adopting cities are similar prior to passage of the law. To provide evidence that this assumption is true, we present summary statistics for both the cities that implemented a low priority law and those that did not in Table 2. The first column gives the summary statistics for all cities in California, while columns 2 and 3 break the sample into adopters and non-adopters. The final column presents the t-statistic of whether the means are statistically different from one another. Looking first at the arrest rate for misdemeanor marijuana offenses, we see that the nonadopters appear to have a higher arrest rate than the adopters. However, when we account for the variance of the samples and test for a difference in means we do not find that the average arrest rate of the two groups are statistically different from one another. While the means are not statistically different from one another, it is important to note how the potential difference would impact our estimates. If the police in the adopting areas were already treating these offenses as the lowest priority, then this would suggest that our point estimates are underestimating the true effect of the policy on the arrest rate. Therefore, we argue that our estimates of the impact of low priority initiatives on arrests of misdemeanor marijuana crimes are actually a lower bound of the true effect of the policy. Looking next at the difference in the means of the other crime rates, we do not find that there is a statistically significant difference between the average crime rate of those cities that implemented low priority laws and those that did not for murder, rape, robbery, assault, burglary, motor vehicle theft, or larceny. We also do not find that the number of police officers per capita is statistically different in the adopting and non-adopting jurisdictions. Overall, these results 13

14 suggest that in terms of police protection and crime rates, there was not a significant difference between the cities that chose to adopt the policy and those that did not. We do see some differences in terms of the socio-economic attributes of the cities that adopted low priority initiatives and those that did not. Specifically, we find statistically significant differences between the adopters and non-adopters with respect to the percent of residents within the city that are black as well as the educational attainment measures. Adopting cities tend to have a higher percent of black residents and also have residents with higher levels of educational attainment on average. We include these variables as controls in our regressions to remove any possible effect these characteristics may have on our variables of interest. 5. Impact of Low Priority Laws on Arrests 5.1. Impact of Low Priority Initiatives on Misdemeanor Marijuana Arrests Table 3 presents estimates of the impact of low priority laws on the log of the number of lowlevel marijuana arrests per 1,000 people. We define low-level marijuana arrests as the number of arrests for misdemeanor marijuana offenses. The first column presents results for the full sample, while columns two and three restrict the sample to cities with populations above 30,000 and 50,000, respectively. 8 We restrict the sample by population to account for the possibility that the full sample results are being driven by small towns versus larger cities. Since larger cities tended to be the adopters of the low priority initiatives, we exclude the smaller jurisdictions 8 We select these cutoffs based on the population of the adopting jurisdictions. West Hollywood has a population of just over 30,000, so we look at both at 30,000 and at 50,000 since we are losing one of the adopting cities when we increase the cutoff. 14

15 so that the two samples are more comparable. Standard errors are reported in parenthesis below the coefficients. Looking at the first column, we see that adoption of low priority initiatives caused a reduction in the number of arrests for minor marijuana offenses. The coefficient, which can be interpreted as a semi-elasticity, suggests that adoption of a low priority law will reduce the arrest rate for misdemeanor marijuana offenses by 0.28%. This effect is statistically different from zero at the 10% level. When we restrict the sample to those cities with larger populations, we find the effect is still negative and approximately the same magnitude, but these estimates are more precise and are now significant at the 5% level. The increase in precision, despite the smaller sample size, supports our exclusion of the smaller, less comparable jurisdictions. Overall, our findings in Table 3 indicate that in the areas that adopted low priority laws, police heeded the mandate and arrested fewer individuals for low level marijuana offenses. We include a variety of other control variables, in addition to MSA and year fixed effects, to control for any unobservable variables that may be present and driving our results. These controls include the log of the number of police officers per capita, as well as a variety of other socio-economic controls. 9 We find a positive effect of the number of police officers on arrests, which suggests that more police officers lead to more arrests. This positive effect is not surprising given that hiring additional officers leads to more individuals to make arrests. In general, the coefficients on the other socio-economic variables do not have a consistent pattern 9 Controls include median age, percent male, percent white, percent black, as well as the measures to control for education. Our socio-economic data does not have information on percentage Hispanic, but this ethnic group is included in the omitted category. 15

16 across the different city sizes. Thus far, we have argued that those cities which adopted low priority initiatives are not different from those that did not. While there are differences across cities with regards to the decision to enact a low priority initiative or not, there are also differences across the adopting cities in terms of how the policy was implemented. We draw upon differences in whether cities adopted through voter initiative or city council vote to see if there are heterogeneous effects of the policy based on how the initiative was passed. Of the seven cities in California that adopted the law, only two, San Francisco and West Hollywood, passed the law through a city council vote. The rest of the cities adopted the policy through voter initiatives. In Table 4, we include an interaction between an indicator variable equal to one if the city adopted a low priority law and another indicator variable equal to one if the law was passed by a city council vote. We find that adoption by the city council does not have a statistically different effect than if the law was passed through a vote by the general public. Therefore, it appears that the method through which cities passed a low priority law did not have a differing effect with regards to the police s response to the mandate Impact of Low Priority Initiatives on Arrests for Felony Crimes Low priority initiatives were intended only to affect how police approached low-level, misdemeanor marijuana possession offenses. Adoption of the policy should not have caused police to adjust arrest behavior with regards to more serious felony marijuana crimes and other 16

17 felonies, such as murder or motor vehicle theft. 10 To examine if police changed their arrest behavior in general or only for the crime indicated by the initiative, we examine in Table 5 the impact of low priority laws on arrests of felony crimes. In the first column, we again include the full sample and then stratify by if the city has a population over 30,000 and over 50,000. The coefficients in the first row provides estimates of the impact of implementation of low priority initiatives on arrests for all felony drug crimes. This would include felony marijuana arrests, as well as arrests for any other type of drug. These types of crimes were not intended to be affected by the low priority laws, as the only offense that the law mandated that police make the lowest priority were misdemeanor marijuana crimes. As we can see in the first row of Table 5, adoption of the law did not have any effect on arrests for felony drug crimes. In addition, we consider the impact of adopting low priority initiatives on the arrest rate for other types of felonies murder, rape, robbery, assault, burglary, motor vehicle theft, and larceny. We do not find that implementation of the policy had an effect on the arrest rate for any other type of crime. 11 These findings indicate that police did not change their arrest behavior in general, only that they reduced arrests for misdemeanor marijuana offenses as instructed by the initiative. 10 The adoption of the policy could have an effect on arrest behavior for other crimes. Since there are additional police resources that can be devoted to catching other types of criminals, this would increase the number of arrests. However, if the additional police officers have a deterrent effect, this would cause fewer crimes to be committed and would thus decrease the number of arrests made. Given these two mechanisms, the possible effect of the policy on arrests is theoretically ambiguous. For that reason, we focus on arrests in this section to determine if there was a change in police behavior in general versus using arrests as a measure of deterrence. 11 We have also looked at misdemeanor offenses that could have been impacted, such as disorderly conduct, trespassing, and vandalism. We continue to find no statistically significant increase in arrests for these low-level crimes. Since the next section focuses on felony arrests, we only show the results for these more serious crimes in the interest of space. 17

18 6. Impact of Low Priority Laws on Crimes Cities enacted low priority initiatives with the stated purpose of freeing up police resources from dealing with minor drug offenses so that police could concentrate efforts on more serious crimes. In the previous section, we found that the enactment of a low priority law led to a reduction in the number of arrests per capita for misdemeanor marijuana offenses. Next, we examine if the reallocation of police resources had a deterrent effect on more serious Index I violent and property crimes murder, rape, robbery, assault, burglary, motor vehicle theft, and larceny. There are two ways that police can affect crime they can stop crimes from happening or they can solve more crimes that do happen. The first relationship can be examined through the crime rate and the second effect can be seen through the clearance rate (number of reported crimes divided by the number of arrests). We first consider the impact of low priority laws on the crime rate for the different violent and property crimes in Table 6a. If police are devoting fewer resources towards low-level marijuana offenses, this frees up resources to devote towards patrolling and preventing more serious crimes. Therefore, if the deterrence hypothesis is true, we expect that adoption of a low priority initiative will cause a reduction in the rate at which more serious crimes are committed. As we can see in Table 6a, we do not find that adoption of low priority initiatives caused a statistically significant reduction in any type of violent or property crime. In fact, the only statistically significant effect we obtain is that adoption of the low priority initiative increases the aggravated assault rate in large cities. In Table 6b, we look at the impact of low priority initiatives on the clearance rate for the 18

19 different types of violent and property crimes. In general, we do not find a statistically significant effect of the law on the clearance rate. The exception to this, however, is that we find adoption of low priority laws reduce the clearance rate for larceny. In other words, police are finding the guilty party for fewer larcenies after adoption of the policy. The goal of the low priority initiative was to reallocate police resources away from low-level marijuana possession offenses towards more serious crimes. While we did find there was an impact of the laws on the arrest behavior for minor marijuana crimes, we did not find an impact on the crime rate for more serious crimes, nor did we find there was a positive impact on the clearance rate for other types of crimes. Overall, our findings indicate that even though police devoted fewer resources towards arresting individuals for misdemeanor marijuana offenses, there was no measurable effect on deterring more serious violent and property crimes through either lower crime rates or higher clearance rates. 7. Conclusions and Policy Implications Low priority laws were adopted to allow police agencies to focus less attention on minor drug crimes and more attention towards violent and property crimes. Using data from California between 2000 and 2009, a large state where seven cities adopted a low priority initiative, we examine first if adoption of low priority laws had an effect on the number of arrests. If we find evidence that police reallocated resources as a result of the policy, we can then examine if this reallocation deterred individuals from committing other crimes. We do find that low priority initiatives caused a reduction in the number of arrests for 19

20 misdemeanor marijuana offenses. In addition, we do not find any effect on arrests for other types of crimes, which suggests that the effect was specific to only these minor marijuana offenses. However, we do not find any significant deterrent effect through either a reduction in the crime rate or an increase in the clearance rate of other crimes, including murder, rape, robbery, aggravated assault, motor vehicle theft, burglary, and larceny. These results hold for both the entire sample, as well as when we focus on only the larger cities. We found that the policy successfully caused police to focus less attention on low-level drug crimes. However, we did not find evidence that after spending less time arresting individuals for these low-level crimes that police spent more time and effort on more serious offenses. There are several reasons why this could be the case. First, while we know that there was a reduction in the number of arrests for minor marijuana offenses, we do not know exactly how much time was created as a result of the policy. Minor marijuana offenses may be an extremely low cost crime for police with regards to amount of time spent on this activity, so one explanation is that not enough additional time was created to have a measurable impact on other crimes. Second, it is possible that police are using the free time to engage in non-productive behavior. Previous studies that have examined changes in the relative allocation of police resources have focused on changes that incentivized officers to spend more time on a specific activity. For example, Baicker and Jacobson (2007), Mast et al. (2000), and Benson et al. (1995) provide empirical evidence that asset forfeiture laws have a positive effect on drug enforcement and that these positive effects are primarily due to the financial incentives created by such laws. Asset forfeiture laws allow police agencies to keep a substantial share of assets they seize during 20

21 drug arrests, thereby increasing their discretionary budget. The low priority initiative mandates low-level marijuana offenses be the lowest priority, but does nothing to incentivize police to allocate those additional resources towards more serious crimes. Therefore, another reason that we may not find an effect on more serious crimes is because police are not reallocating their time towards another activity because they lack any incentive to do so, and instead use that time they were spending arresting misdemeanor marijuana offenses doing nothing. Finally, even if reducing the time spent enforcing minor drug crimes did create a significant amount of additional police time to have a measurable impact on other types of crimes, this does not mean the different activities are close substitutes. The assumption of the policy is that time spent patrolling and arresting individuals for one type of crime versus another are perfect substitutes. There may be different skills and knowledge required to police the different types of activities, which may explain why we do not find an effect of the policy on other types of crime. Future work should consider how substitutable police time is between different types of crime prevention. If police resources and effort towards different crimes are not substitutes, than policies such as low priority initiatives are not viable options to reduce the crime rate of more serious offenses. 21

22 References Baicker, K, and M. Jacobson Finders keepers: Forfeiture laws, policing incentives, and local budgets. Journal of Public Economics 91 (11-12) (12): Becker, G.S Crime and punishment: An economic approach. The Journal of Political Economy, 76(2), Benson, B.L., I. Kim, and D.W. Rasmussen Estimating deterrence effects: A public choice perspective on the economics of crime literature. Southern Economic Journal, 61(1), Benson, B.L., I. Kim, D.W. Rasmussen, and T.W. Zuehlke Is property crime caused by drug use or by drug enforcement policy? Applied Economics, 24 (7), 679. Benson, B.L., and D.W. Rasmussen Relationship between illicit drug enforcement policy and property crimes. Contemporary Policy Issues, 9(4), Benson, B.L., D.W. Rasmussen, and I. Kim Deterrence and public policy: Trade-offs in the allocation of police resources. International Review of Law and Economics, 18(1), Benson, B.L., D.W. Rasmussen, and D.L. Sollars, Police bureaucracies, their incentives, and the war on drugs. Public Choice, 83(1-2), Cameron, S., The economics of crime deterrence: A survey of theory and evidence. Kyklos, 41 (2): Cornwell, C., and W.N. Trumbull, Estimating the economic model of crime with panel data. The Review of Economics and Statistics, 76(2), Corman, H. and H.N. Mocan, A Time-Series Analysis of Crime and Drug Use in New York City. The American Economic Review, 90, Dezhbakhsh, H. and J. Shepherd, The Deterrent Effect of Capital Punishment: Evidence from a Judicial Experiment. Economic Inquiry, 44, Di Tella, R., and E. Schargrodsky, Do Police Reduce Crime? Estimates using the Allocation of Police Forces after a Terrorist Attack. The American Economic Review, 94,

23 Draca, M., S.J. Machin, and R. Witt, Panic in the Streets of London: Police, Crime, and the July 2005 Terrorist Attacks. The American Economic Review, 101, Ehrlich, I., The Deterrent Effect of Capital Punishment: A Question of Life and Death. The American Economic Review, 65, Ehrlich, I., Capital Punishment and Deterrence: Some Further Thoughts and Evidence. Journal of Political Economy, 85, Helland, E. and A. Tabarrok, Does Three Strikes Deter? A Non-Parametric Estimation. The Journal of Human Resources, 22, Iyengar, R., I d Rather be Hanged for a Sheep than a Lamb: The Unintended Consequences of California s Three Strikes Laws. National Bureau of Economic Research Working Paper Number W Klick, J. and A. Tabarrok, Using Terror Alert Levels to Estimate the Effect of Police on Crime. Journal of Law and Economics, 48, Levitt, S.D Using electoral cycles in police hiring to estimate the effect of police on crime. The American Economic Review, 87(3), Levitt, S.D, Using Electoral Cycles in Police Hiring to Estimate the Effects of Police on Crime: Reply. The American Economic Review, 92(4), Mast, B.D., B.L. Benson, and D.W. Rasmussen. Entrepreneurial Police and Drug Enforcement Policy. Public Choice, 104(3-4), Mocan, H., and R. Gittings, Getting off Death Row: Commuted Sentences and the Deterrent Effect of Capital Punishment. Journal of Law and Economics, 46, McCrary, J., Using Electoral Cycles in Police Hiring to Estimate the Effects of Police on Crime: Comment. The American Economic Review, 92(4), Ross, A., Crime, Police, and Truth-in-Sentencing: The Impact of State Sentencing Policy on Local Communities. Regional Science and Urban Economics, 42, Shepherd, J., Police, Prosecutors, Criminals, and Determinate Sentencing: The Truth about Truth-In-Sentencing Laws. Journal of Law and Economics, 45, Sollars, D.L., B.L. Benson, and D.W. Rasmussen Drug enforcement and the deterrence of property crime among local jurisdictions. Public Finance Quarterly, 22(1),

24 Table 1: Cities with Low Priority Initiatives Jurisdiction Year Passed Enacted By Vote Percentage 1 Berkeley, CA 1979 Voter Initiative Not available. 2 Seattle, WA 2003 Voter Initiative Passed with 58% of the vote. 3 Oakland, CA 2004 Voter Initiative Passed with 65% of the vote. 4 Santa Barbara, CA 2006 Voter Initiative Passed with 66% of the vote. 5 Santa Cruz, CA 2006 Voter Initiative Passed with 64% of the vote. 6 San Francisco, CA 2006 City Council San Francisco Board of Supervisors passed the ordinance in an 8-3 vote. 7 Santa Monica, CA 2006 Voter Initiative Passed with 65% of the vote. 8 West Hollywood, CA 2006 City Council West Hollywood City Council passed the resolution in a 4-0 vote. 9 Eureka Springs, AR 2006 Voter Initiative Passed with 62% of the vote. 10 Missoula County, MT 2006 Voter Initiative Passed with 54% of the vote. 11 Denver, CO 2007 Voter Initiative Passed with 55% of the vote. 12 Fayetteville, AR 2008 Voter Initiative Passed with 66% of the vote. 13 Hawaii County, HI 2008 Voter Initiative Passed with 53% of the vote. 14 Hailey, ID 2010 Voter Initiative Passed with 54% of the vote. a 15 Kalamazoo, MI 2011 Voter Initiative Passed with 66% of the vote. 16 Tacoma, WA 2011 Voter Initiative Passed with 65% of the vote. 17 Ypsilanti, MI 2012 Voter Initiative Passed with 74% of the vote. Notes: Information was obtained from Marijuana Policy Project: and The Berkeley Daily Planet: 31/article/23788?headline=Legal-Limbo-for-Pot-Users-By-SUZANNE-LA-BARRE a The initiative passed with 51% of the vote in 2007, and again in 2008 with 54% of the vote. Due to a redaction by a district court judge, the measure did not officially go into effect until

25 Table 2: Summary Statistics Variable Full Sample Adopters Non-Adopters t-statistic Misdemeanor marijuana arrests per capita Murder rate Rape rate Assault rate Robbery rate Burglary rate Motor vehicle theft (MVT) rate Larceny rate Median age Percent male Percent black Percent white Percent with no HS Percent with some HS Percent with HS degree Percent some college Percent BA Percent college or higher Police officers per capita Notes: Misdemeanor marijuana arrests per capita, crime rates (murder, rape, assault, robbery, burglary, MVT, and larceny), and police per capita are all reported per 1,000 residents. The last column reports results for a test of if the mean between the adopters and non-adopters are statistically different. The t-statistic in the last column represents the statistic generated when testing if the two means are different. The values have the standard cutoff values as when determining statistical significance in a regression format. 25

26 Table 3: Impact of Low Priority Laws on the Log of the Number of Misdemeanor Marijuana Arrests per Capita (standard errors are reported in parenthesis) All Cities Cities with Population above 30,000 Cities with Population above 50,000 Low Priority Law in Effect * ** ** [0.159] [0.135] [0.137] Log Number of Police Officers 0.295*** 0.561*** 0.791*** [0.088] [0.162] [0.223] % Male *** 0.086** [0.005] [0.039] [0.024] Median Age *** 0.078*** 0.091*** [0.014] [0.013] [0.013] % Black ** *** [0.010] [0.018] [0.010] % White *** ** [0.008] [0.013] [0.007] % Less than HS *** 0.089** 0.104*** [0.010] [0.037] [0.017] % Some HS 0.051*** [0.015] [0.027] [0.047] %HS Grad *** [0.019] [0.049] [0.040] % Some college *** 0.137*** [0.015] [0.035] [0.052] % Associates Degree *** *** [0.035] [0.060] [0.089] % Bachelor s Degree [0.014] [0.048] [0.035] Observations 3,312 1,698 1,205 R-squared Notes: The dependent variable is the arrest rate per 1,000 residents. Arrest data was obtained from the California Justice Department and the police office data was obtained from the UCR. 26

27 Table 4: Impact of Low Priority Laws on the Log of the Number of Misdemeanor Marijuana Arrests per Capita (standard errors are reported in parenthesis) All Cities Cities with Population above 30,000 Cities with Population above 50,000 Low Priority Law in Effect * ** ** [0.177] [0.150] [0.153] Low Priority * City Council Interaction [0.395] [0.332] [0.337] Log Number of Police Officers 0.294*** 0.559*** 0.788*** [0.088] [0.162] [0.223] % Male *** 0.087** [0.005] [0.039] [0.024] Median Age *** 0.078*** 0.091*** [0.014] [0.013] [0.014] % Black ** *** [0.010] [0.018] [0.010] % White *** ** [0.008] [0.013] [0.007] % Less than HS *** 0.089** 0.104*** [0.010] [0.037] [0.017] % Some HS 0.051*** [0.015] [0.027] [0.047] %HS Grad *** [0.019] [0.049] [0.040] % Some college 0.105*** 0.137*** [0.015] [0.035] [0.052] % Associates Degree *** *** [0.035] [0.060] [0.089] % Bachelor s Degree [0.014] [0.048] [0.035] Observations 3,312 1,698 1,205 R-squared Notes: The dependent variable is the arrest rate per 1,000 residents. The indicator variable is an interaction between whether the city has a low priority law and if it was adopted by the city council. 27

28 Table 5: Impact of Low Priority Laws on the Log of the Number of Arrests per Capita for (standard errors are reported in parenthesis) Dependent Variable All Cities Cities with Population above 30,000 Cities with Population above 50,000 Felony Drug Arrests (0.109) (0.070) (0.064) Murder (0.185) (0.192) (0.196) Rape (0.149) (0.149) (0.143) Robbery (0.143) (0.111) (0.098) Assault (0.079) (0.048) (0.043) Burglary (0.114) (0.072) (0.069) Motor Vehicle Theft (0.144) (0.107) (0.099) Larceny (0.110) (0.063) (0.057) Notes: Each coefficient and standard error reported corresponds to a different regression. Each row represents a regression with a different dependent variable. All dependent variables are rates, and the regressions are run separately for the different population cutoffs. 28

29 Table 6a: Impact of Low Priority Laws on the Log of the Reported Crime Rate for Index I Crimes(standard errors are reported in parenthesis) Dependent Variable All Cities Cities with Population above 30,000 Cities with Population above 50,000 Murder (0.146) (0.148) (0.145) Rape (0.136) (0.108) (0.095) Robbery (0.103) (0.060) (0.057) Assault ** 0.119** (0.077) (0.051) (0.050) Burglary (0.077) (0.054) (0.049) Motor Vehicle Theft (0.092) (0.059) (0.056) Larceny (0.066) (0.046) (0.043) Notes: Each coefficient and standard error reported corresponds to a different regression. Each row represents a regression with a different dependent variable. All dependent variables are rates, and the regressions are run separately for the different population cutoffs. 29