Urban Patterns of Police Misconduct
One month, three cities, many cries of distress
Nightingale authors come from varied disciplines; I’ve seen articles by visualization scientists, sociologists, public health experts, applied ethicists, geographers, etc. This report and its accompanying infographics reflect a Venn diagram of many of those disciplines. The arrival of the trial of Mr. Derek Chauvin this month — in the city where I teach and write — demands an offer be made to Nightingale’s readership of a timely and wide-angle view of overly aggressive police arrests in America.
Crime, from the Latin cri-men, translates as a “cry of distress.” (Silver, 2011). Distress is experienced by all crime stakeholders, including offenders, victims, and law enforcement. This report is both an academic-style “research note” and illustrated essay. It is quantitative, yet fundamentally human-centered, whereby I looked at 648 incidents of crime during August 2019 occurring in the three U.S. big cities of Indianapolis, Baltimore, and Dallas. August often is the most crime-laden of all months in any given year. (Crime in the United States, 1992). Given the tragic death of Mr. George P. Floyd, Jr. (1973–2020), it felt important to me to use a techno-humanistic framework to look at key data. In spirit, it was timely:
“Humanism is defined as the liberation cry of humanity, voiced at times and places where the integrity of life or thought was threatened or compromised, or when fresh horizons beckon.” –Anne Buttimer (Buttimer, 1990)
I sourced open data as posted on city portals to conduct exploratory data analysis (EDA — popular today with data scientists and other disciplines [Blitz, 2017]) — using the variables of police use-of-force (UOF) and racial plurality of urban stakeholders, including police officers and city residents. Some of the latter may well have suffered from “cries of distress” as recipients of overly aggressive force when detained, arrested, injured, and/or hospitalized for their crimes. From a statistical perspective output from EDA was created using traditional scatterplots along with a novel and complex visualization tool known as parallel coordinates. (Kosara, 2010).
Goals
Mr. Floyd’s death, due to ostensibly murderous UOF, was one of the greatest “cries of distress” witnessed so far this century. We have seen hundreds of specific proposals surface–from all corners of America–whose objective was to remedy what happened.
On the other hand, some degree of humanism, perhaps best characterized as “data honesty,” was evident by select police leadership, given good availability of open data for the cities that I studied. In order to ethically balance such factors I set my goals as three: first, conduct comparative EDA with a modest sample of two stakeholder groups, police officers and residents. Second, given action items based upon EDA findings, propose my own concept for potential redress of UOF-centered police misconduct if it were assessed as such. Third, keep my report as compact as feasible for Nightingale readers.
To achieve brevity, I created “infographic-style” broadsheets — those found these days in newspapers and online — to deliver all visualization findings (Corum, 2013). My assumption was that judicious use of infographics would be a compact and compelling means to help guide urban policy-makers and social-justice humanists alike.
Methods
1. I downloaded, consolidated, scrubbed, and formatted data using MS Excel.
2. Next, I imported city data into GeoDa software. I used GeoDa’s bivariate scatterplot tool to calculate descriptive statistics, notably R-squares
3. I then performed multivariate spatial analysis on the data, again with GeoDa. I used parallel coordinate plotting that was supported by more descriptive statistics, this time with calculations of standard deviation; I geovisualized data by pairing maps with companion plots.
4. An online portal — Justice Maps — afforded me the ability to create maps of census demography of each city’s racial plurality that I then visually compared to previous plots and maps.
Results of my methodology required that I gather output from each of the above four steps and enter it into a corresponding infographic that displayed constellations of maps, plots, icons, and captions, with a few key “call-outs.” I refer to infographic-style figures hereafter as “IG.”
Results
Cases of crime incidents (per city) ranged from 169 to 277. Think of each “case” being a resident reported by police to have been subjected to an encounter that resulted in some kind of UOF. The grand total of cases (or observations) was 648. In order to function as spatial variables — and geographical “units of analysis” — I used police districts (in Indianapolis and Baltimore) and council districts (in Dallas), ranging from 5 to 14 per city, for a grand total of 28 districts.
IG1 displayed data that I downloaded and consolidated as follows: 1) I renamed some columns, 2) I revised data attributes of columns (usually to a numbers category), and 3) I performed some scrubbing of rows that were missing appropriate spatial geometry.
The city of Dallas was unique due to its somewhat equal “three-part” racial demography (Hispanic, African-American, Caucasian). In fact, “Bloomberg news service rank[ed] Dallas-Fort Worth … among the most racially diverse metro areas on the country” (Egan, 2018). Given its complex racial plurality, the treatment of Dallas’ open data demanded a simpler path for analysis. Accordingly, I consolidated variables for all stakeholders (i.e., residents and officers) into two categories, those “of color” and those “Caucasian.”
Variables associated with police UOF required that I create working definitions. “Aggressive” UOF, of course, is not a “standard label” for police report spreadsheets across U.S. cities. I learned that clearly-defined descriptive metadata (for their associated open data) derived from most police reports that I had examined were substandard.
The core Indianapolis label for UOF was “Physical-Weight-Leverage” (coded P-W-L in the uppermost table in IG1), the department’s most frequent such arrest behavior. A clear or useful definition of Physical-Weight-Leverage was missing from police metadata. My call to the police department for this definition was unsuccessful. I thus had to make “word-for-word” assumptions, notably how such a phrase might (interpretively) read. In order to justify the attribution “aggressive UOF” I composed following: “the act of an officer applying one’s body weight in making a resident’s arrest, behavior that might include significant physical leveraging — adjacent to or atop — a resident’s body or part/s by and with an officer’s body or part/s (e.g., officer knee/s applied with weight to one’s back or neck).” Data counts in the P-W-L’s 169 cases therefore were labeled with the surrogate phrase “aggressive UOF.” The Baltimore UOF labeling language that could be judged most aggressive was “Forceable takedowns” and “Takedowns with injuries.” Data counts for these — when aggregated — totaled 277, then becoming Baltimore’s surrogate for aggressive UOF. The assumptions made about aggressive UOF in Dallas were based upon language that included “Held suspect down, Takedown body, Feet/leg/elbow/knee/strike, OC spray (oleoresin capsicum or pepper spray), and Taser.” Counts for these incidents — when aggregated — totaled 202 and became Dallas’ surrogate for aggressive UOF.
As IG2 (and later ones) will show, EDA is at core “spatial detective work:”
“As the name suggests, you’re exploring — looking for clues. You’re teasing out trends and patterns, as well as deviations from the model, outliers, and unexpected results, using quantitative and visual methods. What you find out…will help you decide the questions to ask, the research areas to explore and, generally, the next steps to take…In this way, your Exploratory Data Analysis is your detective work.” (Blitz, 2017).
In IG2 I examined bivariate data using scatterplots. My R-squared statistical calculations displayed findings between key variables as paired. All scatterplots with ≥ 0.6 R-squares were shown, they numbered seven.
The R-squares in IG2 are clues that helped explain ≥ 60% of variation. Scanning and comparing scatterplots in IG2 that featured nearly 650 observations were suggestive of police misconduct in all three cities I studied. Misconduct could have been due to overly dire UOF, it could have racist underpinnings, or it could be both. Given high to moderate “explainable variation,” some edgy questions of the presence or absence of social justice need asking, the most compelling of which would lead to hypotheses. Framing tangible, scientific hypotheses resides in the domain of confirmatory data analysis (CDA), something I describe soon.
IG3 featured a half-dozen pairings of maps and plots. As such, they are what visualization experts refer to as “small multiples,” something that—within my oversized infographic — ensure the reader can make rapid visual comparisons with ease. Plots with colored profiles (red for Indianapolis, green for Baltimore, blue for Dallas) fell to the far right in some plot fields while others were more centered. You can see how patterns emerged according to multivariate “rules of thumb:” right-leaning profiles, whose standard deviations exceed +1, tell a dramatic story, their broken lines often were metaphorically akin to possible “broken bones” of urban residents of color. In contrast were the profiles that centered on 0 and/or shifted towards -1 standard deviation. Their story frequently revealed Caucasian residents who, once arrested or detained in the districts shown, appeared to “dodge” aggressive UOF by officers.
I examined geodemography by way of maps created using U.S. census data for each city. In IG4 you can view “racial-plurality” maps (left side) with corresponding parallel coordinate plots (right-side), three of which I selected to offer intrinsic insights. Findings continue to “speak for themselves:” I used captions in this IG to point out the presence of some high standard deviations.
In general, the Eastside police district (Indianapolis), the Northeast patrol district (Baltimore), and the horseshoe-shaped community district #7 (Dallas) were each the locus of very high counts of aggressive UOF force incidents. Why so many incidents in these “hot-spot” districts, I asked myself? If standard deviations tell a reader anything then racist comportment by officers could have been — at least in part — a credible explanatory factor.
It is here that CDA, something of an “older sibling” to EDA that I mentioned previously, could now be a logical, next step. (Blitz, 2017). Its strength is bringing hypotheses to the surface for more-seasoned data analysis using stouter statistics. For example, analysts in that vein might pose something that reads as follows:
A “null” hypothesis: H0
There is no difference between aggressive UOF by police when delivered to arrested residents of color based on Caucasian race of officers within big city populations.An “alternative” hypothesis: H1
Caucasian police officers deliver aggressive UOF to residents of color upon arrest with ≥ 70% variance at p > .05 within a sample population of three U.S. big cities.
Responsible and Holistic Action
I offer two quotations — each one apart from the other by 30 years and by the native disciplines of their authors — yet amazingly alike in their humanistic spirit of wording:
“The Renaissance of humanism calls for an ecumenical rather than a separatist spirit: it calls for excellence in special fields as well as a concern for the whole picture. It beckons sensitivity to what the ‘barbarism’ of our own times might be. And challenges all to seek ways to heal or overcome … in responsible action …”
– Dr. Anne Buttimer, geographer (Buttimer, 1990)
“In these troubling times, a sociology of liberation rooted in empirical observation and theorizing from data rather than ideology is overdue. This sociology is realizable through systematic study and rigorous reasoning in the scholarly tradition pioneered by W. E. B. Du Bois [that] also embraces the idea that our intellectual habits needs interrogation and constant rethinking to generate new insights into an increasingly complex world.”
– Dr. Aldon Morris, sociologist (Morris, 2021)
Though they are social scientists not data visualizers, these quoted luminaries point to a compelling need for both humanistic, responsible action and constant interrogation in the face of our culture’s current “barbaric” zeitgeist.
One specific path to responsible action that I propose is the development of a “social justice sabbatical.” I proffer putting jobs and research on hold in order to substitute that with a huge “mission-centered” cause.
Plenty of scientists use EDA, some include data scientists. The Bureau of Labor Statistics described working data scientists as numbering 60,000. (U.S. Bureau of Labor Statistics, 2021). My proposal is in three steps:
- First, halve the above number to account for those professionals whose visualization/data skills play a central role in their employment; a plausible revision thus might be 30,000.
- Next, let’s conceive of a scenario wherein employers release these data scientists for a one-month “sabbatical” in order to permit them to use tools such as EDA (or a preferred methodology that best fits their skillset) to conduct serious and focused research on the matter of visualizing systemic racism, police brutality, their implications, and their redress. One month is 160 work-hours.
- Lastly, 30,000 x 160 is 4,800,000 person-hours of analysis and writing, the goal of which herewith is to offer scholarly, but tangible, recommendations pertaining to mitigating misconduct and possible racism along with healing communities of color.
I assumed that such an effort had not heretofore happened much. It turns out that it has surfaced–just recently–upon the arrival of COVID-19:
“With a shared vision, a diverse team of professors, postdoctoral researchers, engineers, graduate students and volunteers all willingly put their personal and professional projects on hold to do something that had never been done before. My hope … is that we build on this cooperation, increase funding for the type of fundamental science that made our work possible, and take a science-first leadership approach to stay ahead of future threats and safeguard our health.” (Doudna, 2021).
Acclaims of hope need to eclipse cries of distress. If and when pursued, this proposal might generate nearly five million person-hours to achieve a holistic goal, the timely and innovative investment in scholarly social activism. More so, if other sibling disciplines (e.g., the ones I mentioned previously, sociologists, public health experts, geographers [I’m one of the latter], etc.) were also asked to come on board. The results might lead to a new kind of “citizen’s arrest,” already hundreds of years in the making, arresting and redressing both law enforcement injustice and systemic racism.
Notes
Whereas no review of literature was conducted as part of this research report I’ll note that I prepared a lengthy, annotated bibliography on the subject of police UOF and published it online (in press, Byrne, 2020, pp. 16–21).
Standard deviation clarification for the statistics-minded reader: to account for differences in the range of values between variables the data are shown rescaled (mean = 0, standard deviation = 1).
Replication data: files created for use in this analysis are housed at Figshare online. Indianapolis and Baltimore data here: https://doi.org/10.6084/m9.figshare.13087235.v2
Dallas data here: https://doi.org/10.6084/m9.figshare.13641260.v1
All hyperlinks were last visited on April 4, 2021.
References
Blitz, Shelby. (2017). “Exploratory and Confirmatory Analysis: What’s the Difference?,” Sisense Data and Analytics Platform. Online, https://www.sisense.com/blog/exploratory-confirmatory-analysis-whats-difference/
Byrne, J. Kevin. (2020). “Open Data Visualized From Police Spreadsheets: A Case Study of Where Force, Race, and Place Collided,” 2021 Proceedings of the 52nd Annual Conference of the Environmental Design Research Association, Just Environments: Transdisciplinary Border Crossings, in press. In press, online, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3775925
Buttimer, Anne. (1990). “Geography, Humanism and Global Concern,” Annals of the Association of American Geographers, 80 (1), pp. 1–33.
Corum, Jonathan. (2013). “Storytelling with Data,” February 28, Tapestry Blog. Online, http://style.org/tapestry/
Crime in the United States 1991. (1992). Table 2.1 (p. 6). Washington, D.C.: U.S. Department of Justice.
Doudna, Jennifer. (2021). “The Power of Mission Driven Science,” Wall Street Journal, March 20.
Egan, John. (2018). “This is how racial diversity in Dallas compares to other U.S. cities,” CultureMap, March 6. Online, https://dallas.culturemap.com/news/city-life/03-06-18-dfw-racial-diversity-comparison-census-bloomberg/
GeoDa. (2020). “GeoDa: An Introduction to Spatial Data Analysis,” Center for Spatial Data Science, University of Chicago. Online, https://spatial.uchicago.edu/geoda
Kosara, Robert. (2010). “Parallel Coordinates,” Eager Eyes, May 13. Online, https://eagereyes.org/techniques/parallel-coordinates
Morris, Aldon. (2021). “Emancipatory Sociology: Rising to the Du Boisian Challenge,” 116th Virtual Annual Meeting of the American Sociological Association. Online, https://www.asanet.org/annual-meeting-2021/theme-program-committee
Silver, Lisa. (2011). “Let’s Talk About: The Word ‘Crime,’” Ideablawg, December 29. Online, https://www.ideablawg.ca/blog/2011/12/29/lets-talk-about-the-word-crime.html
U.S. Bureau of Labor Statistics. (2021). Online,
https://www.bls.gov/oes/current/oes152098.htm
Acknowledgments
This research note needs proper acknowledgment as it is a continuation of matters covered in a previous paper of mine (for the Environmental Design Research Association, to be published in the EDRA52 Detroit Conference Proceedings in 2021) and my upcoming presentations (at EDRA52, along with the Annals of the Association of Geographers’ annual conference by way of its SIG — “Towards Computational Praxis for Social Justice”). I enlarged the previous scope in two ways: 1) by offering — as a more ethical umbrella and context — a humanistic open data mindset, and 2) by the inclusion of still more troubling findings gleaned from another deep data dive, this time into the city of Dallas (Texas). I give “permission kudos” to the editor and reviewers at EDRA along with folks at the AAG’s SIG.
Technical language assistance courtesy Dr. Julia Koschinsky. Editing assistance courtesy Kathy Heuer and Mary Aviles. Errors that might remain are mine and mine alone.
This note’s narrative matter and its figures are assigned Creative Commons cc 2021.
J. Kevin Byrne (MA/Minnesota, MFA/Cranbrook, MSc-Cert./Saint Mary’s) is Professor (now Emeritus) at the Minneapolis College of Art and Design (MN/USA). He has published in print and continues to do so online. Byrne’s current roles are those of an urban cartographer, designer, and spatial information analyst interested in mapping a hopeful future for civil rights, a future that must end what scholar Philip A. Goff recently referred to as the “racial terrorism” of today’s policing.
