Racial Inequality and the Neighborhood Context of the ‘Overpolicing-Underpolicing Paradox’
> "*In 2014, Chicago police stopped, questioned, and often frisked more than 718,000 people. That was a big number for a city of 2.7 million. It was more than three times the stop* rate *in the place best known for its aggressive stop & frisk policy, New York City*" [@skogan2022stop, p. 1]
## The neighborhood context of the stop & frisk era in Chicago Stop & Frisk 2012--2015
> "*(...) during its peak year of 2014, 90 percent of those stopped were African Americans and Hispanics. From 2013 to 2015 more police enforcement actions were taken in Chicago's predominately African American neighborhoods than there were people living there, every year*" [@skogan2022stop, p. 2]
> "This process I refer to as the *overpolicing-underpolicing paradox*. Policing seemed to be a ubiquitous part of the lives of many of these marginalized young people; however, the law was rarely there to protect them when they encountered victimization." [@rios2011punished, p. 54]
> "I see cops so often in my neighborhood, but when I see something bad going on, I look around and say ‘where are the cops?" [18-year-old girl from Phildelphia interviewed by @carr2007we, pp. 458-459]
## What are the implications of the era of stop and frisk... $\leadsto$ ...for the neighborhood concentration of police use-of-force? $\leadsto$ ...for the *persistent* neighborhood concentration of civilian complaints about... ::: {style="margin-left: 1.5em;"} - ...*Abusive* policing? - ...*Neglectful* policing? ::: $\leadsto$ ...for people living in such over-policed neighborhoods? # Part 1 -- Neighborhood-level analysis ## Data and measures *Neighborhood-level data* ($N=865$ census tracts) ***[Main independent variable]{.underline}*** * Stop & Frisk records (2013--2015) > Counts of contact cards recorded by the CPD
## Data and measures ***[Main dependent variables]{.underline}*** * Use-of-force records (2013--2015) > Counts of Tactical Respose Records (TRR) recorded by the CPD
* Complaints about abusive policing (2012--2015 \& 2016--2018) > Counts of CPD complaints in which allegations were: "*use of force*", "*excessive force*", "*verbal abuse*", "*illegal search*", or "*false arrest*"
* Complaints about neglectful policing (2012--2015 \& 2016--2018) > Counts of CPD complaints in which the allegations were: "*neglect of duty*", "*inadequate / failure to provide service*", or "*slow / no response*"
## Data and measures ***[Main control variables]{.underline}*** $\leadsto$ Police-related variables * Police-recorded violent crimes (2010--2012)
> Includes previous records of homicides, robberies, and aggravated assault recorded by the CPD
$\leadsto$ Structural antecedents (US Census 2010) * Concentrated disadvantage * Residential stability * Immigrant concentration * Population density
## Estimation strategy Is the spatial concentration of stop & frisk encounters in 2013--2015 associated with more... :::: {style="display: flex; align-items: center; gap: 1em; margin: 0.5em 0;"} ::: {style="flex: 0 0 65%;"} **1)** ...use-of-force records in 2013--2015? ::: :::: :::: {style="display: flex; align-items: center; gap: 1em; margin: 0.5em 0;"} ::: {.fragment fragment-index=2 style="flex: 0 0 65%;"} **2)** ...complaints about abusive policing... ::: ::: {.fragment fragment-index=4 style="flex: 1;"} [→]{style="display: inline-block; transform: rotate(-35deg);"} ...in 2013--2015?\ [→]{style="display: inline-block; transform: rotate(35deg);"} ...in 2016--2018? ::: :::: :::: {style="display: flex; align-items: center; gap: 1em; margin: 0.5em 0;"} ::: {.fragment fragment-index=3 style="flex: 0 0 65%;"} **3)** ...complaints about neglectful policing... ::: ::: {.fragment fragment-index=4 style="flex: 1;"} [→]{style="display: inline-block; transform: rotate(-35deg);"} ...in 2013--2015?\ [→]{style="display: inline-block; transform: rotate(35deg);"} ...in 2016--2018? ::: :::: ::: {style="height: 0.35em;"} ::: $\leadsto$ Poisson SLX models: spatial lags of all covariates, Conley SEs^[Queen contiguity, row-standardised weights; Conley cutoff = 2 km. Linear SLX models with `log(y+1)` and `asinh(y)` outcomes yield substantively identical results.] ## The persistent neighborhood concentration of over- and under-policing {.r-stretch fig-align="center"} > 5x Poisson SLX models controlling for concentrated disadvantage, residential stability, immigrant concentration, population density, and spatial dependence. $N=846$.
# Part 2 -- Individual-level analysis ## Data and measures **Project on Human Development in Chicago Neighborhoods (PHDCN)** - 6200 children from 7 age cohorts, born 1978 to 1996 - 3 interviews from 1995--2003 - Representative of Chicago and its neighborhoods ::: {.fragment fragment-index=2;"} ### {width=200px height=150px} - 60% subsample of 4 cohorts - 1057 interviewed in 2012 - 682 followed-up in 2021 - Followed throughout the US
::: ## Analytic strategy 287 PHDCN+ respondents were living in Chicago from 2012 to 2015, the *era of stop & frisk* [@skogan2022stop] {.r-stretch fig-align="center"} ## Analytic strategy **Independent variable**: Cumulative neighborhood exposure to stop and frisk (2012--2015) > We define $D_{ijt}$ where $T \in \{1, 2, 3, 4\}$ as the number of stop and frisk encounters in neighborhood $j$ where respondent $i$ was residing during period $t$. Periods here refer to years 2012 ($t=1$) through 2015 ($t=5$). We then define cumulative neighborhood exposure to stop and frisk as $cum(D_{i})= \sum_{t=1}^{4}d_{ijt}$, i.e., the sum of all SQFs per individual across this four-year period
::: {.fragment fragment-index=2;"} **Dependent variable**: Gun carrying in adulthood (2021) > Have you ever carried a concealed or hidden gun with you anywhere?\ > 1: Yes, in the last 12 months\ > 0: No, not in the last 12 months
::: ## Estimation strategy Marginal structural models with residual balancing[@zhou2020residual]
> Neighborhood exposure to violent crime is a time-varying confounder, as SQFs and violent crime are dynamically related. Weights created using residual balancing permits appropriatelly controlling for time-varying confounders without introducing over-control bias.
::: {.fragment fragment-index=2;"} Other (pre-exposure) controls: * Gender | Cohort fixed effects | Previous offending behavior * Immigration status | Primary caregiver's education * CBCL internalizing scores | CBCL externalizing scores > The Child Behavior Checklist (CBCL) survey instrument was a parent-report questionnaire on which the child was rated on various behavioral and emotional problems.
::: $\leadsto$ Effects disaggregated by race: White, Black, Hispanic ## Long-term implications of the stop and frisk era for residents Probability of gun carrying in adulthood {.r-stretch fig-align="center"} > Marginal structural models estimated as Firth’s penalized linear probability models with weights created using residual balancing to control for dynamic exposure to neighborhood violence and robust Huber-White standard errors. $N=257$.
# Some concluding thoughts ## Some concluding thoughts In Chicago, aggressive policing tactics were widely used between 2012 and 2015 -- the *stop & frisk era* [@skogan2022stop] > "Confrontational proactive policing methods may have social benefits in crime reduction but costs in intrusion on the rights and privacy of innocent persons" [@manski2017assessing, p. 9308]
::: {.fragment fragment-index=2;"} * Neighborhoods with more SQFs are more likely to have more... * ...incidents of police use-of-force * ...complaints of abuse of police power * ...complaints of neglect and unresponsiveness $\leadsto$ These associations persist even after the end of the *stop & frisk era*
::: ## Some concluding thoughts > "One indicator that poor African Americans in particular have been **excluded from police resources** is that police repeatedly fail to respond, or to respond in a timely fashion, when they are called. (...). In response to **police abandonment**, marginalized people **seeking protection** or redress of grievances have generally turned to **'self-help.'**" [@bell2017police, p. 2116] $\leadsto$ Implications of the "overpolicing-underpolicing paradox" for gun-carrying behavior, even years later ## Some concluding thoughts > "(...) there are thee Chicagos, not one. The city's large White, African American, and Hispanic neighborhoods differ dramatically in character, including their distinctive crime problems and their relationship with the police. In important ways residents live in different worlds" [@skogan2022stop, p. 18]
* Long-term implications for gun carrying among Black and Latino residents of neighborhoods where SQF concentrated in 2012--2015 $\leadsto$ Not for White residents
# Thank you! ## References