MARIJUANA REFORM & RACIAL/ETHNIC DISPARITIES IN ARRESTS: 2012-2019


Final Projects, Visualization

Introduction

For decades, the punitive policies of the War on Drugs have disproportionately impacted Black and Brown Americans. With mandatory minimum criminal drug sentencing contributing to mass incarceration and to the barring individuals with criminal records from accessing employment, housing, and student loans, the War on Drugs has altered the course of millions of Americans lives.

Despite usage rates across races and ethnicity being similar, law enforcement has and continues to disproportionately affected communities of color. With over 700,000 thousand arrests made annually for marijuana-related offenses, states and municipalities have picked up efforts to reform marijuana laws and policies in an effort to reduce the racial and social disparities in their communities.

Marijuana reforms at the state and municipal level include decriminalization, legalization, and changes to enforcement policy. In this study, I will look at the impact of varying reform efforts in the last 8  years and the impact on the rates of racial disparities in cannabis arrests in three major US cities: Los Angeles, New York, and Washington DC.

Process & Design

Expanding upon my Charts and Graphs and Mapping labs, where I charted the rates of marijuana arrests against recent reform efforts and mapped arrests against census demographics in DC, I set out to build a series of small multiples to see what differences or similarities could be found between the cities.

An important departure in this project’s parameters from my previous work with DC’s arrest data is that I incorporated mutually exclusive race/ Hispanic origin data mindful of both Los Angeles and New York’s Hispanic populations and the data that reporting that both Black and Hispanic people are more likely to have numerous interactions with police than white Americans.

With DC complete, I needed out to find both marijuana arrest data with race/ethnicity and geolocation information as well as census shapefiles with complete demographic population information. Unable to meet all parameters in one dataset, I downloaded separate census datasets to merge from County of Los Angeles Open Data portal. Similarly, the New York census shapefile was missing the necessary demographic using the NYC Planning Population FactFinder, again referencing the mutually exclusive race/ Hispanic origin data.

With this demographic data, I the calculated a new percentage metric for each census tract, adding the Hispanic and Black non-Hispanic populations together and dividing by the whole population in each tract.  Next, I downloaded historic arrest data for both Los Angeles and New York, using Open Refine to filter out all arrest information unrelated to marijuana offenses since 2012. Both datasets included both race and ethnicity, that I similarly created a new metric for reporting race and ethnicity together with the following parameters: Hispanic (of any race), Black (non-Hispanic), White (non-Hispanic), and Other/Unknown.

I began the design process in CARTO, where I used the analysis function the add columns to my geospatial file from the demographic datasets, joining the Black and Hispanic population calculation to each census track. I shaded the polygons by percentage value, with the darker areas representing an area with a higher population of Black and Hispanic residents. Over the shaded census tracts, I imported the arrest data, coloring the points by the race and ethnicity column. I purposefully chose hues like red, blue, and yellow that would contrast from the gray background, optimizing pre-attentive visual processing. Then, to reduce overcrowding I used transparency settings and blending to reduce over plotting. Moreover, by applying the blending setting, viewers could observe areas in which both Black and Hispanics are both frequently arrested by the secondary color purple (Figure 1).

Figure 1. Los Angeles marijuana arrests colored by race and ethnicity using blended style and transparency to reduce over-plotting.

Once I had completed my maps in CARTO, I built my accompanying time-series charts in Tableau to compare rates of arrests and racial disparities in arrestees over the period. In order to compare the overall rates of arrests across the three cities, I needed to normalize the data by calculating the rate of arrest per 100,000 people. For simplicity, I calculated the rate per year in a separate Excel and imported to generate a series of line graphs that I annotated with the reform effort that marked the period. Next, I graphed the percentage of total arrests by race and ethnicity over the period ensuring that both the X and Y access reported on the same range to be comparable across cities.

Finally, I created a trellis, or series of small multiplies, with each row telling a story of one city and each column a series of directly comparable graphs, grouped by type. I ensured each X and Y axis for each graph, removing any duplication in reporting the time frame otherwise noted on the bottom X axis. Finally, to situate the series in its proper context I added both a caption and additional text to guide a user’s interpretation of each type of visualization from left to right (Figure 2).

Figure 2. Early draft of visualization including language to guide interpretation.

UX Research

Because my intended audience for this project is both criminal justice activists as well as the general public, it was important to me I both interview an expert in the field as well as a new user, who may not be as knowledgeable of the complexities of marijuana reform. With the new user interview, I focused on design, layout, and clarity of the visualization while focusing on the accuracy and breadth of the content presented.

UX Research – New User

For my non-expert user, I asked my colleague Kathryn Kelly, a Manager for Research and Strategic Partnerships in the Office of the Provost at Pratt Institute. With a background in writing and editing, Kathryn works with faculty to submit proposals for outside funding. It is this experience in coaching and communicating impact that I reached out for her feedback.

I began my interview with asking her to identify her familiarity with the topic of marijuana reform laws and policing policies as either ‘expert,’ ‘familiar,’ ‘somewhat familiar,’ or ‘unfamiliar.’ Kathryn, identified herself as somewhat familiar with marijuana reform laws and policies while clarifying she is familiar with the existing racial disparities in who is arrested.

Next, I asked her to think out loud and describe her thought process for reading and interpreting the visualization. Right off, Kathryn noted that as she started left to right in her interpretation, there was no legend or key to give meaning to the colors visible on the embedded CARTO map. In using the built-in map legend in CARTO, she lost all view of the map. She furthermore had a hard time orientating herself within the portal and it took her a long time to see that the dark shading of the polygons as connected to the percentage population of Black and Hispanic residents. Next, it was not clear to Kathryn right away that each row represented a city, highlighting a need for clearer labeling of the matrix Y axis.

However, considering how she read up and down the columns in addition to left to right, I got a sense that positioning the cities as rows, instead of columns, successfully invited the easy comparison between the type of graphs and series of graphs for each city. She read aloud the rate of arrests and annotations between each city noting how the rate and reform effort of each city varied, although all experienced a decrease in arrests. Finally, she had no problem in interpreting the significance of the final column noting the increase in percentage of total arrests for Black people, while whites dropped across all cities, regardless of reform effort.

Furthermore, in LA, a city with already a relatively low rate of arrests, disparities in arrests deepened with legalization, an interesting observation. However, when working through this, she noted how highlighting and brushing across each city could have aided her experience to more easily consider each city by itself and in relation to the others.

I closed the user experience interview by asking if she believed from the information presented if recent reform efforts were effective in addressing racial disparities, she reported a clear resounding ‘No,’ a confirmation that despite some big design flaws my goal was successfully communicated. Finally, I asked her what questions where provoked or left unanswered by the visualization and she asked how and why where people still arrested, seeking additional information on the arrests that were still occurring.

UX Research – Expert User

For my expert user experience interview, I interviewed a friend, attorney, and activist, Sarah Gersten, who serves as the Executive Director and General Counsel at the Last Prisoner Project, a nonprofit organization bringing restorative justice to the cannabis industry. Like my first meeting, I asked Sarah to self-identify her familiarity with the subject matter before walking through the visualization. An expert in the subject, Sarah not inhibited the reorientation of labels or legends to know exactly what she was looking at: racially targeted policing despite the various ‘reforms.’ As Sarah emphasized, arrests for cannabis continue to disproportionately black and brown Americans with the disparities rising without what she identified as “legalization, the right way.”  To understand more, I then utilized the rest of my time with Sarah to make the connections my visualization didn’t ye make asking, wat important information is missing to understand the rise in racial and ethnic disparities?

Beginning with the recent legalization efforts in Los Angeles, Sarah noted that the rise in arrests of black people could partially be attributed to the lack of equity to the legalized markets to either buy or sell, citing barriers to accessing capital to start a business as well as the high taxes on buying cannabis legally. Furthermore, a legal cannabis industry has a vested interest in eliminating ‘illicit’ market competitors.

Then in New York, Sarah highlighted that issuing tickets instead of arrests is not enough to reduce harm on black and brown communities who interact with police more, citing that multiple violations can be an alternate point of entry the system if you do not have means to pay the fine or are on probation. By not including summons in the visualization, I have not captured the full impact of cannabis related infractions on the black and brown communities.

In the end, Sarah used DC to highlight the ultimate tension overriding all efforts at reforming punitive laws and policing practices: despite local initiatives, marijuana remains illegal at a federal level. Congress rejected the cities’ plans for legalization with commercial regulations in 2019 and possession personal recreational use that is legalized by DC still is criminal on federal property. Also, as the rate of arrest began to increase with a change in federal administration, the ultimate success in any of these localized reforms still largely hinges on who is in power. Another factor not yet considered, I charting arrests with administrations on the X axis of historical arrests data could have offered rich insights.

Conclusions & Next Steps

Ultimately, the dashboard communicates that disparate and localized reform efforts are not enough to address the deep racial disparities in marijuana arrests, but could have the opposite effect until comprehensive ‘legalization, the right way.”  However, the visualization comes up short of its full potential for insights as it not fully interactive nor clear as it could have been, largely due to my use of CARTO.

Although my CARTO maps are successful as individual visualizations, highlighting across all three cities that arrests for marijuana related offenses fall along the edges of white and Black/Hispanic neighborhoods, they did not succeed in their integration into the Tableau dashboard for the purpose of small multiplies. Instead, they are hard to use, disorientating to users, and limiting for full interactivity of the dashboard. If I had been able to achieve the same effect in Tableau’s mapping, I could have linked all data to an interactive legend that could assist in filtering and brushing to be support exploration and interpretation. Although I did incorporate new labels on the Y axis of my trellis visualization and reorient my legend close to the data, there was no other major improvements I could quickly enact for my final dashboard without remapping in Tableau or repositioning content. Furthermore, because the cities selected for this dashboard were so varied in their reform efforts and history of cannabis enforcement, my conclusions can only be broad in scope, and are unable to capture the nuances of the how and why the reforms fell short of addressing racial disparities in arrests. Should I embark on the project again, I would have consulted an expert in the development stages to get a sense if the cities whose reform efforts aligned that could be cross compared for more in depth analysis. For example, it could have been fruitful to compare arrests in cities that otherwise have flouring legalized marijuana dispensaries. Overall, it is still a fruitful exercise, with the visualization highlighting that without comprehensive and equitable legalization, racially targeted arrests could continue to increase.

Explore the final version of the visualization here: