Indeed, Sheriff’s Office officials were among the few public opponents of the LPI resolution, who noted concern regarding “inflexibility” among patrol officers.

fort worth police evidence.com login

Similar to our policing analysis, we conduct event-study analyses based on Eq. 2 but within the context of our prosecutor outcomes. We are interested in estimating the effect of the LPI on the charge rate, dismissal rate and rate of cases plead out across Los Angeles County.Footnote 27 Fig. 4 plots the county-wide monthly misdemeanor marijuana charge rate pre- and post- coefficients at the city-month unit of analysis. We observe no statistically significant increase or decrease in charge rates across the entire county except for in \(6{th}\) and \(7{th}\) months after the initiative, where charge rates are negative relative to the base month (July 2006). Similarly, Fig. 5 suggests that there are some statistically significant negative dismissal rates pre and post LPI (4 months pre and 6-7 months post) relative to the base month. Lastly, Fig. 6 showcases a null effect for the rate of cases plead out in Los Angeles county both pre- and post-LPI with a few exceptions (8 months pre and 11 months post).Footnote 28

We add additional city-level controls, male to female ratio, percent of high school graduates, percent of population White, Black, Asian and Other, median income and percent of females and males aged 65 and over in robustness analysis located in Appendix B, Table 8.

Section 3 of this paper discusses both of these data sets. Importantly for our analysis, while LPIs were adopted in two cities in Los Angeles County—Santa Monica and West Hollywood—we focus on the latter because, as described in Sect. 2, Santa Monica has its own police department, while West Hollywood is the only jurisdiction with an LPI that uses the law enforcement services of the LASD.

We hypothesized that law enforcement could potentially respond to such a nudge in the form of an LPI by reducing the number of arrests and citations for low-level marijuana offenses. But, we also note that law enforcement preferences could have a counterbalancing effect given that law enforcement actors might have stronger punitive preferences than members of a city council that issued an LPI. Furthermore, we expected prosecutors would be impacted by the LPI because of a shift in their incentives: if prosecutors expect a defendant to be more likely to be found not guilty by a jury, they should be more likely to push toward a plea. Alternatively, given that prosecutors are tasked with holding individuals accountable for violating the law and that low level marijuana offenses remain violations of the law post LPI, there remains an expectation for prosecutors to charge these cases in a similar way as pre-LPI.

Misdemeanor Marijuana Arrests Unincorporated Areas (excluding West Hollywood). Fig. 3 showcases event-study coefficients on misdemeanor marijuana arrests across months pre- and post- the LPI in Unincorporated Areas, excluding West Hollywood. We omit 12 months pre, the first month of the data, to be the omitted reference time period. Standard errors are clustered at the reporting district level

Within the LPI framework, the SDID application is as follows. First, the “synthetic control” estimation determines the city locations that most mimic the outcome of interest and re-weights those cities to ensure time trends are parallel pre-intervention. Second, the difference-in-difference estimation is applied to the re-weighted panel to measure the effect of the LPI on misdemeanor marijuana arrests, prosecutor charge rates, prosecutor dismissal rates and prosecutor plea rates. Another deviation from previous results is the unit of analysis. Due to the nature of census data reported at the decennial frequency as well as the noisiness of rate data, SDID is analyzed at the annual-city level where all cities in California are included in the potential donor pool. Similar to our prior analysis, locations directly within a 12 mile radius from the treated location are excluded from the analysis to avoid spillover effects. We believe SDID has a comparative advantage in estimating robust results as locations geographically distant from the treated location can also be used in creating the synthetic donor unit, further decreasing potential spillover effects.

Prosecution figures include a wider, 15 month window since it could take longer for the LPI to be observed in prosecution decisions.

To examine the effect of LPI’s on misdemeanor marijuana arrests we build on the work of DeAngelo et al. (2018),Footnote 14 but further refine the analysis to get closer to the causal effect of the LPI. We start by identifying RDs that are fully contained within West Hollywood, where the LPI was enacted. We also identify RDs that are partially contained in West Hollywood, but partially contained in a neighboring jurisdiction. In total there are 48 RDs that are either fully or partially contained in West Hollywood, which constitute 6.5% (167,435/2,542,640) of the observations in our data. While West Hollywood is a contract city, the RDs that are partially within West Hollywood and partially in a neighboring jurisdictions include both contract cities and unincorporated areas, which we will leverage in our analysis.

OpenEvidence

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Ko, H. (2018). Behavioral law and economics. In A. Marciano & G. Ramello (Eds.), Encyclopedia of Law and Economics. Springer.

We further capture the overall trend of crime across Los Angeles County by comparing the average number of misdemeanor arrests pre and post LPI. We find an average of 2,444 misdemeanor arrests pre LPI and 3,375 arrests post LPI across the county. We further subset our descriptive measurement to treated and untreated locations and find that West Hollywood had an average of 1,426 arrests while untreated locations had an average of 2,456 arrests pre LPI. In the post LPI period, West Hollywood has an average of 2,238 arrests and untreated locations had an average of 3,388 arrests. We further note that the increase in misdemeanor arrests in West Hollywood post LPI differs from the trend of misdemeanor marijuana arrests post LPI in the treated location as showcased in Table 3.

To further explore the locations that are contributing to the significant increase in misdemeanor marijuana arrests, we further explore arrest behavior by those locations where the LASD are contracted to perform policing services versus unincorporated regions where the LASD are statutorily mandated to act as the enforcement agency in Figs. 2 and 3, respectively. We exclude all arrests in West Hollywood to isolate the effect of the LPI on other regions within Los Angeles County.

We explore robustness checks to verify the results presented thus far. A natural extension to a difference-in-difference (DiD) analytical approach is a synthetic control (SC) approach, especially when the current framework estimates on a single treated location. A recent innovation in the empirical methodological literature is a method that couples the difference-in-difference estimation strategy with the synthetic control estimation strategy, the synthetic difference-in-difference (SDID) (Arkhangelsky et al., 2021). We explore the application of SDID within the LPI framework as one of the first empirical applications of this estimation strategy. Below we layout the empirical differences across the three noted strategies as well as the reasons SDID should be considered the most appropriate identification method within the LPI setting.

Our first set of results show that, after the introduction of the LPI, the number of misdemeanor arrests significantly rises throughout the regions of Los Angeles County that the LASD patrols with the exception of West Hollywood. To better understand the dynamics of what is driving this result, we explore heterogeneous results by the region where arrests are being made. The LASD patrols cities that are contracted to use its law enforcement services, such as West Hollywood, and unincorporated areas that do not have their own police department and are not part of any official city. We expect that LASD officers would not adjust their behavior in unincorporated areas, as no contractual arrangement exerts outside pressure on enforcement behavior in the unincorporated areas. Alternatively in contract cities, the LASD performs policing services on behalf of the contract city and may feel pressure to alter their behavior in response to the contract city’s issued guidelines. We find that there is a small effect of the LPI on misdemeanor marijuana arrests when comparing West Hollywood to neighboring contract cities, while there is a large decrease in the number of misdemeanor marijuana arrests in West Hollywood compared to unincorporated areas. Thus, contrary to our initial expectation, it appears that the LASD especially increased their arresting behavior in unincorporated areas that neighbor West Hollywood.

Though a preferential source for covariate characteristics is the American Community Survey (ACS) as this survey is reported more frequently, our treated location of West Hollywood is not located within the list of cities surveyed.

Parker, K. F., & Maggard, S. R. (2005). Structural theories and race-specific drug arrests: What structural factors account for the rise in race-specific drug arrests over time? Crime & Delinquency, 51(4), 521–547.

Convery, F., McDonnell, S., & Ferreira, S. (2007). The most popular tax in europe? Lessons from the irish plastic bags. Environmental and Resource Economics, 38, 1–11.

For the purposes of this analysis we focus exclusively on LASD since our main analysis will focus on one location (West Hollywood) under the LASD’s jurisdiction that imposed an initiative requiring the LASD to change their enforcement behavior. As such, we explain the difference between patrol locations within the LASD’s purview. If a jurisdiction is an officially incorporated city, then they are considered a contract city. Alternatively, if a jurisdiction is not part of an incorporated city, then it is considered an unincorporated area. Contract cities have a choice between providing their own police department (e.g., Compton) or paying for law enforcement services to be provided by the LASD (e.g., West Hollywood). Importantly, if the contract city hires the LASD to act as their enforcement agency, then a contract between the city and the LASD is drawn up. This contractual relationship could be a driving force behind enforcement behavior. Alternatively, unincorporated communities are not a part of any official city and so are governed by the Los Angeles County Board of Supervisors and have their laws enforced by the LASD. In unincorporated areas the LASD does not face any concern about losing a contract.

We accept that reasons why a prosecutor may charge, dismiss or offer a plea deal can be attributed to a myriad of reasons. Some reasons encompass specific offender-case-level facets, such as prior criminal history, offense severity, or the prosecutor’s belief of the offender’s risk to re-offend. We are less concerned with the interaction of prosecutor’s priors with the LPI, as this reform is a small low-level initiative in the grand scheme of prosecutor policy reforms. To better isolate the effect of prosecutor behavior, we do not focus on estimating the change in the levels of prosecutor charges, dismissals or plea offers but rather we estimate effects on the rate of charging, dismissing and plea take-up. By estimating the effect of the LPI on the rate of misdemeanor marijuana charges/dismissals/pleas we control for the fact that incoming misdemeanor cases are a function of arrests made within that particularly city. Comparatively, the rate of misdemeanor marijuana offenses charged is a decision made by the prosecutor.

Misdemeanor Marijuana Charge Rate. Fig. 4 showcases event-study coefficients on misdemeanor marijuana charge rates across months pre- and post- the LPI. We omit the month prior to the LPI and 15 months post as our reference omitted time groups. Standard errors are clustered at the city-level

See robustness analyses in the Appendix for a subset of dismissals pertaining to review prosecutor initiated dismissals.

Given the institutional composition of the criminal justice system in Los Angeles County, as well as the passage of the LPI, we intend to empirically examine the impact of the LPI on the behavior of institutional actors in the criminal justice system. We interpret the LPI as a non-binding guideline toward low-level marijuana offenses, which creates a unique environment to determine how different aspects of the criminal justice system, each with their own organizational preferences and pressures, responds to a specific type of nudging technique.

We explore SDID across four outcomes: incoming arrests for low-level misdemeanor marijuana offenses, low-level misdemeanor marijuana charge rates, low-level misdemeanor marijuana dismissal rates and low-level misdemeanor marijuana plea rates. Our goal in implementing SDID is to create a synthetic West Hollywood that emulates misdemeanor marijuana offense trends in the pre-treatment period, 2000-2006. Our estimated SDID average treatment on the treated (ATT) estimate is graphically showcased in Fig. 7 below. Each panel is estimated on data from CalDOJ censoring misdemeanor marijuana offenses that co-occur with other offense types. Indeed, a benefit of estimating the effect of LPI on CalDOJ data is two-fold. First, we observe each of the arresting department’s suggested charges for each person-date. Second, we observe arrests across cities other than those contained in Los Angeles County. A drawback in utilizing the CalDOJ data in this analysis is the lack of geographical granularity. We cannot observe or measure arrests at the reporting district level as was done in the analysis presented in Table 3. Each panel consists of three graphs. The top figure depicts the difference-in-difference trends. The middle figure displays the synthetic control trends. The bottom figure displays the SDID trends. The arrow represents the average treatment on the treated (ATT) estimate of the LPI in West Hollywood. Specifically, the red line within each panel represents the general trend of the synthetic West Hollywood unit from pre to post LPI, the dashed line describes the counterfactual trend, and the blue line represents the actual trend of the treated location. The counterfactual trend can be thought of similar to difference-in-difference methodologies. Since SDID ensures the synthetic unit and treated unit share parallel trends in the pre-period, we are certain the treated unit would have followed the same trend as the counterfactual had it not been treated. We can then re-anchor the trend the counterfactual had from the pre to post period (the dotted line) at an intercepting point on the treated, blue line. The final step is to compare the gap between the counterfactual dashed line and the real evolution of the treated location (blue line), which is denoted with the arrow.

DeAngelo et al. (2018) examines the effect of the LPI on law enforcement behavior by distance from West Hollywood, noting that the treatment effects increase with distance from West Hollywood.

To dissect the dynamics of what is driving our main result, we explore the relationship between the LASD and the region where arrests are being made. As noted above, the LASD patrols contract cities and unincorporated areas. Indeed, RDs in our control locations contain both contract cities and unincorporated areas. To determine if the new non-binding guidelines in West Hollywood had any impact on misdemeanor marijuana arrests elsewhere, we break our results apart by control locations that are either exclusively in contract cities or exclusively in unincorporated areas. Since the LASD may be under contractual pressure in contract cities, but no such dynamics exist in the unincorporated areas, we hypothesize that the impact of the LPI could have a different impact on each of these regions. We explore this in Table 4. Columns I-II compare treated West Hollywood to cities where LASD was contracted, labeled below as Contract, while columns IV-VI compare treated West Hollywood to unincorporated regions, labeled below as Unincorporated.

Table 6 presents OLS estimates on the effect of the LPI in West Hollywood after the initiative across our three outcomes of interest. The OLS estimation includes city fixed effects in column (I), city and year fixed effects (FE) in column II and city FE, year FE and a control for the total marijuana arrests within a month and city in column III. As the three outcome variables examined are rates bounded between 0 and 1, we interpret our results through a linear probability model. In the first set of results, our coefficient of interest (\(LPI \times Post\)) is consistently statistically significant and suggests a smaller charge rate of 0.479% in West Hollywood compared to other cities in the sample post LPI, which translates to a 32% reduction from the mean. Our second set of results suggest there is no statistically significant change in the rate of overall misdemeanor marijuana case dismissals in West Hollywood compared to other cities in the sample after the passage of the LPI.Footnote 24 As is discussed in Sect. 3.3, these estimates suggest that the number of marijuana misdemeanor charges decreased as a function of fewer arrests made by officers in West Hollywood post-initiative.Footnote 25 Of the arrests made, prosecutors do not appear to change their behavior by increasing or decreasing the rate of overall misdemeanor marijuana charge dismissals. Similar to our posited hypotheses, we observe unchanged prosecutor practices in the form of no statistically significant reduction or increase in overall case dismissal rate after the passage of the LPI. Our final panel of results indicate a statistically significant decrease in the rate of cases plead out in West Hollywood post LPI, representing a 30% decrease from the mean. These results are consistent with the notion that the people of West Hollywood have deemed misdemeanor marijuana cases to be a low priority charge that should not receive harsh penalties, potentially motivating defense attorneys to advise their clients against accepting a plea agreement.Footnote 26

Ross, A., & Walker, A. (2017). The impact of low-priority laws on criminal activity: Evidence from California. Contemporary Economic Policy, 35(2), 239–252.

An alternative explanation for punitiveness by an officer could be prior experience as a victim. See Simmler et al. (2021) for more on this.

MCP’s DataScape™ analytics solution brings native data from multiple sources into a data lake in real-time, including CAD systems. Implementing MCP’s DataScape analytics solution enables agencies to:

As discussed above, we subset dismissals to include only review prosecutor initiated dismissals as charges come in the door, these include reasons such as lack of evidence and for the interest of justice, in the appendix. We include a more extensive set of dismissals observed during every step that a case is handled by the prosecutor in the main results. The main result dismissal rates include case dismissals that are likely attributable to prosecutors dismissals at later points of the case or trial.

As discussed in this section, this work deviates from the previous research in several substantive ways. First, when examining arrest data, we employ a much more refined geographic measure for the treatment. Also, the comparison group is geographically limited to reporting districts immediately adjacent to West Hollywood or, in some instances, partially contained in West Hollywood and partially outside of West Hollywood. Perhaps most substantively, we incorporate data from the Los Angeles District Attorney’s Office to examine prosecution behavior in conjunction with arresting behavior. We further extend on previous research by applying a novel empirical identification strategy, synthetic difference-in-differences as discussed in the robustness analyses in Appendix A.

EvidenceAct

The second data set we utilize is the Criminal Offender Record Information (CORI) from 2000-2010, which is managed by the California Department of Justice (CA DOJ). The raw data are available at the individual step event unit, where individuals show up during each instance of their interaction with the criminal justice system. For instance, each individual’s reason for arrest, arresting agency and case disposition after a court action is recorded. In this analysis we subset our data to Los Angeles county, aggregating data to the unique city-week level.Footnote 11 We leverage the granularity in the prosecutor data by aggregating misdemeanor marijuana offenses to the weekly-level. We conduct our analysis of prosecutor data at the weekly-level to avoid mis-attribution due to the process time lag between the time of arrest and initial intake of an arrest to the prosecutor office. While arrests occur at a specific date and time, a number of factors could impact the length of time until a case arrives at the prosecutor’s office.Footnote 12 Due to the potential time delay in cases arriving at the prosecutor’s office, we aggregate data in an attempt to mitigate potential mis-attribution when cases actually reach the prosecution stage.

Romaniuc, R. (2016). What makes law to change behavior? An experimental study. Review of Law and Economics, 12(2), 447–475.

These rates are constructed as follows: (1) number of misdemeanor marijuana charges divided by total misdemeanor marijuana arrests, (2) number of misdemeanor marijuana cases dismissed divided by total misdemeanor marijuana charges and (3) number of misdemeanor marijuana charges plead out divided by total misdemeanor marijuana charges.

The following section extends our previous results by examining the effect of the LPI on prosecutorial behavior. Since the LPI is intended to change policing behavior, less is known regarding the initiative’s effect on altering prosecutor decisions. We now explore if changes in policing behavior extend to prosecutorial behavior by examining whether misdemeanor marijuana charges from West Hollywood are penalized more stringently compared to misdemeanor marijuana charges brought forth by other cities in Los Angeles County.

Athey, S., & Imbens, G. W. (2017). The state of applied econometrics: Causality and policy evaluation. Journal of Economic perspectives, 31(2), 3–32.

How can we support your mission? From design and procurement to building and management, our national team of experts is here to help…because the mission matters.

As noted in previous literature, synthetic control is arguably the most influential innovations in policy reform evaluation (Athey and Imbens, 2017). This sentiment is highlighted within the current context where our “treatment”, the low priority initiative, is evaluated on a singular treated “unit”, West Hollywood. Another comparative advantage in synthetic control is the robustness gained by comparing the treated unit to another unit that matches on pre-treatment exposure trends and other relevant covariate characteristics. In doing so, this strategy relaxes the strong assumption required within difference-in-differences (parallel pre-trends between treated and control units). If parallel trends are established in the DiD context, we further must believe pre-trends in the treated unit would have evolved similar to the control unit in the absence of intervention. In contrast, the synthetic control design relies on computing a “synthetic unit” composed of weighted untreated “donor” units that match on the treated unit’s pre-treatment outcome. SDID integrates the two approaches by first re-weighting untreated “donor” units to ensure time trends for the outcome are parallel pre-intervention and subsequently measure DID on the revised re-weighted panel. SDID is an attractive alternative method due to the data-driven nature of selecting time weights rather than allowing each time unit to have equal weight in the pre-period, such as in DID, or over-weighting the last pre-treatment period as is typically done in event study analysis (Arkhangelsky et al., 2021).

There are a select number of cities that are out of Los Angeles Prosecutor’s office jurisdiction which are excluded from the following analysis. The list of cities are Burbank, Hawthorne, Hermosa Beach, Inglewood, Long Beach, Los Angeles, Pasadena, Redondo Beach, Santa Monica and Torrance.

DeAngelo, G., Narender, R. & Romaniuc, R. Nudging law enforcement: evidence from low priority initiatives. Eur J Law Econ 58, 321–354 (2024). https://doi.org/10.1007/s10657-024-09816-w

Evidenceexample

Law enforcement in Los Angeles County is largely broken into three groups. First, the Los Angeles Police Department (LAPD), with its approximately 10,000 sworn officers, patrols all of the city of Los Angeles and most Metro trains and buses. Second, the Los Angeles Sheriff’s Department (LASD), which also has approximately 10,000 sworn officers, patrols all unincorporated areas, contract cities, and runs the jails in Los Angeles County. Finally, there are more than 70 independent law enforcement agencies (e.g., Santa Monica Police Department) in Los Angeles County that are not associated with either the LAPD or LASD.

We use two data sets in our analysis. The first data set utilizes incident-level arrests from the Los Angeles Sheriff’s Department. The second data set utilizes prosecution data obtained from the California Department of Justice, which contains information about the final disposition of all charges brought against a citizen. We discuss both of these data sets independently.

West Hollywood City Council has no authority to compel the LASD to enforce the LPI since it was passed by the city council as a resolution, which is not a law and is not legally binding. The city council member who proposed the resolution acknowledged this, stating that the resolution should “send a message to law enforcement that they should focus on more serious crimes.”

Interestingly, misdemeanor marijuana arrests significantly increased in contract cities and persisted, while unincorporated areas had no effect until months 5–9 after the initiative which eventually returned to a zero effect in months 10–12 post LPI. This result is in contrast to the results presented in Table 4, but speaks to the tension between non-binding guidelines issued by the city council, law enforcement biases, and the dynamics between contract cities and the LASD. It appears that law enforcement biases and the contractual dynamics between the LASD and contract cities are a stronger driving force in explaining the changes in misdemeanor arrest behavior. As noted by the Drug Reform Coordination Network, law enforcement announced their opposition to being told how to perform their job.Footnote 20 Moreover, Sheriff BacaFootnote 21 highlighted the importance of the LASD satisfying the terms of the relationship with each contract city. Taken together, it appears that the passage of the LPI resulted in the LASD not merely maintaining the number of misdemeanor marijuana arrests in non-LPI contract cities, but perhaps sending a signal that they would not be shirking on their contractual responsibilities.

DeAngelo, G. J., Ross, A., & Gittings, K. R. (2018). Police incentives, policy spillovers, and the enforcement of drug crimes. Review of Law & Economics, 14(1), 1–29.

A plea offer is extended at the sole discretion of a prosecutor, which we also estimate in our prosecutor analysis. It is true that when aggregating from the single individual case-level to the city-week level we lose the identifying characteristics associated with the individual facing prosecution. However, we are not concerned with losing specific case-level information so long as the guiding principles a prosecutor refers to when deciding to dismiss, plea, or charge a case remain unchanged by the LPI. Further if we believe the types of offenses the DA’s office attends to remain unchanged by the local initiative, any observed differences in outcomes can be associated with a shift in overall prosecution behavior in response to the LPI.

We further explore prosecutor behavior by examining dismissal and plea take up rates. Within the CalDOJ data, a total of 913 disposition codes related to each incoming arrest are recorded. Of these disposition codes, a subset are related to the arrest made, others to prosecutor decisions, and finally decisions resulting from a court disposition. A set of offense codes are attributed to dismissals conducted by review prosecutors as arrests come in the door, which we further explore in Appendix Table 9. These dismissals will not include a dismissal made by other actors within the criminal justice system or dismissals made by prosecutors at later steps within a criminal case proceeding.

We remove 0.5% of our data in this exercise, our reported observation size remains constant in Table 8 since our data is aggregated to the week by year level and so the number of cities captured in our week-year panel remain constant.

Feld, L. P., & Tyran, J.-R. (2006). Achieving compliance when legal sanctions are non-deterrent. Scandinavian Journal of Economics, 108(1), 135–156.

Figure 7, panel (a) displays the effect of the LPI on misdemeanor marijuana arrests. The effect size is consistent with the previously stated difference-in-differences results in Table 3. Panels (b), (c) and (d) depict the effect of the LPI on misdemeanor marijuana charge rates, dismissal rates and plea rates. Each of these estimates also retain consistency with previously measured difference-in-difference analysis in Table 6. A major takeaway from the research introducing the SDID design is the nature of DID and SC methods to commonly overstate the treatment effect while SDID produces more conservative estimates for the magnitude of the ATT. Interestingly, within our current framework, SDID reports similar magnitudes to DID across all estimations, with one notable difference. Namely, the dismissal rates show an increase when estimated using SDID.

DeAngelo et al. (2018) argue that areas that did not adopt an LPI might signal to the police that they want more enforcement, which may explain the increase in these areas.

As this data includes unique offense descriptions for an offender’s arresting and charged offense type, we are able to further subset our data to identify low-level misdemeanor marijuana charges as defined by the offense description. For the purposes of this analysis, we are able to identify three important features pertaining to each arrest that the LADA would handle. First, we can construct a city-week aggregate misdemeanor marijuana charge rate, which is constructed as the number of misdemeanor marijuana charges divided by the total number of incoming misdemeanor marijuana arrests. Second, conditional on the LADA receiving suggested charges from the LASD, we construct the rate at which misdemeanor marijuana charges are dismissed. Mechanically, this is the fraction of dismissed misdemeanor marijuana charges divided by the total number of misdemeanor marijuana charges at the city-week unit of analysis. Finally, conditional on the LADA pursuing the misdemeanor marijuana charges, we construct a measure of the rate at which plea deals are reached, which is the fraction of pursued misdemeanor marijuana charges reaching a plea agreement divided by the total number of pursued misdemeanor marijuana charges. Table 1 displays descriptive statics for each of these constructed measures, showcasing the average and standard deviation in parentheses below, broken apart by pre and post LPI across the full sample, control locations and treated location.

MyEvidence

The following cities are excluded from our data set as LADA does not have jurisdiction over misdemeanor offenses: Burbank, Hawthorne, Hermosa Beach, Inglewood, Long Beach, Los Angeles, Pasadena, Redondo Beach, Santa Monica and Torrance.

Friebel, G., Kosfeld, M., & Thielmann, G. (2019). Trust the police? Self-selection of motivated agents into the German police force. American Economic Journal: Microeconomics, 11(4), 59–78.

To more generally explore whether the introduction of the LPI had an effect even beyond the RDs directly surrounding West Hollywood, we broaden our analysis to examine the effect of the LPI on all of Los Angeles County. To conduct this analysis we estimate a basic event study plot with the unit of analysis at the monthly level. Specifically, we estimate Eq. 2:

Sheriff Baca was the active presiding Los Angeles County Sheriff during the entirety of this analysis. See https://projects.scpr.org/timelines/sheriff-lee-baca/

Dharmapala, D., Garoupa, N., & McAdams, R. H. (2016). Punitive police? Agency costs, law enforcement, and criminal procedure. The Journal of Legal Studies, 45(1), 105–141.

Of note, Steve Cooley believed “undermining those laws via their ordinance powers is counterproductive, and quite frankly, we’re ignoring them” when expressing his plan to undermine council member ordinance to lift bans on prosecuting charges related to medical marijuana dispensaries (Hoeffel, 2009).

Arkhangelsky, D., Athey, S., Hirshberg, D. A., Imbens, G. W., & Wager, S. (2021). Synthetic difference-in-differences. American Economic Review, 111(12), 4088–4118.

Misdemeanor Marijuana Arrests All of Los Angeles County. Fig. 1 showcases event-study coefficients on misdemeanor marijuana arrests across months pre- and post- the LPI in Los Angeles County. We omit 12 months pre, the first month of the data, to be the omitted reference time period. Standard errors are clustered at the reporting district level

This research examines the effect a nudge, in the form of non-binding guidelines, on police behavior and the prosecutor’s decision-making. Previous research in behavioral law and economics looked at the effects of this type of nudges on citizens’ behavior in the context of public good games (Galbiati and Vertova, 2008; Dal Bo and Dal Bo, 2014; Romaniuc, 2016). Our study is the first, to the best of our knowledge, to investigate how law enforcement agents react to the introduction of a nudge, in the form of a low priority initiative that entails no formal sanctions in case of non-enforcement. LPIs mandate that misdemeanor marijuana offenses be considered as the lowest priority for law enforcement. An LPI is passed by city councils and is a non-legally binding guideline. Therefore, their enforcement greatly depends on the law enforcement agents’ preferences to comply with the initiative. We should note, however, that some aspects of LPIs may not strictly satisfy the standard definition of nudges. First, given that an LPI may be the result of constituents’ preferences, law enforcement agents may expect that, by deviating from it, they will expose themselves to non-financial sanctions in the form of distrust from the public. Although we cannot rule out this possibility, such spillover effects are very unlikely. Second, if the Sheriff’s department does not comply with an LPI, the municipality may threaten to provide police services internally. However, such a threat may be perceived as non-credible given the substantial cost of providing police services internally.

In the prosecutorial setting, charges are brought forth by different arresting agencies across the county and prosecuted by the Los Angeles county district attorney’s office. As our unit of treatment is at the city level, we aggregate cases brought forth in each city i within a given week t to be our unit of analysis. Of the cases brought forth, a select number of cases are charged, and of these cases a subset will be dismissed, plead out, or result in a conviction. We follow suit in the California Department of Justice (CA DOJ) definition of charges, dismissals and pleas when identifying these case outcomes in the CORI data.Footnote 22

Tables 8, 9, 10 and 11 expand on original results with the inclusion of month and city-year control characteristics. Undoubtedly, the use of two-way fixed effects is a widely practiced application when estimating the effect of a “shock” or policy change within a panel data setting. Particularly within the LPI framework, we include either a combination of reporting district and year fixed effects when measuring arrest outcomes or city and year fixed effects when measuring prosecution outcomes. Arguably, the addition of geographical and temporal fixed effects is considered the standard application within panel data in light of there being unobserved characteristics within the geographical unit that affect the outcome in a meaningful way. Put differently, geographic and temporal fixed effects control for measured characteristics the researcher believes impact the outcome while also controlling for unmeasured characteristics that also impact the outcome. Furthermore, hand picking the control covariates the researcher believes are the most pertinent when measuring the outcome leads to over weighting the importance of these observable characteristics which are limited to how often this data is sampled and what variables are measured in survey data. Though it can be argued that the selection of specific covariates based on institutional and theoretical framework is the preferred approach. To ensure previously reported estimates are robust when including a vector of covariates, we report columns (3) and (6) in tables 8, 9, 10 and 11. Our full set of controls considered are gathered from the decennial Census encompassing the years 2000 and 2010.Footnote 31 We include an exhaustive vector of city-level covariates that were available across our panel including ratio of males to female, percent of population that received a high school education, percent of Black population, percent of White population, percent of Asian population, percent of Other population, median income, percent of female population ages 65 and over, and percent of male population ages 65 and over. Table 7 breaks apart these city demographics by pre and post LPI periods across the full sample, control cities and treated city.

To be clear, our outcome measure identifies the probability that an arrest involves a misdemeanor marijuana offense. This is a proxy for an offender’s conditional probability of being arrested for committing a misdemeanor marijuana offense.

Misdemeanor Marijuana Dismissal Rate. Fig. 5 showcases event-study coefficients on misdemeanor marijuana dismissal rates across months pre- and post- the LPI. We omit the month prior to the LPI and 15 months post as our reference omitted time groups. Standard errors are clustered at the city-level

Table 2 displays the average number of misdemeanor marijuana arrests for the full sample, West Hollywood, contract cities (excluding West Hollywood), and unincorporated areas before and after the introduction of the LPI.Footnote 17 We also examine the difference in misdemeanor marijuana arrests and perform a t-test in the third column. The number of misdemeanor marijuana arrests significantly rises after the introduction of LPI throughout the regions of Los Angeles County that the LASD patrol, however they do not significantly rise in West Hollywood. Both contract cities and unincorporated areas experience an increase in misdemeanor marijuana arrests by approximately the same amount.

While the unconditional means display interesting changes in the number of misdemeanor marijuana arrests before and after the introduction of the LPI, they could be masking unobserved variation in citizen and law enforcement behavior over time. In Table 3 we examine the effect of the LPI on misdemeanor marijuana arrests using a fixed effects OLS model. In column I we include RD fixed effects, while column II includes the gender and ethnicity controls listed above. Finally, in column III we include RD and year fixed effects as well as gender and ethnicity controls.

The effect of non-binding guidelines (nudges) on citizens’ compliance with law has been studied theoretically and empirically in behavioral law and economics. Less is known about the impact of non-binding guidelines on the behavior of law enforcement agents and prosecution. Our work fills this gap. We study whether non-binding guidelines affect law enforcement and prosecution practices when the guidelines are not necessarily aligned with legal actors’ preferences. Our empirical analysis focuses on the impact of the adoption of a low priority initiative (LPI) on police and prosecutor behavior in Los Angeles County. Our results suggest that following the introduction of an LPI there is a rise in the number of misdemeanor arrests, but not in the rate that misdemeanor marijuana offenses are dismissed. We conclude that law enforcement preferences have a counterbalancing effect given that police officers may have strong punitive preferences. Prosecutors do not appear to respond to non-binding guidelines as evidenced by no change in the rate of dismissing cases.

The first data set includes all arrests obtained from the Los Angeles Sheriff’s Department (LASD) from 2000-2007. Since these are arrest-level data, we can identify low-level, misdemeanor marijuana offenses, which are precisely the offenses that the LPI initiative aimed to prevent from being issued. This data is also geo-coded so that we can determine the jurisdiction where the arrest occurred. Also, the police reporting district (RD) where the arrest occurred is coded into this data. Lastly, we can determine if the jurisdiction is a contract city or an unincorporated community.

Gaston, S. (2019). Producing race disparities: A study of drug arrests across place and race. Criminology, 57(3), 424–451.

There is also a political force at work, although the effect is more indirect. Since the county sheriff is an elected position, it could be that the sheriff aims to comply with potential voter’s desires to increase their likelihood of re-election.

Galbiati, R., & Vertova, P. (2008). Obligations and cooperative behavior in public good games. Games and Economic Behavior, 64, 146–170.

Importantly for our analysis, while the LASD have law enforcement jurisdiction throughout Los Angeles County, many jurisdictions have their own police force (e.g., Long Beach, Los Angeles, Pomona, and Santa Monica) and so the LASD does not patrol these regions. Given that Santa Monica has their own police department, the only jurisdiction that the LASD patrol that implemented an LPI is West Hollywood. Thus, throughout our analysis, we will only be examining jurisdictions where the LASD patrol and West Hollywood will be the only “treated” location in our analysis.

AxonEvidencelogin

The low priority initiative (LPI) mandates that minor, misdemeanor marijuana possession offenses be considered the lowest priority for the law enforcement agency with jurisdiction.Footnote 6 The mandate is only intended to affect offenses where marijuana was intended for adult personal use, which does not include possession or selling of marijuana to minors. Also, the LPI only applies to the private use of marijuana. So, offenses committed on public property are not impacted by the LPI.

Unlike law enforcement in Los Angeles County, there is a single prosecutor’s office with an elected district attorney that is charged with prosecuting criminal conduct. Indeed, the Los Angeles District Attorney’s Office (LADA) was lead by the elected district attorney, Steve Cooley, for the entirety of our analysis.Footnote 5 There are nearly 1,000 deputy district attorneys that prosecute felony crimes throughout Los Angeles County and misdemeanor crimes in unincorporated areas of the county and in all cities except Burbank, Hawthorne, Hermosa Beach, Inglewood, Long Beach, Los Angeles, Pasadena, Redondo Beach, Santa Monica and Torrance. Thus, the prosecutor’s office has jurisdiction over West Hollywood and most of the surrounding regions.

As previously noted, there are three outcomes of interest: misdemeanor marijuana charge rates, misdemeanor marijuana dismissal rates and the rate of misdemeanor marijuana cases plead out.Footnote 23 The coefficient of interest, \(\beta _3\), is a difference-in-difference estimate that explains the effect of the LPI on the rate of misdemeanor marijuana cases charged, dismissed and plead out. To account for unobserved time-varying changes across cities, we include year controls, \(\tau _y\), and to absorb time-invariant variation across cities, we include city controls, \(City_c\). We estimate various specifications in the Appendix section Tables 8, 9, 10 and 11 including month fixed effects and a vector of city-level characteristics and find consistent results to our main findings. The magnitude of our observed estimates are strongest in our model including a vector of control characteristics, discussed extensively in the appendix section. Provided that our location level fixed effects account for both observed and unobserved variation across cities, we retain these estimates within our main findings.

Since the outcome variable in this analysis is an indicator variable for whether an arrest involved a misdemeanor marijuana arrest, we are utilizing a linear probability model. The coefficient of interest (LPI \(\times \) Post) is consistently statistically significant, indicating that the number of misdemeanor citations fell by approximately 50% (0.008/0.015) in West Hollywood after the passage of the LPI.Footnote 18 As noted in DeAngelo et al. (2018), however, this result is driven by law enforcement increasing arrests for misdemeanor marijuana arrests in regions immediately adjacent to West Hollywood, which drives the main result.Footnote 19

Variance in the length of time between an arrest being made and the case arriving in the prosecutor’s office could be due to many, often unobserved factors. Examples of such factors include case backlog in the prosecutor’s office, shortages of prosecution staff and/or court room shortages, a delay in the time taken to collect evidence by the law enforcement agency, etc.

We are able to utilize this low priority initiative as an opportunity to examine how law enforcement and prosecutors respond to a specific type of nudging in the form of non-binding guidelines. This analysis empirically examines the change in arresting behavior by law enforcement using a difference-in-difference framework and find that following an LPI, law enforcement agents engage in making more arrests in areas that neighbor the region that passes the LPI.Footnote 29 We also identify that financial contracts appear to play a driving role in determining how law enforcement make arrests in the aftermath of the passing of an LPI. We also explore changes in prosecution behavior, finding that prosecutors filed fewer charges in the treated location, while overall dismissals in West Hollywood remained unchanged. Plea bargain rates for low-level misdemeanor charges fell, which could be a product of changes in strategies that the defense pursues, especially given that the city council has expressed a desire to not punish such offenses. To better disentangle the weight of upstream arrests on downstream prosecutorial behavior, we add a third column to control for total misdemeanor marijuana arrests within a city-month. This control yielded robust estimates. Though a major limitation of this control is that it is derived from the second administrative data source, it suggests that the treatment effect for charge rates can be partially explained through prosecutor behavior.Footnote 30 However, the prosecutor’s office did experience changes in the number of low-level marijuana arrests made by LASD and possibly a change in defense attorney’s strategies. Further, our extension in estimating the effect of the LPI under the SDID framework showcases that our main results are consistent in magnitude and direction. By implementing SDID, we ensure that the results showcased are not present due to chance and allow the relaxation of the parallel trends assumption. We can therefore provide greater confidence in the nature of our estimates to be estimating the effect from the LPI reform.

The relatively large standard errors and insignificant estimates are likely attributable to pleas being observed at low frequencies in our data, making it difficult to precisely estimate changes in cases plead out.

AxonEvidence

Nix, J., Pickett, J., & Wolfe, S. (2020). Ltesting a theoretical model of perceived audience legitimacy: The neglected linkage in the dialogic model of police-community relations. Journal of Research in Crime and Delinquency, 57(2), 217–259.

To test this hypothesis, we run a separate set of estimates where our outcome of interest is the rate of misdemeanor marijuana charges resulting in a guilty verdict divided by total misdemeanor marijuana charges. We find that the rate of guilty convictions decreased by 39% post LPI in West Hollywood compared to other cities in the sample, a 121% decrease from the mean. This finding suggests that post- LPI, misdemeanor Marijuana infractions in West Hollywood result in a guilty plea at much smaller rates.

MCP is the industry leader in CAD procurement and implementation. We strive to help agencies improve their public safety CAD systems by:

Misdemeanor Marijuana Arrests Contract Cities (excluding West Hollywood). Fig. 2 showcases event-study coefficients on misdemeanor marijuana arrests across months pre- and post- the LPI in Los Angeles County, excluding West Hollywood. We omit 12 months pre, the first month of the data, to be the omitted reference time period. Standard errors are clustered at the reporting district level

Another potential concern when interpreting our results is the nature of misdemeanor marijuana offenses occurring concurrently with other offense types. A possible scenario could be that our results are conditional on whether a misdemeanor arrest occurs coupled with other offense types. We need not be concerned with the coupling of offense types so long as the nature of the types and frequency of co-occurring offenses remain unchanged after the introduction of the LPI. To ensure our results are not attributed to the “stacking” of offenses or types of co-occuring offenses post LPI, we censor our sample to include arrests where the only offense type is misdemeanor marijuana offenses in columns (3-6). We find that our estimates retain magnitude, direction and remain statistically significant, showcasing that upon removing arrests where misdemeanor marijuana arrests co-occur with other offense types our results are robust.Footnote 32

Rengifo, A. F., Slocum, L. A., & Vijay, C. (2018). From impressions to intentions: Direct and indirect effects of police contact on willingness to report crimes to law enforcement. Journal of Research in Crime and Delinquency, 56, 1–39.

Open access funding provided by SCELC, Statewide California Electronic Library Consortium. Open access funding provided by SCELC, Statewide California Electronic Library Consortium. Open access funding provided by SCELC, Statewide California Electronic Library Consortium. There is no funding associated with this research.

Dickinson, D., Masclet, D., & Villeval, M.-C. (2015). Norm enforcement in social dilemmas: An experiment with police commissioners. Journal of Public Economics, 126, 74–85.

Simmler, M., Stempkowski, M., & Markwalder, N. (2021). Punitive attitudes and victimization among police officers in Switzerland: An empirical exploration. Police Practice and Research, 22(2), 1191–1208.

Ambrosino, A., Faralla, V., & Novarese, M. (2018). Nudge. In A. Marciano & G. Ramello (Eds.), Encyclopedia of law and economics. Springer.

Dal Bo, E., & Dal Bo, P. (2014). “do the right thing:’’ The effects of moral suasion on cooperation. Journal of Public Economics, 117, 28–38.

LPIs present one way that local communities can express their desire for legal changes without going through the formal channels of changing laws. In this way, LPIs can be thought of as non-binding guidelines, and used to evaluate the effect of a specific form of nudging on the decision-making actors in the criminal justice system. And while criminal justice leadership might believe that it is necessary to comply with an LPI for financial and political reasons, such pressure might not impact patrol officers or line prosecutors.Footnote 10

The LASD is the agency most directly impacted by the passage of the LPI. As noted above, West Hollywood contracts law enforcement services from the LASD. As such, we expect that the LASD would be the criminal justice agency most likely to respond to non-binding guidelines by reducing the number of arrests or citations associated with low-level marijuana offenses. However, it could also be the case that law enforcement preferences could go against the non-binding guidelines if these imply being less punitive. One reason is that police agents may have a special taste for punishment, as it has been shown experimentally by Dickinson et al. (2015) and Friebel et al. (2019). Moreover, it is possible that marijuana dispensaries opening in West Hollywood prior to LPI further influenced LASD to not prioritize misdemeanor marijuana offenses.Footnote 13 We are unable to disentangle the degree to which medical dispensaries influenced recreational marijuana offenses which is a limitation to our current analysis.

It is our prior that prosecutor behavior can be impacted directly and indirectly. First, prosecutors could choose to follow the non-binding guidelines by complying with the LPI. Second, there may be a strategic reason for prosecutors to adjust to the implementation of an LPI. Prosecutors may be more likely to push toward a plea if they expect that a jury is more likely to find a defendant not guilty following the implementation of an LPI. Third, there may be fewer cases arriving at the DA’s office if officers are making fewer marijuana arrests.

where \(Pre_{i}\) and \(Post_{i}\) are indicators for the 12 periods leading up to and after the passage of the LPI, respectively. We plot the county-wide monthly marijuana arrest coefficients in Fig. 1, where the unit of analysis is an RD-month. It is evident that, in the aftermath of the passage of the LPI, misdemeanor marijuana arrests grew by, on average, 0.2–0.3 arrests per month in an RD. Given that the average number of misdemeanor marijuana arrests per RD-month is 0.56 in the pre-period, it appears that the number of misdemeanor marijuana arrests grew by approximately 50%, on average, in Los Angeles County after the passage of the LPI.

In today’s complex, mission-critical communications environment, every network, system, and application impacts the entire emergency-response process. Because emergency response begins in the emergency communications center (ECC), every agency needs a team of experts who have a deep understanding of computer aided dispatch (CAD), not only the system and the applications that operate on it, and the data that it generates.

MCP’s provides comprehensive, end-to-end support includes conducting needs and feasibility assessments, developing technical specifications and procurement documents, facilitating the procurement process, and implementing and managing the new solution. If you have a public-safety technology upgrade/replacement project on the horizon, contact MCP to learn more.

Due to the infrequent occurrence of misdemeanor marijuana arrests within our treated location, we measure the number a similar offense (public intoxication) arrests to draw a comparison of the relative magnitude of our treatment effect. On average there are 0.444 public intoxication arrests prior to the marijuana LPI and 1.298 weekly average arrests after the LPI throughout Los Angeles County, which amounts to a 180% increase. Within West Hollywood, there were, on average, 0.737 weekly arrests before the LPI and 2.132 weekly average arrests after the LPI, which amounts to a 187% increase. This indicates that a similar offense did not decline, but rather increased after the LPI, suggesting that our treatment effects are isolated to the LPI.

Synthetic Difference-in-Differences. Panel (a) estimates the effect of the LPI on arrest outcomes while Panels(b-d) estimate the effect of the LPI on misdemeanor marijuana charge rates, dismissal rates and plea rates. Within each panel, a comparison of difference-in-difference, synthetic control and synthetic difference-in-differences is estimated on the effect of the LPI at the annual-city level

Importantly for our analysis, except in rare circumstances, the prosecutor’s office is at the mercy of law enforcement agencies, as they can only make charging decisions on cases where law enforcement has made an arrest. As such, the cases that are brought to the LADA have already been impacted by law enforcement response to LPIs. Since the LADA does not have any explicit contract with specific regions within Los Angeles County to perform prosecution services, the response of the office is purely to the LPI and potential re-election pressure that the elected district attorney faces.

Santa Monica and West Hollywood were the only municipalities that adopted LPIs in Los Angeles County. Since Santa Monica has its own police department and is not prosecuted by the LADA, we cannot compare citations and prosecution behavior in Santa Monica to other locations, since the law enforcement and prosecution agencies differ.Footnote 7 Since West Hollywood is a contract city that employs the LASD for policing services and falls under the jurisdiction of LADA, we can examine the impact of the LPI in West Hollywood on changes in LASD enforcement and LADA prosecution behavior within West Hollywood and in neighboring jurisdictions that did not implement an LPI. As noted above, the relationship between the city of West Hollywood and the LASD is governed by a somewhat complicated contractual structure. On the one hand, the LASD is a law enforcement agency and the LPI is not a legal change that requires that the LASD alter their behavior.Footnote 8 On the other hand, West Hollywood could end their contract with the LASD if the agency does not honor West Hollywood’s request to make misdemeanor marijuana arrests the lowest priority effort of the law enforcement agent’s activities.Footnote 9 Similarly, the LPI should not impact prosecution behavior by the LADA since it does not change the laws, but to the extent that the LADA responds to non-binding guidelines, we can examine this behavior.

Our second set of results examine the behavior of the prosecutor’s office in response to the LPI. Unlike the LASD, which has contractual arrangements with some parts of Los Angeles County and is, by law, to perform policing services in other parts of the county, the prosecutor’s office provides prosecution services to the entire county.Footnote 3 As such, the prosecutor’s office does not face the same contractual pressures that the sheriff’s department faces. Our difference-in-difference analysis finds that the fraction of misdemeanor marijuana charges fell, which is a product of law enforcement decisions. The rate of misdemeanor marijuana charge dismissals remain unchanged when considering all possible dismissals.Footnote 4 Lastly, the rate of plea bargaining declined, which we discuss in Sect. 4.2. We measure the robustness of these results with the application of a novel empirical strategy, synthetic difference-in-differences. When comparing our difference-in-difference estimates to our synthetic difference-in-difference estimates, we find the same magnitude and direction in our results across both methods. We discuss the underpinning methodological differences across the two methods extensively in the robustness analyses (Appendix A).

The extant behavioral literature has exclusively focused on the effect of non-binding guidelines or nudges on citizens’ compliance with formal and informal rules (Galbiati and Vertova, 2008; Romaniuc, 2016; Engel, 2021). However, we are not aware of any research that studied the impact of this type of nudging on the conduct of law enforcement agents. In this article, we study whether legal actors, such as police officers and prosecutors, can also be influenced by nudges in the form of non-binding guidelines. We focus on a particular type of non-binding guidelines that are low priority initiatives (LPI). LPIs are local mandates which state that police should make the enforcement of minor marijuana offenses the “lowest enforcement priority.” Within Los Angeles County, two jurisdictions adopted such initiatives in 2006–Santa Monica and West Hollywood. The LPI passed by city councils is a non-legally binding resolution. Given that there are no penalties for not complying with LPIs nor are police and prosecution actors forced to enforce LPIs, the implementation of such initiatives can be considered a nudging strategy used by city councils to influence the enforcement agents’ behavior.Footnote 1

We estimate the effect of the LPI in West Hollywood by measurement on the rate of misdemeanor marijuana charges, rate of misdemeanor marijuana cases dismissed, and the rate of misdemeanor marijuana cases plead out. First, we formalize our estimation strategy through a difference-in-difference design in Eq. 3, nearly identical to the estimation in Eq. 1, where our treatment group (Treat) is defined by charges brought forward in the city of West Hollywood and our treatment period (Post) is the time period after the LPI was initiated (July 1, 2006). This framework is similar to that estimated in our policing results, but differs in that our unit of analysis is at the city level (\(City_c\)) rather than the reporting district level (\(RD_i\)). Additionally, we adopt the most conservative definition of treated location as only containing the city of West Hollywood. This definition is pertinent as the district attorney’s office operates at the county level excluding a select few cities previously described. The unit of analysis between our analysis of police and prosecution differs due to the timing of criminal justice processes. To account for differences in the criminal case processing time, we aggregate the prosecution data at the weekly-level (\(Week_w\)). While arrest timing is precisely determined, the date that cases arrive at the prosecutor’s office are a function of a number of features (backlogs in law enforcement agencies, time required to gather evidence, etc.). As such, we aggregate our unit of analysis to the week level in examining the prosecution data in an attempt to overcome some of these issues.

There are various factors that could explain the legal actors’ decision to comply (or not) with locally enacted LPIs. The willingness to seek social approval in human beings has been shown to play an important role in our societies (Veblen, 1889; Elster, 1989; Bicchieri, 2005) and may be one factor explaining why police officers/management may comply with local LPIs.Footnote 2 In the case of police officers, they may seek approval of members of the community. Indeed, a City Council’s decision to enact an LPI is likely the expression of the preferences of its constituents. If police officers want to guarantee that their actions are in line with the local community’s preferences, as argued by Nix et al. (2020), then this may increase police’s compliance with the LPI. As shown in Rengifo et al. (2018) using survey data, the more individuals see the police as aligned with their own values, the more they are willing to support and cooperate with them. The Sheriff’s decision can also be affected by a desire to obtain approval from local communities, as this is an elected position.

With respect to policy implications, one major obstacle in modifying legislation to reflect the evolution of community beliefs is the slow and adversarial nature of legislative reform. Our analysis may introduce an alternative avenue when considering small-scale reforms in de-criminalizing non-violent low-level offenses. Though an important consideration to be aware of from our arrest analysis is that the community concerned with the low-level misdemeanor marijuana offenses appears to have experienced a reduction in arrests, this is not accurate. Rather, neighboring locations that did not pass a LPI experienced an increase in arrests, while regions passing the LPI experienced no increase in arrests. Thus, when the difference-in-difference analysis is performed, there is an appearance of a reduction in arrests in the LPI region even though there was no increase in arrests. This result shows that a non-binding guideline had significant implications for law enforcement behavior within the community of interest. But, law enforcement behavior does not seem to have been altered in regions where the non-binding guideline did not apply. With respect to the prosecutor’s office, the nature and quantity of cases across all cities encompassed in LA County makes it so that the LPI only played a significant role in reducing the volume of incoming arrests and in defendant’s optimal plea take-up rate strategy for a very small share of cases.

Columns I-III of Table 4 examines RDs in the control set that are located in contract cities, while columns IV-VI examines RDs in the control set that are located in unincorporated areas. Once again, we conduct a linear probability model to explore the effect of the LPI on misdemeanor marijuana arrests. The difference in effects between contract cities and unincorporated areas is quite stark. There is a small and weakly significant effect of the LPI on misdemeanor marijuana arrests when comparing West Hollywood to neighboring contract cities. However, in comparing West Hollywood to neighboring unincorporated areas, we find a large statistically significant reduction in misdemeanor marijuana arrests. Thus, it appears that the LPI had a differential impact on law enforcement behavior depending on the dynamics of the relationship between the LASD and the location where the RD was located. The LASD increased the number of misdemeanor marijuana arrests in both contract cities and unincorporated areas in the treatment period. However, the increase in misdemeanor marijuana arrests for contract cities is approximately 38.5 percent relative to the pre-period, whereas the unincorporated regions saw an increase of 107.1 percent relative to the pre-period. Thus, it appears that the LASD increased their arresting behavior in unincorporated regions, to a much larger extent, where no contractual relationship exists.

We estimate the effects of the LPI on review prosecutor initiated case dismissals and find an increase in the rate of these dismissals in West Hollywood post LPI (Appendix Table 8).

Recent research in behavioral law and economics has shown that behavior change can be achieved even when laws are backed by non-deterrent sanctions (McAdams, 2015; Ko, 2018). For example, Convery et al. (2007) showed that the legal requirement to charge small fees for plastic bags led to a reduction in use on the order of 90% and an associated gain in the form of reduced littering and negative landscape effects. Such a legal requirement that comes with only minor changes in incentives is similar to a nudge that expresses what is acceptable and unacceptable behavior. There is experimental evidence from the laboratory that this type of nudging can increase individual contributions to public goods. For example, Feld and Tyran (2006) found that experimenter guidelines backed by non-deterrent sanctions (i.e., when the experimenter states what is the right thing to do and deviations are mildly sanctioned) can increase individual contributions to a public good. Such guidelines are non-deterrent or non-binding because they do not change the individual’s dominant strategy, which remains free-riding. Thus, non-binding guidelines have the same properties as nudges: they do not significantly change one’s monetary incentives and do not entail any obligation to behave in a specific way (Sunstein, 2012; Ambrosino et al., 2018).

If we find a comparative reduction in misdemeanor marijuana arrests in West Hollywood to other reporting districts, we would expect there to be mechanically fewer cases charged in West Hollywood compared to other cities.

To study whether law enforcement is influenced by non-binding guidelines, we examine two parts of the criminal justice system in our analysis. The first data set utilizes incident-level arrests from the Los Angeles Sheriff’s Department (LASD). The second data set utilizes prosecution data obtained from the California Department of Justice, which contains information about the initial charging decisions and final disposition of all charges brought against a citizen. Prosecution is a unique setting to study the impact of non-binding guidelines on system actors’ behavior since there are no contractual relationships between a particular municipality and the prosecutor’s office. Therefore, the prosecutor’s behavior can only be impacted indirectly. First, prosecutors could choose to follow the non-binding guidelines by complying with LPIs, thus reducing the number of charges and increase dismissal/pleas after the implementation of an LPI. Second, there may be fewer cases arriving at the DA’s office if officers are making fewer marijuana arrests. Furthermore, there may be a strategic reason for prosecutors to adapt their behavior to the LPI. Defense-attorneys might discourage their clients from taking a plea bargain because a jury of their peers is more likely to find them not guilty knowing that there is an LPI in place. If prosecutors care about their success rates at trial, then prosecutors should be less likely to charge if the arrestee is from an LPI-area and more likely to push toward a plea.

Given that the law remained the same after the implementation of an LPI, police agents may have incentives to ignore the non-binding guidelines and continue to enforce formal laws. Finally, the threat from the municipality to provide police services internally if the Sheriff’s department does not comply with an LPI may be perceived as non-credible.

In the CORI data, case dispositions are organized by various disposition codes. For this analysis, we define charges with codes associated with cases that result in conviction or dismissal. Dismissals are defined as cases resulting in a review prosecutor dismissal, court dismissal, and dismissal for various reasons such as “furtherance of justice” and pleas are identified by a regular expression extraction detecting the phrase “PLEA” in the offense disposition description for an individual’s case.

bcpd.evidence.com login

The coefficient of interest (\(\beta _3\)) is our difference-in-difference estimate that explains the effect of the LPI on misdemeanor marijuana arrests. To ensure that we are not falsely attributing unobserved time-invariant variation in RDs, we include fixed effects for RDs (\(\delta _{i}\)) in our analysis. We also include yearly time controls (\(\gamma _{t}\)) to absorb time-varying unobserved variation across all RDs. Lastly, we include controls for suspect gender and racial demographics, which include whether the suspect was identified as male, White, Black, Hispanic, or Asian.Footnote 16 Undoubtedly these controls act as a subset of observed factors associated with the variation in misdemeanor marijuana arrests. We believe gender and racial composition to be two of the most predictive and endogenous attributes associated with low-level unconcealed drug arrests, which have been identified in the literature (Parker and Maggard, 2005; Gaston, 2019; Mitchell and Caudy, 2015; Mauer et al., 1999). In particular, our goal in adding controls and fixed effects in the model is to better isolate the effect of LPI on low-level misdemeanor marijuana arrests by accounting for other observed variables that contribute to this relationship. While we include exhaustive time, location and demographic controls, it is still possible that unobserved factors could influence our results. For instance, if LADA’s office is short staffed leading to less emphasis being made on low-level offenses at the same time as the passage of the LPI, we would not be able to capture the effects of this hypothetical reality on our estimations, leading to misattribution of the treatment effects.

Our primary outcome variable is whether an arrest involved a misdemeanor marijuana arrest.Footnote 15 We will analyze a standard difference-in-difference framework where we define the treatment group (Treat) as those arrests occurring within West Hollywood and our control group as arrests occurring in RDs that are partially contained in West Hollywood, but where the actual arrest did not occur in West Hollywood. Lastly, we will define the treatment period (Post) as the time period after the LPI was passed (July 1, 2006). The main specification that we utilize is described in Eq. 1:

Misdemeanor Marijuana Plea Rate. Fig. 6 showcases event-study coefficients on misdemeanor marijuana plea rates across months pre- and post- the LPI. We omit the month prior to the LPI and 15 months post as our reference omitted time groups. Standard errors are clustered at the city-level

DeAngelo, G., & Owens, E. G. (2017). Learning the ropes: General experience, task-specific experience, and the output of police officers. Journal of Economic Behavior & Organization, 142, 368–377.

When upgrading or replacing public-safety technology,  it is vital that the selected network, system, or application is flexible, scalable, and capable of meeting your agency’s current and future needs. Mission Critical Partners (MCP) supports this goal by applying big-picture thinking, breaking down barriers, navigating issues, and applying a vendor-neutral approach that results in the implementation of best-of-breed solutions.

Table 5 depicts summary statistics for the various rates considered for the full sample, West Hollywood and non-West Hollywood cities. Column 1 presents unconditional averages prior to the initiative, column 2 presents the sample average after the initiative and column 3 formally tests the difference in pre- and post- initiative means with a t-test. We observe statistically significant differences in all three panels in the Full Sample and non-West Hollywood cities, while West Hollywood experiences significant differences for charge and dismissals rates. Namely, we observe an increase in charge and dismissal rates after the intervention in both the full sample and non-west Hollywood cities when testing the unconditional means. In the third panel, we observe the average rate of cases plead out to be larger prior to the intervention relative to after intervention in both the full sample and non-West Hollywood cities. In West Hollywood we observe statistically significant differences in means for charge and dismissal rates before and after the initiative, but we do not observe a statistically different average plea rate. To further examine the LPI, we address potential caveats in relying on a simple difference in means by estimating an OLS model across cities and weeks pre- and post-initiative to better capture unobserved variation in prosecutorial behavior over time.

Mauer, M., Potler, C., & Wolf, R. (1999). Gender and justice: Women, drugs, and sentencing policy. Sentencing Project Washington.

However, despite the approval-seeking factor that may explain law enforcement agents’ compliance with low priority initiatives, adopting new enforcement practices may not seem natural to many police agents. First, habits are difficult to change even in professional contexts such as law enforcement (DeAngelo and Owens, 2017). Second, as discussed in Dharmapala et al. (2016), police agents may have a special taste for punishment, which could provoke resistance from law enforcement agents if they are explicitly mandated (but not by law) to not enforce specific laws. In effect, police agents may have intrinsic preferences for punishing wrongdoers and therefore derive direct utility from enforcing laws. Dharmapala et al. (2016) propose a theoretical model where police agents derive pecuniary (salary) and non-pecuniary (a personal satisfaction from punishing wrongdoers) utility from their job. Their model identifies circumstances in which “punitive” individuals will self-select into law enforcement jobs that offer the opportunity to punish wrongdoers. According to the authors, such “punitive” individuals will accept a lower salary but will create the need to provide suspects with strong criminal procedure protections. While Dharmapala et al. (2016)’s work is theoretical and assumes the self-selection into police of “punitive” individuals, there is empirical work supporting their claim. Dickinson et al. (2015) used a subject pool that includes 87 French police commissioners as well as non-police subjects. Police and non-police subjects participated in experimental economics games that measured people’s preferences to punish antisocial behavior. In their experiment, punishment was costly to the punisher and came with no material benefits. The authors found that police subjects were willing to incur greater costs to impose punishment than non-police subjects. However, their experiment did not disentangle a self-selection effect (“punitive” individuals are attracted into police) from a police training effect. Friebel et al. (2019) showed experimentally that “punitive” individuals self-select into police. They used as their police subjects high school students who had applied to join the police forces but who had no police training and compared their punishing behavior to non-applicant high school students. They found that high school students who had applied to enter the police were willing to incur greater costs to punish antisocial behavior than non-applicant high school subjects.

Ross and Walker (2017) and DeAngelo et al. (2018) studied how the introduction of LPIs influenced law enforcement agents’ behavior in the areas concerned by the LPI and outside of these areas. However, they do not include prosecutors in their analysis. Furthermore, their research does not connect LPIs with the behavioral law and economics literature investigating the effect of non-binding guidelines. Our paper fills these two gaps.