How do police compare demographically & politically to civilians they serve? We investigate w/ data on officers from 97 of the 100 largest US agencies, more than ⅓ of local police. We also test whether Dem and Rep officers behave differently in Chicago. 🧵scholar.princeton.edu/sites/default/… Theories of “representative bureaucracy” hold agencies function better when staffed by people who share common experiences/values w/ citizens. Most studies examining this idea in policing focus on race. We set out to compare officers to civilians on a host of dimensions. 2/n

Seems like a great day for an IV thread!

Here's a simulated case inspired by Angrist, Imbens & Rubin (1996) showing how our new automatic causal bounding approach can alert users when a key assumption (e.g. "no defiers") is violated. 🧵

https://twitter.com/jonmummolo/status/1443215691721674764

Imagine an encouragement Z for a treatment X that causes an outcome Y (for simplicity, suppose everything is binary). The goal of IV is to address confounding of X & Y by an unobserved U. This confounding means the average treatment effect (ATE) of X on Y is not identified. 2/

Our new working paper: tell the computer what you know (and don't know!) about a causal question w/ discrete data → automatically get most precise possible answer (bounds, or a point estimate). Joint w/ @guilhermejd1 @nsfinkelstein @dean_c_knox Shpitser.🧵arxiv.org/abs/2109.13471 Scholars have developed many causal research designs (SOO, IV, DiD, RD, RCTs, etc.). All rely on bundles of assumptions. But applied settings are messy. What if an assumption fails? Most common choices: give up, or ignore the problem. 2/

Our team, Research on Policing Reform and Accountability (RoPRA), analyzed 1000s of civilian complaints against Philly police. In new @PHLPAC report we show complaints almost never lead to discipline. @bocar_a @dean_c_knox R.Mariman @MAranzazuRU 1/n phila.gov/media/20210520… The vast majority of cases don’t advance beyond initial Internal Affairs investigation. Even when IA finds policy violation, most misconduct results only in training/counseling, not discipline. Overall, only 0.5% of civilian complaints result in anything beyond a reprimand. 2/n

In a new @ScienceMagazine study, we find Black, Hispanic & female officers engage in less enforcement and violence than white & male officers facing common circumstances. @bocar_a @dean_c_knox @romangrivera1 1/ science.sciencemag.org/cgi/doi/10.112… The central hurdle in evaluating whether officers from different groups treat civilians differently has been accounting for context. Without the ability to compare officers facing common circumstances, any differences in their behavior become ambiguous. 2/

Early studies of disbursement of military gear to local police claimed to show public safety benefits. Two new articles in @NatureHumBehav show when flaws in a primary data source are addressed, evidence for those benefits vanishes. My commentary here: nature.com/articles/s4156… thanks to @tom_s_clark @milo_phd @annagunderson
@elishaacohen @kaylynjackson Adam Glynn & Ken Lowande for very valuable contributions to this literature.
Great examples of what can be learned when we don't take data at face value. nature.com/articles/s4156…

nature.com/articles/s4156…

Post-treatment selection bias can distort studies of police violence. GCBCSH proposes new stat theory for causal identification, despite selection & confounding—a huge breakthrough, if credible. We formally assess it.1/n @dean_c_knox @conjugateprior Paper: dropbox.com/s/nx8pe8gmw41d… Studies of racial bias in policing (e.g. Fryer 2019) often rely on detainment data (stops/arrests) to estimate discrimination in subsequent actions, e.g. use of force. If racial bias affects detainment, @sndurlauf & Heckman call this a "classic route to selection bias."

Our paper shows traditional analyses understate racial bias in police violence. A newly posted critique claims those approaches work great (if we assume away the problem). Given the bad science going around on the topic: thread 5harad.com/papers/post-tr… Our paper asks: given what we know about police-civilian interactions, how can we estimate racial bias using detainment records (stops/arrests) alone? We identify minimal assumptions, and show racial bias can be bounded using only data on stops. @dean_c_knox @conjugateprior

New @The_JOP by @seanjwestwood @SolomonMg @ylelkes shows probabilistic election forecasts like @FiveThirtyEight confuse voters & decrease turnout, mostly among Dems. It’s thorough and innovative experimental behavioral research.

A fan thread. (1/n) Unlike polls that show candidates' expected vote share, prob. election forecasts convey the estimated probability that a candidate will win. Problem: folks don’t understand probabilities. This paper demonstrates severity of this confusion, and its political consequences. (2/n)

@PNASNews published a study last year claiming no racial bias in police shootings. The study's central claim was mathematically unsupported. @dean_c_knox & I submitted critique to PNAS, which was rejected. We appealed. Today PNAS published our critique.1/n pnas.org/content/117/3/… The original study by Johnson et al. claimed to find “White officers are not more likely to shoot minority civilians than non-White officers.” i.e., the fatal shooting rate of minorities by white officers is ≤ that of nonwhite officers. 2/n

Collaborative experiments w/ gov. agencies are the gold standard in policy research, but are collaborating agencies representative? New working paper evaluates selection into research partnerships w/ police. Seeking feedback! 1/n @smgoerger @seanjwestwood papers.ssrn.com/sol3/papers.cf… We sent ~3,000 local police chiefs/sheriffs in 48 states *sincere* invitations to discuss a potential collaboration (no deception). Two goals. 1) Assess the correlates of willingness to collaborate. 2) Assess the role of agency reputation in collaboration decisions. 2/n

Brett Stephens is crediting "Broken Windows" policing strategies like "Stop, Question and Frisk" (SQF) for the nationwide crime drop that began in the 1990s. Let's take a look at the evidence. 1/n

https://twitter.com/prageru/status/1177209958493913089

It is true that Broken Windows expanded in the mid 90s in NYC and crime also fell in that period. But crime also fell in cities across the country that did not embrace Broken Windows. 2/n

New study of fatal police-involved shootings reported “no overall evidence of anti-Black ... disparities.” @dean_c_knox & I submitted a letter to the editor showing study's approach can't support claim. Journal declined to publish. Sharing here so future research can improve. 1/N The study claims its approach “sidesteps the benchmark debate”---the problem of picking a baseline to use to evaluate shooting rates across racial groups. We show this is not true.The study implicitly and wrong assumes black/white civilians encounter police in equal numbers.2/N

Estimates of racial bias using police data are wrong if police discriminate in who they stop. New paper w/ @dean_c_knox, @conjugateprior: analysis in Fryer (forthcoming) likely masks hundreds of thousands of instances of discriminatory police violence. 1/n goo.gl/yfPM5r The study of racial bias in policing is a causal inference problem, but prior work rarely makes causal estimands and attendant assumptions explicit, making analyses difficult to evaluate. We formalize the problem using principal stratification in a causal mediation framework.

My latest research out today in @PNASNews. “Militarization fails to enhance police safety or reduce crime but may harm police reputation.” (thread)

Article (open access): pnas.org/content/early/…

[Photo credit: Shutterstock/JPL Designs] Four yrs ago this month, coverage of the heavily armed police response to protests in Ferguson, MO fueled a national debate re: police militarization. Police claim militarized units enhance public/officer safety. Critics claim they target racial minorities/erode trust in police.